Refresh all metrics, add reflex's launch post, raise the recency credit
Browse filesMetrics: re-read every item against X, GitHub (via the authenticated gh CLI,
5,000/hr instead of 60) and the Hub, 43 of 46 entries updated. The snapshot
moves to 2026-09-19 in the footer, README and BASE_DATE.
The refresh reshaped the top of the list: Laya went 163 -> 1,410 GitHub stars
and 108 -> 577 Hub likes (raw 1,679 -> 11,666), which lifted it above LocalJev
and dropped LocalJev from #3 to #5 at the old credit of 5000. The credit is
raised to 10000, which puts LocalJev back at #2 with a 926-point cushion over
#4 rather than the 51-point one a credit of 9000 would give -- a margin thin
enough that the next refresh would flip it back. Top 5 is now Laya, LocalJev,
45-second TL;DR, Qwen-2.5-1B-RLCD, openjev.com.
Note the engagement gate now saturates for 20 of 46 items, so for anything
with real traction the credit acts as a flat recency tier and only age
separates items. That is the intended behaviour here -- it is what keeps a
0-engagement entry off the list -- but it does mean the credit is no longer
"small", which was the original constraint.
reflex (Kshetrajna Raghavan, @kshetrajna ) was already listed; this adds its
launch post and WebGPU demo, a real description of what shipped since (a
/v1/systemone server, temperature scaling to 0.039 ECE against Jev's 0.031,
and an open LoRA recipe), its author name, and fresh metrics (330 likes,
138k views, 77 stars).
Also corrected typesafe-ai-benchmark to its actual tweet date (2026-09-17),
found by comparing every entry against its tweet's created_at.
README.md and og.html (plus the re-rendered og.png, now showing LocalJev at
#2) are updated to match.
Co-Authored-By: Claude Code <noreply@anthropic.com>
- README.md +2 -2
- index.html +50 -49
- og.html +50 -49
- og.png +2 -2
- refresh.js +9 -6
|
@@ -22,7 +22,7 @@ This static Space tracks the artifacts in one combined grid, color-coded by kind
|
|
| 22 |
- **Prior art**: "this already exists" claims
|
| 23 |
- **Explainers**: architecture speculation, explainers, benchmarks and roundups
|
| 24 |
|
| 25 |
-
Cards sort by a **trending** score: ♥ likes on X + 5 × GitHub stars + 8 × Hub likes + views ÷ 500, plus a **recency credit** (up to
|
| 26 |
|
| 27 |
…plus a section on what is still **not** in the open: TypeSafe's weights, the RLCD algorithm, and any open model matching Jev's calibration claims.
|
| 28 |
|
|
@@ -36,7 +36,7 @@ Hugging Face models and Spaces are first-class artifacts here, not only GitHub r
|
|
| 36 |
|
| 37 |
All data lives in one JS array (`ITEMS`) near the bottom of `index.html`. Open a PR on this Space to add or correct an entry. Please include the announcement post on X if there is one, and the Hub / GitHub link.
|
| 38 |
|
| 39 |
-
Engagement metrics are a snapshot (2026-09-
|
| 40 |
|
| 41 |
## Credits
|
| 42 |
|
|
|
|
| 22 |
- **Prior art**: "this already exists" claims
|
| 23 |
- **Explainers**: architecture speculation, explainers, benchmarks and roundups
|
| 24 |
|
| 25 |
+
Cards sort by a **trending** score: ♥ likes on X + 5 × GitHub stars + 8 × Hub likes + views ÷ 500, plus a **recency credit** (up to 10000, halving daily) so a fresh release is not buried under older, louder ones. The credit is scaled by the item's own engagement, so a brand-new entry nobody has reacted to yet cannot climb on its date alone. Cards added within a day of the metrics snapshot get a green **new** badge. Use the category and "has" chips to filter, or switch the sort.
|
| 26 |
|
| 27 |
…plus a section on what is still **not** in the open: TypeSafe's weights, the RLCD algorithm, and any open model matching Jev's calibration claims.
|
| 28 |
|
|
|
|
| 36 |
|
| 37 |
All data lives in one JS array (`ITEMS`) near the bottom of `index.html`. Open a PR on this Space to add or correct an entry. Please include the announcement post on X if there is one, and the Hub / GitHub link.
|
| 38 |
|
| 39 |
+
Engagement metrics are a snapshot (2026-09-19) from public post metadata and will drift.
|
| 40 |
|
| 41 |
## Credits
|
| 42 |
|
|
@@ -315,7 +315,7 @@
|
|
| 315 |
<div>
|
| 316 |
<p><b>Missing something?</b> Open a PR against this Space's <code class="mono">index.html</code> (the data is one JS array at the bottom of the file) or ping <a href="https://huggingface.co/multimodalart" target="_blank" rel="noopener">@multimodalart</a>. Hugging Face models and Spaces count as first-class artifacts here, not just GitHub repos.</p>
|
| 317 |
<p><b>Trending</b> = ♥ likes on X + 5 × GitHub stars + 8 × Hub likes + views ÷ 500. All categories share one grid, color-coded; use the category chips to filter. Promised-but-unreleased items always sort last.</p>
|
| 318 |
-
<p>Engagement numbers are a
|
| 319 |
<p>Visual identity and outlined Huggies from <a href="https://huggingface.co/spaces/Chunte/HFBA" target="_blank" rel="noopener">HF Huggiverse (Chunte/HFBA)</a>. Unofficial, community-maintained, not affiliated with TypeSafe AI.</p>
|
| 320 |
</div>
|
| 321 |
<img src="huggies/rocket.png" alt="">
|
|
@@ -348,117 +348,118 @@ const ITEMS = [
|
|
| 348 |
desc: 'Batches every JSON key at once and reads category probabilities from stock Qwen 2.5 1B logits, ~5× faster than generating JSON.',
|
| 349 |
quote: 'They were building in stealth for 2 years, I was building in stealth for 2 hours… No new training required.',
|
| 350 |
links: { tweet: 'https://x.com/harshagundal/status/2100044305536889015', hf: 'https://huggingface.co/harshatheg/Qwen-2.5-1B-RLCD', space: 'https://huggingface.co/spaces/drinkmoonshine/parallel-constrained-decoding' },
|
| 351 |
-
m: { likes:
|
| 352 |
{ cat: 'decode', title: 'openjev-sglang: Jev-compatible public API', by: { name: 'Eric Zhang', handle: 'ekzhang1' },
|
| 353 |
desc: 'A drop-in /v1/systemone endpoint over Qwen3.6-35B-A3B. SGLang\'s radix cache keeps the shared prefill hot, so 64 parallel tasks finish in under a second. Prefill-only, no generation.',
|
| 354 |
quote: 'It runs a comparable open model (Qwen3.6-35B-A3B), and just uses SGLang radix cache to preserve the prefill reuse — 64 tasks in <1s.',
|
| 355 |
links: { tweet: 'https://x.com/ekzhang1/status/2100651678110515383', gh: 'https://github.com/ekzhang/openjev-sglang' },
|
| 356 |
-
m: { likes:
|
| 357 |
{ cat: 'decode', title: 'openjev + openjev.com', by: { name: 'Theo Lee', handle: 'theoleecj' },
|
| 358 |
desc: 'Reads option logits on Qwen3.5-4B instead of generating text: 21 decisions in 1.02s vs 5.33s for JSON. CUDA plus a WebGPU browser demo.',
|
| 359 |
quote: 'On the Jev waitlist? You can use not Jev (but close enough) right here.',
|
| 360 |
links: { tweet: 'https://x.com/theoleecj/status/2100243609060856264', gh: 'https://github.com/TheoLeeCJ/openjev', demo: 'https://openjev.com' },
|
| 361 |
-
m: { likes: 10, rts: 0, views:
|
| 362 |
{ cat: 'decode', title: 'PocketJev: on-device iPhone visual decisions', by: { name: 'aqty', handle: 'aqty' },
|
| 363 |
desc: 'Qwen3-VL running locally on an iPhone via MLX. Camera frame + question + 3 options → direct option logits in ~1s, no photos saved.',
|
| 364 |
quote: 'iPhone上でQwen 2Bをローカル実行して、カメラ画像+質問+3択から文章生成なしで直接判定。',
|
| 365 |
links: { tweet: 'https://x.com/aqty/status/2100549517171175807', gh: 'https://github.com/NullPo-jp/PocketJev' },
|
| 366 |
-
m: { likes: 6, rts: 4, views:
|
| 367 |
{ cat: 'decode', title: 'jevmlx: Jev-style decisions for any MLX model', by: { name: 'Beni', handle: 'beni_il_' },
|
| 368 |
desc: 'Typed schema fields broadcast across the KV cache, one batch row per field, all scored in a single forward pass on any MLX model.',
|
| 369 |
quote: 'If you can run Qwen on your machine, it can behave very similarly to Jev.',
|
| 370 |
links: { tweet: 'https://x.com/beni_il_/status/2100617387116568956', gh: 'https://github.com/bnsd55/jevmlx' },
|
| 371 |
-
m: { likes: 7, rts: 0, views:
|
| 372 |
{ cat: 'decode', title: 'jev-on-a-laptop: the honest benchmark', by: { name: 'Richard Becker', handle: 'RichardBcker1' },
|
| 373 |
desc: 'Qwen 2.5 1.5B/7B and Qwen3 8B on a laptop vs Jev\'s published cases: 7B hits 73.8% agreement to Jev\'s 86.6%, 0.4–2s per decision. Finds confidence does not reliably flag errors.',
|
| 374 |
quote: 'All code, benchmarks, model-size comparisons and the full write-up (unofficial, no affiliation with TypeSafe).',
|
| 375 |
links: { tweet: 'https://x.com/RichardBcker1/status/2100195276107530393', gh: 'https://github.com/rorshopping/jev-on-a-laptop', space: 'https://huggingface.co/spaces/rorshopping/parallel-constrained-decisions' },
|
| 376 |
-
m: { likes: 1, rts: 0, views:
|
| 377 |
{ cat: 'decode', title: 'jevfire: one context, many decisions on CUDA', by: { name: 'Enrique Alcázar', handle: 'enriquealcazarg' },
|
| 378 |
desc: 'Parallel decisions for CUDA LLMs behind a vLLM-style API, with game-agent examples. Claims 10×+ over regular inference, runs in the browser, accepts images.',
|
| 379 |
quote: 'Here you have an open-source approach to the same problem… uses the same principles as JEV, is type-safe, runs in your browser, and accepts images.',
|
| 380 |
links: { tweet: 'https://x.com/enriquealcazarg/status/2100267330022375858', gh: 'https://github.com/kikoncuo/jevfire', site: 'https://kikoncuo.github.io/jevfire/' },
|
| 381 |
-
m: { likes: 1, rts: 0, views:
|
| 382 |
{ cat: 'decode', title: 'LFM2.5-RLCD (350M & 2.6B): PCD on Liquid models', by: { name: 'notnotsamuel · monotykamary', handle: null },
|
| 383 |
desc: 'Parallel constrained decoding on unchanged Liquid LFM2.5 weights, branching attention and convolution state. Inference-only: 8–63× faster, field accuracy still ~60%.',
|
| 384 |
links: { hf: 'https://huggingface.co/notnotsamuel/LFM2.5-350M-RLCD', hf2: 'https://huggingface.co/monotykamary/LFM2.5-2.6B-RLCD' },
|
| 385 |
-
m: { hfLikes:
|
| 386 |
{ cat: 'decode', title: 'parallelConstraintDecoding (Java + Python)', by: { name: 'Stephan Janssen', handle: 'Stephan007' },
|
| 387 |
desc: 'A Java port of the decode trick alongside a Python reference.',
|
| 388 |
quote: 'Here\'s a java version of Jev 🤩',
|
| 389 |
links: { tweet: 'https://x.com/Stephan007/status/2100673219967557906', gh: 'https://github.com/stephanj/parallelConstraintDecoding' },
|
| 390 |
-
m: { likes: 3, rts: 0, views:
|
| 391 |
{ cat: 'decode', title: 'system-one (snellingio): "System One Lite"', by: { name: 'snellingio', handle: null },
|
| 392 |
desc: 'Choice / Score / Noul server over MLX Qwen3, mirroring TypeSafe\'s three primitives. No launch post found.',
|
| 393 |
links: { gh: 'https://github.com/snellingio/system-one' },
|
| 394 |
-
m: { stars:
|
| 395 |
{ cat: 'decode', title: 'open-jev (daseinlabs): Gemma 3 4B on MLX', by: { name: 'Dasein Labs', handle: null },
|
| 396 |
desc: 'Prefix-cache option scoring on a local Gemma 3 4B via MLX, a Doom demo, and a /v1/systemone shim. Only seen in roundups.',
|
| 397 |
links: { gh: 'https://github.com/daseinlabs/open-jev' },
|
| 398 |
-
m: { stars:
|
| 399 |
{ cat: 'decode', title: 'openvons (open-Jev, Japanese)', by: { name: 'genai-craft', handle: null },
|
| 400 |
desc: 'Decision layer answering finite-option questions with probabilities, from text, images or Japanese voice commands.',
|
| 401 |
links: { gh: 'https://github.com/genai-craft/openvons' },
|
| 402 |
-
m: { stars:
|
| 403 |
{ cat: 'decode', title: 'openjev (zhihz): bilingual local decisions', by: { name: 'zhihz', handle: null },
|
| 404 |
desc: 'Local bilingual probability decisions from context, questions and candidate answers.',
|
| 405 |
links: { gh: 'https://github.com/zhihz/openjev' },
|
| 406 |
-
m: { stars:
|
| 407 |
{ cat: 'decode', title: 'system-one (sgoedecke): batched single-token choice', by: { name: 'Sean Goedecke', handle: null },
|
| 408 |
desc: 'Batched single-token choice inference for open models, wire-compatible with TypeSafe.',
|
| 409 |
links: { gh: 'https://github.com/sgoedecke/system-one' },
|
| 410 |
-
m: { stars:
|
| 411 |
-
{ cat: 'decode', title: 'reflex: stock Qwen3.5 / Qwen3-VL behind /v1/systemone', by: { name: '
|
| 412 |
-
desc: '
|
| 413 |
-
|
| 414 |
-
|
|
|
|
| 415 |
{ cat: 'decode', title: 'LitJev: Jev-schema decision layer for the Qwen family', by: { name: 'Zhengxu Yu', handle: 'yuzxfred' },
|
| 416 |
desc: 'Shared prefill, forked KV/recurrent state, one cached branch per question, option logits read from lm_head on any Qwen checkpoint (default Qwen3.8-27B), behind the exact Jev /v1/systemone schema. Playground, MMLU-Pro eval, optional temperature calibration.',
|
| 417 |
quote: 'Qwen3.8-27B,10 道 MMLU-Pro 一次完成,0.3 s。比让模型只生成答案快 2.9x。非官方,基于公开信息。',
|
| 418 |
links: { tweet: 'https://x.com/yuzxfred/status/2100652136878981337', gh: 'https://github.com/zhengxuyu/litjev', blog: 'https://x.com/yuzxfred/status/2100896621173047367' },
|
| 419 |
-
m: { likes: 2, rts: 0, views:
|
| 420 |
|
| 421 |
/* ---------------- trained ---------------- */
|
| 422 |
{ cat: 'decode', title: 'snapjudge: TypeSafe-compatible decisions on a Mac, scored against Jev', by: { name: 'Michael Gross', handle: 'michagross' },
|
| 423 |
desc: 'Choice / Score / Noul behind /v1/systemone on MLX Qwen3.5-family models, probabilities read straight from the logits. On TypeSafe\'s 20 public cases (373 decisions) Qwen3.8-27B reaches 81.2% agreement vs Jev\'s 87.7%: level on short texts, behind on invoices.',
|
| 424 |
links: { tweet: 'https://x.com/michagross/status/2101062157290856604', gh: 'https://github.com/Micha0827/snapjudge' },
|
| 425 |
-
m: { stars:
|
| 426 |
{ cat: 'trained', title: 'Bespoke Nimble: open data, open model, open recipe', by: { name: 'Mahesh Sathiamoorthy', handle: 'madiator' },
|
| 427 |
desc: 'The first fully open recipe: LoRA on Qwen3.5-9B with the data curation published too. Contrastive curation flips one fact to make negatives, so training needs no probabilities. 90.1% agreement with Jev vs 66.4% for base Qwen.',
|
| 428 |
quote: 'Introducing Bespoke Nimble: an open data, open model, open recipe for an open Jev.',
|
| 429 |
links: { tweet: 'https://x.com/madiator/status/2100990591215783946', gh: 'https://github.com/bespokelabsai/nimble', hf: 'https://huggingface.co/bespokelabs/Bespoke-Nimble-9B' },
|
| 430 |
-
m: { likes:
|
| 431 |
{ cat: 'trained', title: 'jevlike: train your own one-pass option scorer', by: { name: 'Vinny LaRouge', handle: 'vinnylarouge' },
|
| 432 |
desc: 'A tiny byte encoder (or frozen HF encoder) plus an option head maps context + N options to one probability each. Doom, chess and Wikispeedia demos; parent of several forks.',
|
| 433 |
quote: 'I reverse-engineered a jev-like architecture given its type. You can find the repo here to train your own jevlikes.',
|
| 434 |
links: { tweet: 'https://x.com/vinnylarouge/status/2100170846346097083', gh: 'https://github.com/vinnylarouge/jevlike', demo: 'https://x.com/vinnylarouge/status/2100281651930513460' },
|
| 435 |
-
m: { likes:
|
| 436 |
{ cat: 'trained', title: 'openjev (AlexWortega): Qwen3.5 + MLP head, real weights', by: { name: 'Alex Wortega', handle: 'justALEXWORTEGA' },
|
| 437 |
desc: 'The only trained "openjev" checkpoint on the Hub: Qwen3.5-4B as an NLI cross-encoder, argmax entailment is the move. Plays Doom zero-shot from text and pixels; 35B-A3B heads included.',
|
| 438 |
quote: 'Typesafe: pnewed 💨 Jev: liberated 🫡 I trained an MLP on top of qwen 4b and it works literally like JEV',
|
| 439 |
links: { tweet: 'https://x.com/justALEXWORTEGA/status/2100341039986798930', hf: 'https://huggingface.co/AlexWortega/openjev', gh: 'https://github.com/mikesmullin/openjev' },
|
| 440 |
-
m: { likes:
|
| 441 |
{ cat: 'trained', title: 'decider-2b: full fine-tune of Qwen3.5-2B-Base', by: { name: 'Mapika', handle: null },
|
| 442 |
desc: 'Full fine-tune of Qwen3.5-2B-Base on 942k examples from 64 decision datasets, plus a vision variant. Weights on the Hub.',
|
| 443 |
links: { gh: 'https://github.com/Mapika/decider', hf: 'https://huggingface.co/Mapika/decider-2b', hf2: 'https://huggingface.co/Mapika/decider-2b-vision' },
|
| 444 |
-
m: { stars:
|
| 445 |
{ cat: 'trained', title: 'rlcd-modernbert-151m ("OpenJev Verdict")', by: { name: 'heman10x', handle: null },
|
| 446 |
desc: 'Post-trained ModernBERT (151M) scoring 25 candidate slots, abstention included, in under 35ms. WebGPU playground, Apache-2.0.',
|
| 447 |
links: { hf: 'https://huggingface.co/heman10x/rlcd-modernbert-151m', gh: 'https://github.com/Heman10x-NGU/Verdict-open-jev' },
|
| 448 |
-
m: { hfLikes:
|
| 449 |
{ cat: 'trained', title: 'open-jev (pngwn): Qwen3.5-4B decision scorer + demo Space', by: { name: 'pngwn', handle: 'evilpingwin' },
|
| 450 |
desc: 'Qwen3.5-4B LoRA fine-tune that scores each candidate answer for a typed question, with temperature scaling so confidence tracks accuracy. Weights (v1, v2b) and a demo Space on the Hub.',
|
| 451 |
quote: 'Since everyone is open-jevving. I set ML Intern (on hugging chat) to work. After a few experiments and a little guidance and we did ok.',
|
| 452 |
links: { tweet: 'https://x.com/evilpingwin/status/2100721265107751121', space: 'https://huggingface.co/spaces/pngwn/open-jev', hf: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer-v2b', hf2: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer' },
|
| 453 |
-
m: { likes:
|
| 454 |
{ cat: 'trained', title: 'jevbetter: a stronger jevlike', by: { name: 'olanotolu', handle: null },
|
| 455 |
desc: 'jevlike with a fancier head: hashed n-gram encoder, rival-aware attention, temperature scaling. Benchmarked against jevlike.',
|
| 456 |
links: { gh: 'https://github.com/olanotolu/jevbetter' },
|
| 457 |
-
m: { stars:
|
| 458 |
{ cat: 'trained', title: 'system-one-open: Gemma 4 E2B / Gemma 3 270M + head', by: { name: 'mithalouni', handle: null },
|
| 459 |
desc: 'Typed calibrated decisions in one pass on Gemma 4 E2B and Gemma 3 270M, trained on Modal.',
|
| 460 |
links: { gh: 'https://github.com/mithalouni/system-one-open' },
|
| 461 |
-
m: { stars:
|
| 462 |
{ cat: 'trained', title: 'qwen-rlcd: Qwen3.5-0.8B Choice/Score/Noul', by: { name: 'shamazharikh', handle: null },
|
| 463 |
desc: 'Choice / Score / Noul on Qwen3.5-0.8B. Inference works; training marked on hold.',
|
| 464 |
links: { gh: 'https://github.com/shamazharikh/qwen-rlcd' },
|
|
@@ -466,11 +467,11 @@ const ITEMS = [
|
|
| 466 |
{ cat: 'trained', title: 'system-one-gemma: Gemma 3 270M scoring head', by: { name: 'akash-kamat', handle: null },
|
| 467 |
desc: 'Gemma 3 270M with a scoring head; single forward pass, no text generation.',
|
| 468 |
links: { gh: 'https://github.com/akash-kamat/system-one-gemma' },
|
| 469 |
-
m: { stars:
|
| 470 |
{ cat: 'trained', title: 'Laya: 421M RLCD decision model + demo Space', by: { name: 'Convai Innovations', handle: null },
|
| 471 |
desc: 'ModernBERT-large fine-tuned plus a decision head trained from scratch (421M total) with RLCD against proper scoring rules: choice/score/noul with calibrated probabilities, all questions in one pass. Weights and a 7-tab Gradio demo on the Hub.',
|
| 472 |
links: { gh: 'https://github.com/NandhaKishorM/laya', hf: 'https://huggingface.co/convaiinnovations/laya', space: 'https://huggingface.co/spaces/convaiinnovations/laya-demo', blog: 'https://dev.to/nandakishor_m_6cc0adfde9f/i-built-non-autoregressive-decision-models-a-year-ago-then-a-frontier-lab-called-it-a-18me', site: 'https://www.reddit.com/r/LocalLLaMA/comments/1wjieap/made_the_horizontal_opensource_model_for_jev_with/' },
|
| 473 |
-
m: { stars:
|
| 474 |
{ cat: 'trained', title: 'System One Mini: 69M DistilBERT, five fixed decisions', by: { name: 'David Hatley', handle: null },
|
| 475 |
desc: '69M DistilBERT with five fixed typed heads over synthetic software-diagnosis summaries. Its card says: not a reproduction of Jev or RLCD.',
|
| 476 |
links: { hf: 'https://huggingface.co/DavidHatley/system-one-mini' },
|
|
@@ -478,86 +479,86 @@ const ITEMS = [
|
|
| 478 |
{ cat: 'trained', title: 'NanoJev: 0.6B Qwen3 with a real decision head', by: { name: 'TianyuCodings', handle: null },
|
| 479 |
desc: 'A 0.6B Qwen3 backbone with an actual trained head emitting Choice / Boolean / Score across 2–255 candidates in one forward pass. Full training pipeline, maze and snake runs, weights and data on the Hub. The one Chinese and Korean roundups are pushing hardest.',
|
| 480 |
links: { gh: 'https://github.com/TianyuCodings/NanoJev', hf: 'https://huggingface.co/C-Tianyu/NanoJev' },
|
| 481 |
-
m: { stars:
|
| 482 |
{ cat: 'trained', title: 'mini-jev: a preregistered measurement of the trick', by: { name: 'r-ms', handle: null },
|
| 483 |
desc: 'Not another clone: a preregistered study of letter-logit versus grammar-JSON decoding on frozen Qwen3-4B and CLINC150. An actual measurement, no weights trained.',
|
| 484 |
links: { gh: 'https://github.com/r-ms/mini-jev' },
|
| 485 |
-
m: { stars:
|
| 486 |
|
| 487 |
{ cat: 'trained', promised: true, title: 'Archer Hume\'s open-weight Jev (promised)', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 488 |
desc: 'Announced alongside the architecture teardown: an open-weight version trained on the inferred design. Training in progress, nothing released yet.',
|
| 489 |
quote: 'I decided I\'d release an open-weight version. BUT training takes time, so while we all wait I decided I\'d drop the sauce.',
|
| 490 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 491 |
-
m: { likes:
|
| 492 |
|
| 493 |
/* ---------------- prior ---------------- */
|
| 494 |
{ cat: 'prior', title: 'GLiNER 2.5: "Weights are on Hugging Face."', by: { name: 'George (Fastino / GLiNER)', handle: 'george_onx' },
|
| 495 |
desc: 'Points the launch thread at the GLiNER 2.5 collection: a pretrained zero-shot encoder-classifier already doing single-pass classification over caller-supplied labels.',
|
| 496 |
quote: 'We\'ve been reading the comments. Weights are on Hugging Face.',
|
| 497 |
links: { tweet: 'https://x.com/george_onx/status/2100293114808119379', hf: 'https://huggingface.co/collections/fastino/gliner25-models' },
|
| 498 |
-
m: { likes:
|
| 499 |
{ cat: 'prior', title: 'DeepMost sales-conversion RL model + papers', by: { name: 'Nandakishor M', handle: 'Nandakishorm1' },
|
| 500 |
desc: 'Claims the idea shipped in 2025 papers (arXiv 2503.23303, 2510.01237) with a Hub model and dataset. PPO over embeddings; loosely related.',
|
| 501 |
quote: 'Also the second work published in September 2025 was exactly the same one jev proposed now.',
|
| 502 |
links: { tweet: 'https://x.com/Nandakishorm1/status/2100451360873611666', hf: 'https://huggingface.co/DeepMostInnovations/sales-conversion-model-reinf-learning', paper: 'https://arxiv.org/abs/2510.01237' },
|
| 503 |
-
m: { likes:
|
| 504 |
|
| 505 |
/* ---------------- arch ---------------- */
|
| 506 |
{ cat: 'explain', title: 'Jev\'s Architecture Unmasked', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 507 |
desc: 'Argues Jev is a causal transformer with shared-state encoding, per-question branches and direct probability readouts, likely on a sparse MoE backbone, inferred from latency probing.',
|
| 508 |
quote: 'All the grifters are completely wrong about Jev\'s architecture so I decided I\'d release an open-weight version. BUT training takes time…',
|
| 509 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 510 |
-
m: { likes:
|
| 511 |
|
| 512 |
/* ---------------- diffusion ---------------- */
|
| 513 |
{ cat: 'diffusion', title: 'DiffusionGemma-as-Jev + vLLM structured-generation patch', by: { name: 'Matt Mastracci', handle: 'mmastrac' },
|
| 514 |
desc: 'Adds a Jev-like structured mode for DiffusionGemma to vLLM (PR #57250) and runs live evals: not slower than Jev\'s API and roughly tied on quality.',
|
| 515 |
quote: 'Is Jev faster than DiffusionGemma? No ❌ Is Jev smarter than DiffusionGemma? No ❌ (they\'re roughly tied!)',
|
| 516 |
links: { tweet: 'https://x.com/mmastrac/status/2100626193943052784', pr: 'https://github.com/vllm-project/vllm/pull/57250' },
|
| 517 |
-
m: { likes:
|
| 518 |
{ cat: 'diffusion', title: 'razorback16/openjev + api.codiv.ai host', by: { name: 'razorback16', handle: null },
|
| 519 |
desc: 'The same vLLM PR #57250 route, packaged as a Jev-wire /v1/systemone service: a DiffusionGemma 26B-A4B canvas with only the label slots left as noise, one read-only denoise step reads out the distribution. Public host at api.codiv.ai.',
|
| 520 |
links: { gh: 'https://github.com/razorback16/openjev', site: 'https://api.codiv.ai/v1/systemone' },
|
| 521 |
-
m: { stars:
|
| 522 |
{ cat: 'diffusion', title: 'open-jev (JoshuaSP): typed JSON with DiffusionGemma', by: { name: 'Joshua SP', handle: null },
|
| 523 |
desc: 'Typed JSON inference with DiffusionGemma, benchmarked against Every and Jev.',
|
| 524 |
links: { gh: 'https://github.com/JoshuaSP/open-jev' },
|
| 525 |
-
m: { stars:
|
| 526 |
{ cat: 'diffusion', title: 'LocalJev (GitHub Next): oMLX Jev API + model bake-off', by: { name: 'GitHub Next', handle: 'GitHubNext' },
|
| 527 |
desc: 'Official GitHub Next bridge, since not everyone there has Jev access: a Bun server putting oMLX DiffusionGemma behind /v1/systemone. Prompts for JSON probabilities instead of reading logits: wire-compatible, not equivalent. Ships a 1,200-request bake-off.',
|
| 528 |
quote: 'Not everyone on the team has access to Jev yet. Spent a morning cobbling together a poor man\'s Jev on top of omlx for local use.',
|
| 529 |
links: { tweet: 'https://x.com/GitHubNext/status/2101193436816920798', gh: 'https://github.com/githubnext/localjev' },
|
| 530 |
-
m: { likes:
|
| 531 |
|
| 532 |
/* ---------------- explain ---------------- */
|
| 533 |
{ cat: 'explain', title: '45-second TL;DR on Jev (video)', by: { name: 'Matija Sosic', handle: 'MatijaSosic' },
|
| 534 |
desc: 'The most-shared explainer of the launch: a 45-second video that strips Jev down to its core idea, made because the official launch video was hard to follow.',
|
| 535 |
quote: 'I find the core idea beautifully simple, but the video made it really hard to understand.',
|
| 536 |
links: { tweet: 'https://x.com/MatijaSosic/status/2100190746389135772' },
|
| 537 |
-
m: { likes:
|
| 538 |
{ cat: 'explain', title: 'typesafe-ai-benchmark: an "imposter Jev" on Cerebras', by: { name: 'Shannon Duncan', handle: 'iamMrDuncan' },
|
| 539 |
desc: 'An "imposter Jev" gateway on Cerebras with Qwen 3.8 27B, compared side by side. Similar quality and speed; TypeSafe far cheaper.',
|
| 540 |
quote: 'Built an alternative version of @typesafeai but on @cerebras with Qwen 3.8 27b. Similar quality, similar performance, but vastly different cost.',
|
| 541 |
links: { tweet: 'https://x.com/iamMrDuncan/status/2100467548298899918', gh: 'https://github.com/iammrduncan/typesafe-ai-benchmark' },
|
| 542 |
-
m: { likes:
|
| 543 |
{ cat: 'explain', title: 'awesome-jev lists (AnotiaWang · hellogumbo)', by: { name: 'AnotiaWang · hellogumbo', handle: null },
|
| 544 |
desc: 'Two curated indexes; AnotiaWang\'s "Research & Open Models" section is the tight one, hellogumbo\'s catches the long tail.',
|
| 545 |
links: { gh: 'https://github.com/AnotiaWang/awesome-jev', gh2: 'https://github.com/hellogumbo/awesome-jev' },
|
| 546 |
-
m: { stars:
|
| 547 |
{ cat: 'explain', title: 'jev-benchmarks: probability-aware eval harness', by: { name: 'AbdelStark', handle: null },
|
| 548 |
desc: 'Calibration, selective risk and latency evals for typed decision models, Jev or open.',
|
| 549 |
links: { gh: 'https://github.com/AbdelStark/jev-benchmarks' },
|
| 550 |
-
m: { stars:
|
| 551 |
{ cat: 'explain', title: 'Thread: System 1 / System 2 agents and the open recreations', by: { name: 'hhkkmon', handle: 'hhkkmon' },
|
| 552 |
desc: '14-part explainer framing agents as "expensive LLM calls pretending to be if statements", listing the open recreations at the end.',
|
| 553 |
quote: 'Big models think. Small decision models react.',
|
| 554 |
links: { tweet: 'https://x.com/hhkkmon/status/2100443314957038010' },
|
| 555 |
-
m: { likes: 2, rts: 0, views:
|
| 556 |
{ cat: 'explain', title: 'Long-form: "OpenJev is NOT Jev open-sourced"', by: { name: 'mygtmhire', handle: 'mygtmhire' },
|
| 557 |
desc: 'Most thorough single post on TheoLeeCJ/openjev, including the warning that its probabilities are not calibrated confidence.',
|
| 558 |
quote: 'State in → probabilities out → code acts. No conversation required.',
|
| 559 |
links: { tweet: 'https://x.com/mygtmhire/status/2100666488625734103' },
|
| 560 |
-
m: { likes:
|
| 561 |
{ cat: 'explain', title: 'jev-behavior-study: probing the closed model', by: { name: 'RINNECODER', handle: null },
|
| 562 |
desc: 'Black-box behavioural study of the closed Jev 1.13.0 itself. Not a replica — it measures the thing everyone is trying to reproduce.',
|
| 563 |
links: { gh: 'https://github.com/RINNECODER/jev-behavior-study' },
|
|
@@ -608,7 +609,7 @@ const likesOf = it => it.m?.likes ?? -1;
|
|
| 608 |
brand-new entry nobody has reacted to yet cannot climb on its date alone.
|
| 609 |
The credit therefore rewards "new AND real", never novelty by itself. */
|
| 610 |
const W = { likes: 1, stars: 5, hfLikes: 8, views: 1/500 };
|
| 611 |
-
const RECENCY = { credit:
|
| 612 |
const BASE_DATE = '2026-09-19'; /* the snapshot the metrics were last refreshed against */
|
| 613 |
const ageDays = it => Math.max(0, (new Date(BASE_DATE) - new Date(it.date)) / 864e5);
|
| 614 |
const recency = it => {
|
|
@@ -690,7 +691,7 @@ function orderedCats() {
|
|
| 690 |
function cardHTML(it, cat, idx, q) {
|
| 691 |
const m = it.m || {};
|
| 692 |
const metrics = [];
|
| 693 |
-
const tr = trending(it); if (tr) metrics.push(`<span class="tr" title="trending = ♥X + 5×★GitHub + 8×🤗likes + views/500 + recency credit (up to
|
| 694 |
if (m.likes != null) metrics.push(`<span title="likes on X">${ICON.heart}${fmt(m.likes)}</span>`);
|
| 695 |
if (m.rts) metrics.push(`<span title="reposts">${ICON.rt}${fmt(m.rts)}</span>`);
|
| 696 |
if (m.views != null) metrics.push(`<span title="views on X">${ICON.eye}${fmt(m.views)}</span>`);
|
|
|
|
| 315 |
<div>
|
| 316 |
<p><b>Missing something?</b> Open a PR against this Space's <code class="mono">index.html</code> (the data is one JS array at the bottom of the file) or ping <a href="https://huggingface.co/multimodalart" target="_blank" rel="noopener">@multimodalart</a>. Hugging Face models and Spaces count as first-class artifacts here, not just GitHub repos.</p>
|
| 317 |
<p><b>Trending</b> = ♥ likes on X + 5 × GitHub stars + 8 × Hub likes + views ÷ 500. All categories share one grid, color-coded; use the category chips to filter. Promised-but-unreleased items always sort last.</p>
|
| 318 |
+
<p>Engagement numbers are a snapshot from public X post metadata on 2026-09-19 and will drift. GitHub stars and Hub likes were read the same day. Descriptions paraphrase each project's own README and launch post; corrections welcome.</p>
|
| 319 |
<p>Visual identity and outlined Huggies from <a href="https://huggingface.co/spaces/Chunte/HFBA" target="_blank" rel="noopener">HF Huggiverse (Chunte/HFBA)</a>. Unofficial, community-maintained, not affiliated with TypeSafe AI.</p>
|
| 320 |
</div>
|
| 321 |
<img src="huggies/rocket.png" alt="">
|
|
|
|
| 348 |
desc: 'Batches every JSON key at once and reads category probabilities from stock Qwen 2.5 1B logits, ~5× faster than generating JSON.',
|
| 349 |
quote: 'They were building in stealth for 2 years, I was building in stealth for 2 hours… No new training required.',
|
| 350 |
links: { tweet: 'https://x.com/harshagundal/status/2100044305536889015', hf: 'https://huggingface.co/harshatheg/Qwen-2.5-1B-RLCD', space: 'https://huggingface.co/spaces/drinkmoonshine/parallel-constrained-decoding' },
|
| 351 |
+
m: { likes: 5796, rts: 355, views: 1269064, hfLikes: 429 }, date: '2026-09-16', tags: 'qwen json mac m4 on-device forks epsilon3 shreyansh26 drinkmoonshine' },
|
| 352 |
{ cat: 'decode', title: 'openjev-sglang: Jev-compatible public API', by: { name: 'Eric Zhang', handle: 'ekzhang1' },
|
| 353 |
desc: 'A drop-in /v1/systemone endpoint over Qwen3.6-35B-A3B. SGLang\'s radix cache keeps the shared prefill hot, so 64 parallel tasks finish in under a second. Prefill-only, no generation.',
|
| 354 |
quote: 'It runs a comparable open model (Qwen3.6-35B-A3B), and just uses SGLang radix cache to preserve the prefill reuse — 64 tasks in <1s.',
|
| 355 |
links: { tweet: 'https://x.com/ekzhang1/status/2100651678110515383', gh: 'https://github.com/ekzhang/openjev-sglang' },
|
| 356 |
+
m: { likes: 977, rts: 60, views: 139265, stars: 196 }, date: '2026-09-17', tags: 'sglang qwen3.6 moe api systemone radix cache' },
|
| 357 |
{ cat: 'decode', title: 'openjev + openjev.com', by: { name: 'Theo Lee', handle: 'theoleecj' },
|
| 358 |
desc: 'Reads option logits on Qwen3.5-4B instead of generating text: 21 decisions in 1.02s vs 5.33s for JSON. CUDA plus a WebGPU browser demo.',
|
| 359 |
quote: 'On the Jev waitlist? You can use not Jev (but close enough) right here.',
|
| 360 |
links: { tweet: 'https://x.com/theoleecj/status/2100243609060856264', gh: 'https://github.com/TheoLeeCJ/openjev', demo: 'https://openjev.com' },
|
| 361 |
+
m: { likes: 10, rts: 0, views: 3291, stars: 1976 }, date: '2026-09-16', tags: 'webgpu browser rtx 3090 qwen3.5 logits mit' },
|
| 362 |
{ cat: 'decode', title: 'PocketJev: on-device iPhone visual decisions', by: { name: 'aqty', handle: 'aqty' },
|
| 363 |
desc: 'Qwen3-VL running locally on an iPhone via MLX. Camera frame + question + 3 options → direct option logits in ~1s, no photos saved.',
|
| 364 |
quote: 'iPhone上でQwen 2Bをローカル実行して、カメラ画像+質問+3択から文章生成なしで直接判定。',
|
| 365 |
links: { tweet: 'https://x.com/aqty/status/2100549517171175807', gh: 'https://github.com/NullPo-jp/PocketJev' },
|
| 366 |
+
m: { likes: 6, rts: 4, views: 424, stars: 0 }, date: '2026-09-17', tags: 'ios iphone mlx vision qwen3-vl camera japan' },
|
| 367 |
{ cat: 'decode', title: 'jevmlx: Jev-style decisions for any MLX model', by: { name: 'Beni', handle: 'beni_il_' },
|
| 368 |
desc: 'Typed schema fields broadcast across the KV cache, one batch row per field, all scored in a single forward pass on any MLX model.',
|
| 369 |
quote: 'If you can run Qwen on your machine, it can behave very similarly to Jev.',
|
| 370 |
links: { tweet: 'https://x.com/beni_il_/status/2100617387116568956', gh: 'https://github.com/bnsd55/jevmlx' },
|
| 371 |
+
m: { likes: 7, rts: 0, views: 1183, stars: 40 }, date: '2026-09-17', tags: 'mlx apple silicon kv cache hebrew' },
|
| 372 |
{ cat: 'decode', title: 'jev-on-a-laptop: the honest benchmark', by: { name: 'Richard Becker', handle: 'RichardBcker1' },
|
| 373 |
desc: 'Qwen 2.5 1.5B/7B and Qwen3 8B on a laptop vs Jev\'s published cases: 7B hits 73.8% agreement to Jev\'s 86.6%, 0.4–2s per decision. Finds confidence does not reliably flag errors.',
|
| 374 |
quote: 'All code, benchmarks, model-size comparisons and the full write-up (unofficial, no affiliation with TypeSafe).',
|
| 375 |
links: { tweet: 'https://x.com/RichardBcker1/status/2100195276107530393', gh: 'https://github.com/rorshopping/jev-on-a-laptop', space: 'https://huggingface.co/spaces/rorshopping/parallel-constrained-decisions' },
|
| 376 |
+
m: { likes: 1, rts: 0, views: 67, stars: 16 }, date: '2026-09-16', tags: 'benchmark study agreement fraud apple silicon m5 writeup explainer' },
|
| 377 |
{ cat: 'decode', title: 'jevfire: one context, many decisions on CUDA', by: { name: 'Enrique Alcázar', handle: 'enriquealcazarg' },
|
| 378 |
desc: 'Parallel decisions for CUDA LLMs behind a vLLM-style API, with game-agent examples. Claims 10×+ over regular inference, runs in the browser, accepts images.',
|
| 379 |
quote: 'Here you have an open-source approach to the same problem… uses the same principles as JEV, is type-safe, runs in your browser, and accepts images.',
|
| 380 |
links: { tweet: 'https://x.com/enriquealcazarg/status/2100267330022375858', gh: 'https://github.com/kikoncuo/jevfire', site: 'https://kikoncuo.github.io/jevfire/' },
|
| 381 |
+
m: { likes: 1, rts: 0, views: 250, stars: 8 }, date: '2026-09-16', tags: 'cuda vllm browser images games primeagen' },
|
| 382 |
{ cat: 'decode', title: 'LFM2.5-RLCD (350M & 2.6B): PCD on Liquid models', by: { name: 'notnotsamuel · monotykamary', handle: null },
|
| 383 |
desc: 'Parallel constrained decoding on unchanged Liquid LFM2.5 weights, branching attention and convolution state. Inference-only: 8–63× faster, field accuracy still ~60%.',
|
| 384 |
links: { hf: 'https://huggingface.co/notnotsamuel/LFM2.5-350M-RLCD', hf2: 'https://huggingface.co/monotykamary/LFM2.5-2.6B-RLCD' },
|
| 385 |
+
m: { hfLikes: 18 }, date: '2026-09-16', tags: 'liquid lfm hub weights renamed inference only pcd' },
|
| 386 |
{ cat: 'decode', title: 'parallelConstraintDecoding (Java + Python)', by: { name: 'Stephan Janssen', handle: 'Stephan007' },
|
| 387 |
desc: 'A Java port of the decode trick alongside a Python reference.',
|
| 388 |
quote: 'Here\'s a java version of Jev 🤩',
|
| 389 |
links: { tweet: 'https://x.com/Stephan007/status/2100673219967557906', gh: 'https://github.com/stephanj/parallelConstraintDecoding' },
|
| 390 |
+
m: { likes: 3, rts: 0, views: 103, stars: 10 }, date: '2026-09-17', tags: 'java jvm port' },
|
| 391 |
{ cat: 'decode', title: 'system-one (snellingio): "System One Lite"', by: { name: 'snellingio', handle: null },
|
| 392 |
desc: 'Choice / Score / Noul server over MLX Qwen3, mirroring TypeSafe\'s three primitives. No launch post found.',
|
| 393 |
links: { gh: 'https://github.com/snellingio/system-one' },
|
| 394 |
+
m: { stars: 38 }, date: '2026-09-16', tags: 'mlx qwen3 choice score noul server' },
|
| 395 |
{ cat: 'decode', title: 'open-jev (daseinlabs): Gemma 3 4B on MLX', by: { name: 'Dasein Labs', handle: null },
|
| 396 |
desc: 'Prefix-cache option scoring on a local Gemma 3 4B via MLX, a Doom demo, and a /v1/systemone shim. Only seen in roundups.',
|
| 397 |
links: { gh: 'https://github.com/daseinlabs/open-jev' },
|
| 398 |
+
m: { stars: 54 }, date: '2026-09-17', tags: 'gemma mlx doom systemone shim' },
|
| 399 |
{ cat: 'decode', title: 'openvons (open-Jev, Japanese)', by: { name: 'genai-craft', handle: null },
|
| 400 |
desc: 'Decision layer answering finite-option questions with probabilities, from text, images or Japanese voice commands.',
|
| 401 |
links: { gh: 'https://github.com/genai-craft/openvons' },
|
| 402 |
+
m: { stars: 9 }, date: '2026-09-17', tags: 'japanese voice image multimodal' },
|
| 403 |
{ cat: 'decode', title: 'openjev (zhihz): bilingual local decisions', by: { name: 'zhihz', handle: null },
|
| 404 |
desc: 'Local bilingual probability decisions from context, questions and candidate answers.',
|
| 405 |
links: { gh: 'https://github.com/zhihz/openjev' },
|
| 406 |
+
m: { stars: 14 }, date: '2026-09-16', tags: 'chinese bilingual local' },
|
| 407 |
{ cat: 'decode', title: 'system-one (sgoedecke): batched single-token choice', by: { name: 'Sean Goedecke', handle: null },
|
| 408 |
desc: 'Batched single-token choice inference for open models, wire-compatible with TypeSafe.',
|
| 409 |
links: { gh: 'https://github.com/sgoedecke/system-one' },
|
| 410 |
+
m: { stars: 14 }, date: '2026-09-17', tags: 'compatible batched single token' },
|
| 411 |
+
{ cat: 'decode', title: 'reflex: stock Qwen3.5 / Qwen3-VL behind /v1/systemone', by: { name: 'Kshetrajna Raghavan', handle: 'kshetrajna' },
|
| 412 |
+
desc: 'Runs Jev\'s shape on stock Qwen3.5: choice / score / noul in one batched pass, a /v1/systemone server, and a WebGPU demo on a 0.8B model. Temperature scaling gets Qwen3.5-4B to 0.039 ECE vs Jev\'s 0.031; a LoRA recipe trains honest probabilities on public data.',
|
| 413 |
+
quote: '"Wonder if we could build that?" is a pretty normal response to new tech at Shopify. Jev got me curious, so I built Reflex: a Qwen-based experiment in structured decisions + probabilities running on WebGPU.',
|
| 414 |
+
links: { tweet: 'https://x.com/kshetrajna/status/2100739853101195744', gh: 'https://github.com/kshetrajna12/reflex', demo: 'https://kshetrajna12.github.io/reflex/' },
|
| 415 |
+
m: { likes: 330, rts: 22, views: 138281, stars: 77 }, date: '2026-09-17', tags: 'systemone choice score noul qwen3-vl image shared state server webgpu browser demo lora calibration ece shopify mmlu' },
|
| 416 |
{ cat: 'decode', title: 'LitJev: Jev-schema decision layer for the Qwen family', by: { name: 'Zhengxu Yu', handle: 'yuzxfred' },
|
| 417 |
desc: 'Shared prefill, forked KV/recurrent state, one cached branch per question, option logits read from lm_head on any Qwen checkpoint (default Qwen3.8-27B), behind the exact Jev /v1/systemone schema. Playground, MMLU-Pro eval, optional temperature calibration.',
|
| 418 |
quote: 'Qwen3.8-27B,10 道 MMLU-Pro 一次完成,0.3 s。比让模型只生成答案快 2.9x。非官方,基于公开信息。',
|
| 419 |
links: { tweet: 'https://x.com/yuzxfred/status/2100652136878981337', gh: 'https://github.com/zhengxuyu/litjev', blog: 'https://x.com/yuzxfred/status/2100896621173047367' },
|
| 420 |
+
m: { likes: 2, rts: 0, views: 148, stars: 19 }, date: '2026-09-17', tags: 'qwen qwen3.8 27b systemone schema compatible calibration mmlu-pro doom chess vision transformers chinese explainer writeup limitations' },
|
| 421 |
|
| 422 |
/* ---------------- trained ---------------- */
|
| 423 |
{ cat: 'decode', title: 'snapjudge: TypeSafe-compatible decisions on a Mac, scored against Jev', by: { name: 'Michael Gross', handle: 'michagross' },
|
| 424 |
desc: 'Choice / Score / Noul behind /v1/systemone on MLX Qwen3.5-family models, probabilities read straight from the logits. On TypeSafe\'s 20 public cases (373 decisions) Qwen3.8-27B reaches 81.2% agreement vs Jev\'s 87.7%: level on short texts, behind on invoices.',
|
| 425 |
links: { tweet: 'https://x.com/michagross/status/2101062157290856604', gh: 'https://github.com/Micha0827/snapjudge' },
|
| 426 |
+
m: { stars: 3, likes: 1, rts: 0, views: 11 }, date: '2026-09-18', tags: 'mlx apple silicon qwen3.6 qwen3.8 systemone api benchmark agreement german runner game' },
|
| 427 |
{ cat: 'trained', title: 'Bespoke Nimble: open data, open model, open recipe', by: { name: 'Mahesh Sathiamoorthy', handle: 'madiator' },
|
| 428 |
desc: 'The first fully open recipe: LoRA on Qwen3.5-9B with the data curation published too. Contrastive curation flips one fact to make negatives, so training needs no probabilities. 90.1% agreement with Jev vs 66.4% for base Qwen.',
|
| 429 |
quote: 'Introducing Bespoke Nimble: an open data, open model, open recipe for an open Jev.',
|
| 430 |
links: { tweet: 'https://x.com/madiator/status/2100990591215783946', gh: 'https://github.com/bespokelabsai/nimble', hf: 'https://huggingface.co/bespokelabs/Bespoke-Nimble-9B' },
|
| 431 |
+
m: { likes: 1210, rts: 151, views: 415617, stars: 509, hfLikes: 59 }, date: '2026-09-18', tags: 'bespokelabs nimble qwen3.5 9b lora contrastive data curation open recipe synthetic distillation free parallel scorer 100ms h100 macbook' },
|
| 432 |
{ cat: 'trained', title: 'jevlike: train your own one-pass option scorer', by: { name: 'Vinny LaRouge', handle: 'vinnylarouge' },
|
| 433 |
desc: 'A tiny byte encoder (or frozen HF encoder) plus an option head maps context + N options to one probability each. Doom, chess and Wikispeedia demos; parent of several forks.',
|
| 434 |
quote: 'I reverse-engineered a jev-like architecture given its type. You can find the repo here to train your own jevlikes.',
|
| 435 |
links: { tweet: 'https://x.com/vinnylarouge/status/2100170846346097083', gh: 'https://github.com/vinnylarouge/jevlike', demo: 'https://x.com/vinnylarouge/status/2100281651930513460' },
|
| 436 |
+
m: { likes: 2018, rts: 168, views: 164627, stars: 999 }, date: '2026-09-16', tags: 'train scorer encoder doom chess wikispeedia rl' },
|
| 437 |
{ cat: 'trained', title: 'openjev (AlexWortega): Qwen3.5 + MLP head, real weights', by: { name: 'Alex Wortega', handle: 'justALEXWORTEGA' },
|
| 438 |
desc: 'The only trained "openjev" checkpoint on the Hub: Qwen3.5-4B as an NLI cross-encoder, argmax entailment is the move. Plays Doom zero-shot from text and pixels; 35B-A3B heads included.',
|
| 439 |
quote: 'Typesafe: pnewed 💨 Jev: liberated 🫡 I trained an MLP on top of qwen 4b and it works literally like JEV',
|
| 440 |
links: { tweet: 'https://x.com/justALEXWORTEGA/status/2100341039986798930', hf: 'https://huggingface.co/AlexWortega/openjev', gh: 'https://github.com/mikesmullin/openjev' },
|
| 441 |
+
m: { likes: 774, rts: 47, views: 103797, hfLikes: 213, stars: 2 }, date: '2026-09-16', tags: 'nli cross-encoder mlp qwen3.5 doom vision weights hub' },
|
| 442 |
{ cat: 'trained', title: 'decider-2b: full fine-tune of Qwen3.5-2B-Base', by: { name: 'Mapika', handle: null },
|
| 443 |
desc: 'Full fine-tune of Qwen3.5-2B-Base on 942k examples from 64 decision datasets, plus a vision variant. Weights on the Hub.',
|
| 444 |
links: { gh: 'https://github.com/Mapika/decider', hf: 'https://huggingface.co/Mapika/decider-2b', hf2: 'https://huggingface.co/Mapika/decider-2b-vision' },
|
| 445 |
+
m: { stars: 96, hfLikes: 49 }, date: '2026-09-16', tags: 'fine-tune qwen3.5 2b calibrated vision gh200 hub weights' },
|
| 446 |
{ cat: 'trained', title: 'rlcd-modernbert-151m ("OpenJev Verdict")', by: { name: 'heman10x', handle: null },
|
| 447 |
desc: 'Post-trained ModernBERT (151M) scoring 25 candidate slots, abstention included, in under 35ms. WebGPU playground, Apache-2.0.',
|
| 448 |
links: { hf: 'https://huggingface.co/heman10x/rlcd-modernbert-151m', gh: 'https://github.com/Heman10x-NGU/Verdict-open-jev' },
|
| 449 |
+
m: { hfLikes: 12, stars: 25 }, date: '2026-09-17', tags: 'modernbert encoder webgpu abstention gliclass hub weights' },
|
| 450 |
{ cat: 'trained', title: 'open-jev (pngwn): Qwen3.5-4B decision scorer + demo Space', by: { name: 'pngwn', handle: 'evilpingwin' },
|
| 451 |
desc: 'Qwen3.5-4B LoRA fine-tune that scores each candidate answer for a typed question, with temperature scaling so confidence tracks accuracy. Weights (v1, v2b) and a demo Space on the Hub.',
|
| 452 |
quote: 'Since everyone is open-jevving. I set ML Intern (on hugging chat) to work. After a few experiments and a little guidance and we did ok.',
|
| 453 |
links: { tweet: 'https://x.com/evilpingwin/status/2100721265107751121', space: 'https://huggingface.co/spaces/pngwn/open-jev', hf: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer-v2b', hf2: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer' },
|
| 454 |
+
m: { likes: 230, rts: 36, views: 13184, hfLikes: 10 }, date: '2026-09-17', tags: 'qwen3.5 scoring head temperature scaling calibration ml intern huggingchat hub weights space demo' },
|
| 455 |
{ cat: 'trained', title: 'jevbetter: a stronger jevlike', by: { name: 'olanotolu', handle: null },
|
| 456 |
desc: 'jevlike with a fancier head: hashed n-gram encoder, rival-aware attention, temperature scaling. Benchmarked against jevlike.',
|
| 457 |
links: { gh: 'https://github.com/olanotolu/jevbetter' },
|
| 458 |
+
m: { stars: 12 }, date: '2026-09-16', tags: 'n-gram attention temperature scaling calibration' },
|
| 459 |
{ cat: 'trained', title: 'system-one-open: Gemma 4 E2B / Gemma 3 270M + head', by: { name: 'mithalouni', handle: null },
|
| 460 |
desc: 'Typed calibrated decisions in one pass on Gemma 4 E2B and Gemma 3 270M, trained on Modal.',
|
| 461 |
links: { gh: 'https://github.com/mithalouni/system-one-open' },
|
| 462 |
+
m: { stars: 11 }, date: '2026-09-17', tags: 'gemma modal scoring head' },
|
| 463 |
{ cat: 'trained', title: 'qwen-rlcd: Qwen3.5-0.8B Choice/Score/Noul', by: { name: 'shamazharikh', handle: null },
|
| 464 |
desc: 'Choice / Score / Noul on Qwen3.5-0.8B. Inference works; training marked on hold.',
|
| 465 |
links: { gh: 'https://github.com/shamazharikh/qwen-rlcd' },
|
|
|
|
| 467 |
{ cat: 'trained', title: 'system-one-gemma: Gemma 3 270M scoring head', by: { name: 'akash-kamat', handle: null },
|
| 468 |
desc: 'Gemma 3 270M with a scoring head; single forward pass, no text generation.',
|
| 469 |
links: { gh: 'https://github.com/akash-kamat/system-one-gemma' },
|
| 470 |
+
m: { stars: 2 }, date: '2026-09-17', tags: 'gemma 270m tiny' },
|
| 471 |
{ cat: 'trained', title: 'Laya: 421M RLCD decision model + demo Space', by: { name: 'Convai Innovations', handle: null },
|
| 472 |
desc: 'ModernBERT-large fine-tuned plus a decision head trained from scratch (421M total) with RLCD against proper scoring rules: choice/score/noul with calibrated probabilities, all questions in one pass. Weights and a 7-tab Gradio demo on the Hub.',
|
| 473 |
links: { gh: 'https://github.com/NandhaKishorM/laya', hf: 'https://huggingface.co/convaiinnovations/laya', space: 'https://huggingface.co/spaces/convaiinnovations/laya-demo', blog: 'https://dev.to/nandakishor_m_6cc0adfde9f/i-built-non-autoregressive-decision-models-a-year-ago-then-a-frontier-lab-called-it-a-18me', site: 'https://www.reddit.com/r/LocalLLaMA/comments/1wjieap/made_the_horizontal_opensource_model_for_jev_with/' },
|
| 474 |
+
m: { stars: 1410, hfLikes: 577 }, date: '2026-09-18', tags: 'modernbert rlcd calibrated choice score noul support triage guardrails rag moderation routing hub weights space demo github pypi reddit devto' },
|
| 475 |
{ cat: 'trained', title: 'System One Mini: 69M DistilBERT, five fixed decisions', by: { name: 'David Hatley', handle: null },
|
| 476 |
desc: '69M DistilBERT with five fixed typed heads over synthetic software-diagnosis summaries. Its card says: not a reproduction of Jev or RLCD.',
|
| 477 |
links: { hf: 'https://huggingface.co/DavidHatley/system-one-mini' },
|
|
|
|
| 479 |
{ cat: 'trained', title: 'NanoJev: 0.6B Qwen3 with a real decision head', by: { name: 'TianyuCodings', handle: null },
|
| 480 |
desc: 'A 0.6B Qwen3 backbone with an actual trained head emitting Choice / Boolean / Score across 2–255 candidates in one forward pass. Full training pipeline, maze and snake runs, weights and data on the Hub. The one Chinese and Korean roundups are pushing hardest.',
|
| 481 |
links: { gh: 'https://github.com/TianyuCodings/NanoJev', hf: 'https://huggingface.co/C-Tianyu/NanoJev' },
|
| 482 |
+
m: { stars: 975, hfLikes: 35 }, date: '2026-09-17', tags: 'nanojev 0.6b qwen3 decision head choice boolean score maze snake training pipeline weights hub china korea' },
|
| 483 |
{ cat: 'trained', title: 'mini-jev: a preregistered measurement of the trick', by: { name: 'r-ms', handle: null },
|
| 484 |
desc: 'Not another clone: a preregistered study of letter-logit versus grammar-JSON decoding on frozen Qwen3-4B and CLINC150. An actual measurement, no weights trained.',
|
| 485 |
links: { gh: 'https://github.com/r-ms/mini-jev' },
|
| 486 |
+
m: { stars: 24 }, date: '2026-09-17', tags: 'preregistered study letter logits grammar json clinc150 qwen3-4b measurement' },
|
| 487 |
|
| 488 |
{ cat: 'trained', promised: true, title: 'Archer Hume\'s open-weight Jev (promised)', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 489 |
desc: 'Announced alongside the architecture teardown: an open-weight version trained on the inferred design. Training in progress, nothing released yet.',
|
| 490 |
quote: 'I decided I\'d release an open-weight version. BUT training takes time, so while we all wait I decided I\'d drop the sauce.',
|
| 491 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 492 |
+
m: { likes: 2242, rts: 212, views: 376500 }, date: '2026-09-17', tags: 'promised coming soon open-weight training archer' },
|
| 493 |
|
| 494 |
/* ---------------- prior ---------------- */
|
| 495 |
{ cat: 'prior', title: 'GLiNER 2.5: "Weights are on Hugging Face."', by: { name: 'George (Fastino / GLiNER)', handle: 'george_onx' },
|
| 496 |
desc: 'Points the launch thread at the GLiNER 2.5 collection: a pretrained zero-shot encoder-classifier already doing single-pass classification over caller-supplied labels.',
|
| 497 |
quote: 'We\'ve been reading the comments. Weights are on Hugging Face.',
|
| 498 |
links: { tweet: 'https://x.com/george_onx/status/2100293114808119379', hf: 'https://huggingface.co/collections/fastino/gliner25-models' },
|
| 499 |
+
m: { likes: 1584, rts: 94, views: 224123 }, date: '2026-09-16', tags: 'gliner fastino zero-shot encoder classifier collection' },
|
| 500 |
{ cat: 'prior', title: 'DeepMost sales-conversion RL model + papers', by: { name: 'Nandakishor M', handle: 'Nandakishorm1' },
|
| 501 |
desc: 'Claims the idea shipped in 2025 papers (arXiv 2503.23303, 2510.01237) with a Hub model and dataset. PPO over embeddings; loosely related.',
|
| 502 |
quote: 'Also the second work published in September 2025 was exactly the same one jev proposed now.',
|
| 503 |
links: { tweet: 'https://x.com/Nandakishorm1/status/2100451360873611666', hf: 'https://huggingface.co/DeepMostInnovations/sales-conversion-model-reinf-learning', paper: 'https://arxiv.org/abs/2510.01237' },
|
| 504 |
+
m: { likes: 1, rts: 0, views: 84, hfLikes: 132 }, date: '2026-09-17', tags: 'ppo embeddings arxiv prior art sales' },
|
| 505 |
|
| 506 |
/* ---------------- arch ---------------- */
|
| 507 |
{ cat: 'explain', title: 'Jev\'s Architecture Unmasked', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 508 |
desc: 'Argues Jev is a causal transformer with shared-state encoding, per-question branches and direct probability readouts, likely on a sparse MoE backbone, inferred from latency probing.',
|
| 509 |
quote: 'All the grifters are completely wrong about Jev\'s architecture so I decided I\'d release an open-weight version. BUT training takes time…',
|
| 510 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 511 |
+
m: { likes: 2242, rts: 212, views: 376501 }, date: '2026-09-17', tags: 'moe teardown speculation open-weight promise blog' },
|
| 512 |
|
| 513 |
/* ---------------- diffusion ---------------- */
|
| 514 |
{ cat: 'diffusion', title: 'DiffusionGemma-as-Jev + vLLM structured-generation patch', by: { name: 'Matt Mastracci', handle: 'mmastrac' },
|
| 515 |
desc: 'Adds a Jev-like structured mode for DiffusionGemma to vLLM (PR #57250) and runs live evals: not slower than Jev\'s API and roughly tied on quality.',
|
| 516 |
quote: 'Is Jev faster than DiffusionGemma? No ❌ Is Jev smarter than DiffusionGemma? No ❌ (they\'re roughly tied!)',
|
| 517 |
links: { tweet: 'https://x.com/mmastrac/status/2100626193943052784', pr: 'https://github.com/vllm-project/vllm/pull/57250' },
|
| 518 |
+
m: { likes: 1003, rts: 75, views: 326616 }, date: '2026-09-17', tags: 'diffusion gemma vllm dgx spark evals' },
|
| 519 |
{ cat: 'diffusion', title: 'razorback16/openjev + api.codiv.ai host', by: { name: 'razorback16', handle: null },
|
| 520 |
desc: 'The same vLLM PR #57250 route, packaged as a Jev-wire /v1/systemone service: a DiffusionGemma 26B-A4B canvas with only the label slots left as noise, one read-only denoise step reads out the distribution. Public host at api.codiv.ai.',
|
| 521 |
links: { gh: 'https://github.com/razorback16/openjev', site: 'https://api.codiv.ai/v1/systemone' },
|
| 522 |
+
m: { stars: 134 }, date: '2026-09-17', tags: 'diffusiongemma 26b a4b nvfp4 vllm 57250 hosted api codiv systemone denoise canvas' },
|
| 523 |
{ cat: 'diffusion', title: 'open-jev (JoshuaSP): typed JSON with DiffusionGemma', by: { name: 'Joshua SP', handle: null },
|
| 524 |
desc: 'Typed JSON inference with DiffusionGemma, benchmarked against Every and Jev.',
|
| 525 |
links: { gh: 'https://github.com/JoshuaSP/open-jev' },
|
| 526 |
+
m: { stars: 16 }, date: '2026-09-16', tags: 'diffusion gemma benchmark every' },
|
| 527 |
{ cat: 'diffusion', title: 'LocalJev (GitHub Next): oMLX Jev API + model bake-off', by: { name: 'GitHub Next', handle: 'GitHubNext' },
|
| 528 |
desc: 'Official GitHub Next bridge, since not everyone there has Jev access: a Bun server putting oMLX DiffusionGemma behind /v1/systemone. Prompts for JSON probabilities instead of reading logits: wire-compatible, not equivalent. Ships a 1,200-request bake-off.',
|
| 529 |
quote: 'Not everyone on the team has access to Jev yet. Spent a morning cobbling together a poor man\'s Jev on top of omlx for local use.',
|
| 530 |
links: { tweet: 'https://x.com/GitHubNext/status/2101193436816920798', gh: 'https://github.com/githubnext/localjev' },
|
| 531 |
+
m: { likes: 1327, rts: 168, views: 185193, stars: 449 }, date: '2026-09-19', tags: 'github next localjev omlx bun typescript diffusiongemma 26b a4b json probabilities bakeoff m5 max benchmark qwen gemma moe wire compatible promptcoding' },
|
| 532 |
|
| 533 |
/* ---------------- explain ---------------- */
|
| 534 |
{ cat: 'explain', title: '45-second TL;DR on Jev (video)', by: { name: 'Matija Sosic', handle: 'MatijaSosic' },
|
| 535 |
desc: 'The most-shared explainer of the launch: a 45-second video that strips Jev down to its core idea, made because the official launch video was hard to follow.',
|
| 536 |
quote: 'I find the core idea beautifully simple, but the video made it really hard to understand.',
|
| 537 |
links: { tweet: 'https://x.com/MatijaSosic/status/2100190746389135772' },
|
| 538 |
+
m: { likes: 9277, rts: 668, views: 1584798 }, date: '2026-09-16', tags: 'video tldr explainer launch' },
|
| 539 |
{ cat: 'explain', title: 'typesafe-ai-benchmark: an "imposter Jev" on Cerebras', by: { name: 'Shannon Duncan', handle: 'iamMrDuncan' },
|
| 540 |
desc: 'An "imposter Jev" gateway on Cerebras with Qwen 3.8 27B, compared side by side. Similar quality and speed; TypeSafe far cheaper.',
|
| 541 |
quote: 'Built an alternative version of @typesafeai but on @cerebras with Qwen 3.8 27b. Similar quality, similar performance, but vastly different cost.',
|
| 542 |
links: { tweet: 'https://x.com/iamMrDuncan/status/2100467548298899918', gh: 'https://github.com/iammrduncan/typesafe-ai-benchmark' },
|
| 543 |
+
m: { likes: 209, rts: 25, views: 25998, stars: 32 }, date: '2026-09-17', tags: 'cerebras qwen gateway benchmark comparison cost' },
|
| 544 |
{ cat: 'explain', title: 'awesome-jev lists (AnotiaWang · hellogumbo)', by: { name: 'AnotiaWang · hellogumbo', handle: null },
|
| 545 |
desc: 'Two curated indexes; AnotiaWang\'s "Research & Open Models" section is the tight one, hellogumbo\'s catches the long tail.',
|
| 546 |
links: { gh: 'https://github.com/AnotiaWang/awesome-jev', gh2: 'https://github.com/hellogumbo/awesome-jev' },
|
| 547 |
+
m: { stars: 155 }, date: '2026-09-16', tags: 'awesome list index curated roundup' },
|
| 548 |
{ cat: 'explain', title: 'jev-benchmarks: probability-aware eval harness', by: { name: 'AbdelStark', handle: null },
|
| 549 |
desc: 'Calibration, selective risk and latency evals for typed decision models, Jev or open.',
|
| 550 |
links: { gh: 'https://github.com/AbdelStark/jev-benchmarks' },
|
| 551 |
+
m: { stars: 11 }, date: '2026-09-17', tags: 'calibration selective risk eval harness' },
|
| 552 |
{ cat: 'explain', title: 'Thread: System 1 / System 2 agents and the open recreations', by: { name: 'hhkkmon', handle: 'hhkkmon' },
|
| 553 |
desc: '14-part explainer framing agents as "expensive LLM calls pretending to be if statements", listing the open recreations at the end.',
|
| 554 |
quote: 'Big models think. Small decision models react.',
|
| 555 |
links: { tweet: 'https://x.com/hhkkmon/status/2100443314957038010' },
|
| 556 |
+
m: { likes: 2, rts: 0, views: 119 }, date: '2026-09-17', tags: 'thread explainer system 1 system 2 agents' },
|
| 557 |
{ cat: 'explain', title: 'Long-form: "OpenJev is NOT Jev open-sourced"', by: { name: 'mygtmhire', handle: 'mygtmhire' },
|
| 558 |
desc: 'Most thorough single post on TheoLeeCJ/openjev, including the warning that its probabilities are not calibrated confidence.',
|
| 559 |
quote: 'State in → probabilities out → code acts. No conversation required.',
|
| 560 |
links: { tweet: 'https://x.com/mygtmhire/status/2100666488625734103' },
|
| 561 |
+
m: { likes: 2, rts: 0, views: 123 }, date: '2026-09-17', tags: 'explainer openjev routing calibration' },
|
| 562 |
{ cat: 'explain', title: 'jev-behavior-study: probing the closed model', by: { name: 'RINNECODER', handle: null },
|
| 563 |
desc: 'Black-box behavioural study of the closed Jev 1.13.0 itself. Not a replica — it measures the thing everyone is trying to reproduce.',
|
| 564 |
links: { gh: 'https://github.com/RINNECODER/jev-behavior-study' },
|
|
|
|
| 609 |
brand-new entry nobody has reacted to yet cannot climb on its date alone.
|
| 610 |
The credit therefore rewards "new AND real", never novelty by itself. */
|
| 611 |
const W = { likes: 1, stars: 5, hfLikes: 8, views: 1/500 };
|
| 612 |
+
const RECENCY = { credit: 10000, halfLifeDays: 1, ref: 400 };
|
| 613 |
const BASE_DATE = '2026-09-19'; /* the snapshot the metrics were last refreshed against */
|
| 614 |
const ageDays = it => Math.max(0, (new Date(BASE_DATE) - new Date(it.date)) / 864e5);
|
| 615 |
const recency = it => {
|
|
|
|
| 691 |
function cardHTML(it, cat, idx, q) {
|
| 692 |
const m = it.m || {};
|
| 693 |
const metrics = [];
|
| 694 |
+
const tr = trending(it); if (tr) metrics.push(`<span class="tr" title="trending = ♥X + 5×★GitHub + 8×🤗likes + views/500 + recency credit (up to 10000, scaled by engagement, halving daily)">▲ ${fmt(tr)}</span>`);
|
| 695 |
if (m.likes != null) metrics.push(`<span title="likes on X">${ICON.heart}${fmt(m.likes)}</span>`);
|
| 696 |
if (m.rts) metrics.push(`<span title="reposts">${ICON.rt}${fmt(m.rts)}</span>`);
|
| 697 |
if (m.views != null) metrics.push(`<span title="views on X">${ICON.eye}${fmt(m.views)}</span>`);
|
|
@@ -250,117 +250,118 @@ const ITEMS = [
|
|
| 250 |
desc: 'Batches every JSON key at once and reads category probabilities from stock Qwen 2.5 1B logits, ~5× faster than generating JSON.',
|
| 251 |
quote: 'They were building in stealth for 2 years, I was building in stealth for 2 hours… No new training required.',
|
| 252 |
links: { tweet: 'https://x.com/harshagundal/status/2100044305536889015', hf: 'https://huggingface.co/harshatheg/Qwen-2.5-1B-RLCD', space: 'https://huggingface.co/spaces/drinkmoonshine/parallel-constrained-decoding' },
|
| 253 |
-
m: { likes:
|
| 254 |
{ cat: 'decode', title: 'openjev-sglang: Jev-compatible public API', by: { name: 'Eric Zhang', handle: 'ekzhang1' },
|
| 255 |
desc: 'A drop-in /v1/systemone endpoint over Qwen3.6-35B-A3B. SGLang\'s radix cache keeps the shared prefill hot, so 64 parallel tasks finish in under a second. Prefill-only, no generation.',
|
| 256 |
quote: 'It runs a comparable open model (Qwen3.6-35B-A3B), and just uses SGLang radix cache to preserve the prefill reuse — 64 tasks in <1s.',
|
| 257 |
links: { tweet: 'https://x.com/ekzhang1/status/2100651678110515383', gh: 'https://github.com/ekzhang/openjev-sglang' },
|
| 258 |
-
m: { likes:
|
| 259 |
{ cat: 'decode', title: 'openjev + openjev.com', by: { name: 'Theo Lee', handle: 'theoleecj' },
|
| 260 |
desc: 'Reads option logits on Qwen3.5-4B instead of generating text: 21 decisions in 1.02s vs 5.33s for JSON. CUDA plus a WebGPU browser demo.',
|
| 261 |
quote: 'On the Jev waitlist? You can use not Jev (but close enough) right here.',
|
| 262 |
links: { tweet: 'https://x.com/theoleecj/status/2100243609060856264', gh: 'https://github.com/TheoLeeCJ/openjev', demo: 'https://openjev.com' },
|
| 263 |
-
m: { likes: 10, rts: 0, views:
|
| 264 |
{ cat: 'decode', title: 'PocketJev: on-device iPhone visual decisions', by: { name: 'aqty', handle: 'aqty' },
|
| 265 |
desc: 'Qwen3-VL running locally on an iPhone via MLX. Camera frame + question + 3 options → direct option logits in ~1s, no photos saved.',
|
| 266 |
quote: 'iPhone上でQwen 2Bをローカル実行して、カメラ画像+質問+3択から文章生成なしで直接判定。',
|
| 267 |
links: { tweet: 'https://x.com/aqty/status/2100549517171175807', gh: 'https://github.com/NullPo-jp/PocketJev' },
|
| 268 |
-
m: { likes: 6, rts: 4, views:
|
| 269 |
{ cat: 'decode', title: 'jevmlx: Jev-style decisions for any MLX model', by: { name: 'Beni', handle: 'beni_il_' },
|
| 270 |
desc: 'Typed schema fields broadcast across the KV cache, one batch row per field, all scored in a single forward pass on any MLX model.',
|
| 271 |
quote: 'If you can run Qwen on your machine, it can behave very similarly to Jev.',
|
| 272 |
links: { tweet: 'https://x.com/beni_il_/status/2100617387116568956', gh: 'https://github.com/bnsd55/jevmlx' },
|
| 273 |
-
m: { likes: 7, rts: 0, views:
|
| 274 |
{ cat: 'decode', title: 'jev-on-a-laptop: the honest benchmark', by: { name: 'Richard Becker', handle: 'RichardBcker1' },
|
| 275 |
desc: 'Qwen 2.5 1.5B/7B and Qwen3 8B on a laptop vs Jev\'s published cases: 7B hits 73.8% agreement to Jev\'s 86.6%, 0.4–2s per decision. Finds confidence does not reliably flag errors.',
|
| 276 |
quote: 'All code, benchmarks, model-size comparisons and the full write-up (unofficial, no affiliation with TypeSafe).',
|
| 277 |
links: { tweet: 'https://x.com/RichardBcker1/status/2100195276107530393', gh: 'https://github.com/rorshopping/jev-on-a-laptop', space: 'https://huggingface.co/spaces/rorshopping/parallel-constrained-decisions' },
|
| 278 |
-
m: { likes: 1, rts: 0, views:
|
| 279 |
{ cat: 'decode', title: 'jevfire: one context, many decisions on CUDA', by: { name: 'Enrique Alcázar', handle: 'enriquealcazarg' },
|
| 280 |
desc: 'Parallel decisions for CUDA LLMs behind a vLLM-style API, with game-agent examples. Claims 10×+ over regular inference, runs in the browser, accepts images.',
|
| 281 |
quote: 'Here you have an open-source approach to the same problem… uses the same principles as JEV, is type-safe, runs in your browser, and accepts images.',
|
| 282 |
links: { tweet: 'https://x.com/enriquealcazarg/status/2100267330022375858', gh: 'https://github.com/kikoncuo/jevfire', site: 'https://kikoncuo.github.io/jevfire/' },
|
| 283 |
-
m: { likes: 1, rts: 0, views:
|
| 284 |
{ cat: 'decode', title: 'LFM2.5-RLCD (350M & 2.6B): PCD on Liquid models', by: { name: 'notnotsamuel · monotykamary', handle: null },
|
| 285 |
desc: 'Parallel constrained decoding on unchanged Liquid LFM2.5 weights, branching attention and convolution state. Inference-only: 8–63× faster, field accuracy still ~60%.',
|
| 286 |
links: { hf: 'https://huggingface.co/notnotsamuel/LFM2.5-350M-RLCD', hf2: 'https://huggingface.co/monotykamary/LFM2.5-2.6B-RLCD' },
|
| 287 |
-
m: { hfLikes:
|
| 288 |
{ cat: 'decode', title: 'parallelConstraintDecoding (Java + Python)', by: { name: 'Stephan Janssen', handle: 'Stephan007' },
|
| 289 |
desc: 'A Java port of the decode trick alongside a Python reference.',
|
| 290 |
quote: 'Here\'s a java version of Jev 🤩',
|
| 291 |
links: { tweet: 'https://x.com/Stephan007/status/2100673219967557906', gh: 'https://github.com/stephanj/parallelConstraintDecoding' },
|
| 292 |
-
m: { likes: 3, rts: 0, views:
|
| 293 |
{ cat: 'decode', title: 'system-one (snellingio): "System One Lite"', by: { name: 'snellingio', handle: null },
|
| 294 |
desc: 'Choice / Score / Noul server over MLX Qwen3, mirroring TypeSafe\'s three primitives. No launch post found.',
|
| 295 |
links: { gh: 'https://github.com/snellingio/system-one' },
|
| 296 |
-
m: { stars:
|
| 297 |
{ cat: 'decode', title: 'open-jev (daseinlabs): Gemma 3 4B on MLX', by: { name: 'Dasein Labs', handle: null },
|
| 298 |
desc: 'Prefix-cache option scoring on a local Gemma 3 4B via MLX, a Doom demo, and a /v1/systemone shim. Only seen in roundups.',
|
| 299 |
links: { gh: 'https://github.com/daseinlabs/open-jev' },
|
| 300 |
-
m: { stars:
|
| 301 |
{ cat: 'decode', title: 'openvons (open-Jev, Japanese)', by: { name: 'genai-craft', handle: null },
|
| 302 |
desc: 'Decision layer answering finite-option questions with probabilities, from text, images or Japanese voice commands.',
|
| 303 |
links: { gh: 'https://github.com/genai-craft/openvons' },
|
| 304 |
-
m: { stars:
|
| 305 |
{ cat: 'decode', title: 'openjev (zhihz): bilingual local decisions', by: { name: 'zhihz', handle: null },
|
| 306 |
desc: 'Local bilingual probability decisions from context, questions and candidate answers.',
|
| 307 |
links: { gh: 'https://github.com/zhihz/openjev' },
|
| 308 |
-
m: { stars:
|
| 309 |
{ cat: 'decode', title: 'system-one (sgoedecke): batched single-token choice', by: { name: 'Sean Goedecke', handle: null },
|
| 310 |
desc: 'Batched single-token choice inference for open models, wire-compatible with TypeSafe.',
|
| 311 |
links: { gh: 'https://github.com/sgoedecke/system-one' },
|
| 312 |
-
m: { stars:
|
| 313 |
-
{ cat: 'decode', title: 'reflex: stock Qwen3.5 / Qwen3-VL behind /v1/systemone', by: { name: '
|
| 314 |
-
desc: '
|
| 315 |
-
|
| 316 |
-
|
|
|
|
| 317 |
{ cat: 'decode', title: 'LitJev: Jev-schema decision layer for the Qwen family', by: { name: 'Zhengxu Yu', handle: 'yuzxfred' },
|
| 318 |
desc: 'Shared prefill, forked KV/recurrent state, one cached branch per question, option logits read from lm_head on any Qwen checkpoint (default Qwen3.8-27B), behind the exact Jev /v1/systemone schema. Playground, MMLU-Pro eval, optional temperature calibration.',
|
| 319 |
quote: 'Qwen3.8-27B,10 道 MMLU-Pro 一次完成,0.3 s。比让模型只生成答案快 2.9x。非官方,基于公开信息。',
|
| 320 |
links: { tweet: 'https://x.com/yuzxfred/status/2100652136878981337', gh: 'https://github.com/zhengxuyu/litjev', blog: 'https://x.com/yuzxfred/status/2100896621173047367' },
|
| 321 |
-
m: { likes: 2, rts: 0, views:
|
| 322 |
|
| 323 |
/* ---------------- trained ---------------- */
|
| 324 |
-
{ cat: 'decode', title: 'snapjudge: TypeSafe-compatible decisions on a Mac, scored against Jev', by: { name: 'Michael Gross', handle:
|
| 325 |
desc: 'Choice / Score / Noul behind /v1/systemone on MLX Qwen3.5-family models, probabilities read straight from the logits. On TypeSafe\'s 20 public cases (373 decisions) Qwen3.8-27B reaches 81.2% agreement vs Jev\'s 87.7%: level on short texts, behind on invoices.',
|
| 326 |
-
links: { gh: 'https://github.com/Micha0827/snapjudge' },
|
| 327 |
-
m: { stars: 0 }, date: '2026-09-18', tags: 'mlx apple silicon qwen3.6 qwen3.8 systemone api benchmark agreement german runner game' },
|
| 328 |
{ cat: 'trained', title: 'Bespoke Nimble: open data, open model, open recipe', by: { name: 'Mahesh Sathiamoorthy', handle: 'madiator' },
|
| 329 |
desc: 'The first fully open recipe: LoRA on Qwen3.5-9B with the data curation published too. Contrastive curation flips one fact to make negatives, so training needs no probabilities. 90.1% agreement with Jev vs 66.4% for base Qwen.',
|
| 330 |
quote: 'Introducing Bespoke Nimble: an open data, open model, open recipe for an open Jev.',
|
| 331 |
links: { tweet: 'https://x.com/madiator/status/2100990591215783946', gh: 'https://github.com/bespokelabsai/nimble', hf: 'https://huggingface.co/bespokelabs/Bespoke-Nimble-9B' },
|
| 332 |
-
m: { likes:
|
| 333 |
{ cat: 'trained', title: 'jevlike: train your own one-pass option scorer', by: { name: 'Vinny LaRouge', handle: 'vinnylarouge' },
|
| 334 |
desc: 'A tiny byte encoder (or frozen HF encoder) plus an option head maps context + N options to one probability each. Doom, chess and Wikispeedia demos; parent of several forks.',
|
| 335 |
quote: 'I reverse-engineered a jev-like architecture given its type. You can find the repo here to train your own jevlikes.',
|
| 336 |
links: { tweet: 'https://x.com/vinnylarouge/status/2100170846346097083', gh: 'https://github.com/vinnylarouge/jevlike', demo: 'https://x.com/vinnylarouge/status/2100281651930513460' },
|
| 337 |
-
m: { likes:
|
| 338 |
{ cat: 'trained', title: 'openjev (AlexWortega): Qwen3.5 + MLP head, real weights', by: { name: 'Alex Wortega', handle: 'justALEXWORTEGA' },
|
| 339 |
desc: 'The only trained "openjev" checkpoint on the Hub: Qwen3.5-4B as an NLI cross-encoder, argmax entailment is the move. Plays Doom zero-shot from text and pixels; 35B-A3B heads included.',
|
| 340 |
quote: 'Typesafe: pnewed 💨 Jev: liberated 🫡 I trained an MLP on top of qwen 4b and it works literally like JEV',
|
| 341 |
links: { tweet: 'https://x.com/justALEXWORTEGA/status/2100341039986798930', hf: 'https://huggingface.co/AlexWortega/openjev', gh: 'https://github.com/mikesmullin/openjev' },
|
| 342 |
-
m: { likes:
|
| 343 |
{ cat: 'trained', title: 'decider-2b: full fine-tune of Qwen3.5-2B-Base', by: { name: 'Mapika', handle: null },
|
| 344 |
desc: 'Full fine-tune of Qwen3.5-2B-Base on 942k examples from 64 decision datasets, plus a vision variant. Weights on the Hub.',
|
| 345 |
links: { gh: 'https://github.com/Mapika/decider', hf: 'https://huggingface.co/Mapika/decider-2b', hf2: 'https://huggingface.co/Mapika/decider-2b-vision' },
|
| 346 |
-
m: { stars:
|
| 347 |
{ cat: 'trained', title: 'rlcd-modernbert-151m ("OpenJev Verdict")', by: { name: 'heman10x', handle: null },
|
| 348 |
desc: 'Post-trained ModernBERT (151M) scoring 25 candidate slots, abstention included, in under 35ms. WebGPU playground, Apache-2.0.',
|
| 349 |
links: { hf: 'https://huggingface.co/heman10x/rlcd-modernbert-151m', gh: 'https://github.com/Heman10x-NGU/Verdict-open-jev' },
|
| 350 |
-
m: { hfLikes:
|
| 351 |
{ cat: 'trained', title: 'open-jev (pngwn): Qwen3.5-4B decision scorer + demo Space', by: { name: 'pngwn', handle: 'evilpingwin' },
|
| 352 |
desc: 'Qwen3.5-4B LoRA fine-tune that scores each candidate answer for a typed question, with temperature scaling so confidence tracks accuracy. Weights (v1, v2b) and a demo Space on the Hub.',
|
| 353 |
quote: 'Since everyone is open-jevving. I set ML Intern (on hugging chat) to work. After a few experiments and a little guidance and we did ok.',
|
| 354 |
links: { tweet: 'https://x.com/evilpingwin/status/2100721265107751121', space: 'https://huggingface.co/spaces/pngwn/open-jev', hf: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer-v2b', hf2: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer' },
|
| 355 |
-
m: { likes:
|
| 356 |
{ cat: 'trained', title: 'jevbetter: a stronger jevlike', by: { name: 'olanotolu', handle: null },
|
| 357 |
desc: 'jevlike with a fancier head: hashed n-gram encoder, rival-aware attention, temperature scaling. Benchmarked against jevlike.',
|
| 358 |
links: { gh: 'https://github.com/olanotolu/jevbetter' },
|
| 359 |
-
m: { stars:
|
| 360 |
{ cat: 'trained', title: 'system-one-open: Gemma 4 E2B / Gemma 3 270M + head', by: { name: 'mithalouni', handle: null },
|
| 361 |
desc: 'Typed calibrated decisions in one pass on Gemma 4 E2B and Gemma 3 270M, trained on Modal.',
|
| 362 |
links: { gh: 'https://github.com/mithalouni/system-one-open' },
|
| 363 |
-
m: { stars:
|
| 364 |
{ cat: 'trained', title: 'qwen-rlcd: Qwen3.5-0.8B Choice/Score/Noul', by: { name: 'shamazharikh', handle: null },
|
| 365 |
desc: 'Choice / Score / Noul on Qwen3.5-0.8B. Inference works; training marked on hold.',
|
| 366 |
links: { gh: 'https://github.com/shamazharikh/qwen-rlcd' },
|
|
@@ -368,11 +369,11 @@ const ITEMS = [
|
|
| 368 |
{ cat: 'trained', title: 'system-one-gemma: Gemma 3 270M scoring head', by: { name: 'akash-kamat', handle: null },
|
| 369 |
desc: 'Gemma 3 270M with a scoring head; single forward pass, no text generation.',
|
| 370 |
links: { gh: 'https://github.com/akash-kamat/system-one-gemma' },
|
| 371 |
-
m: { stars:
|
| 372 |
{ cat: 'trained', title: 'Laya: 421M RLCD decision model + demo Space', by: { name: 'Convai Innovations', handle: null },
|
| 373 |
desc: 'ModernBERT-large fine-tuned plus a decision head trained from scratch (421M total) with RLCD against proper scoring rules: choice/score/noul with calibrated probabilities, all questions in one pass. Weights and a 7-tab Gradio demo on the Hub.',
|
| 374 |
links: { gh: 'https://github.com/NandhaKishorM/laya', hf: 'https://huggingface.co/convaiinnovations/laya', space: 'https://huggingface.co/spaces/convaiinnovations/laya-demo', blog: 'https://dev.to/nandakishor_m_6cc0adfde9f/i-built-non-autoregressive-decision-models-a-year-ago-then-a-frontier-lab-called-it-a-18me', site: 'https://www.reddit.com/r/LocalLLaMA/comments/1wjieap/made_the_horizontal_opensource_model_for_jev_with/' },
|
| 375 |
-
m: { stars:
|
| 376 |
{ cat: 'trained', title: 'System One Mini: 69M DistilBERT, five fixed decisions', by: { name: 'David Hatley', handle: null },
|
| 377 |
desc: '69M DistilBERT with five fixed typed heads over synthetic software-diagnosis summaries. Its card says: not a reproduction of Jev or RLCD.',
|
| 378 |
links: { hf: 'https://huggingface.co/DavidHatley/system-one-mini' },
|
|
@@ -380,86 +381,86 @@ const ITEMS = [
|
|
| 380 |
{ cat: 'trained', title: 'NanoJev: 0.6B Qwen3 with a real decision head', by: { name: 'TianyuCodings', handle: null },
|
| 381 |
desc: 'A 0.6B Qwen3 backbone with an actual trained head emitting Choice / Boolean / Score across 2–255 candidates in one forward pass. Full training pipeline, maze and snake runs, weights and data on the Hub. The one Chinese and Korean roundups are pushing hardest.',
|
| 382 |
links: { gh: 'https://github.com/TianyuCodings/NanoJev', hf: 'https://huggingface.co/C-Tianyu/NanoJev' },
|
| 383 |
-
m: { stars:
|
| 384 |
{ cat: 'trained', title: 'mini-jev: a preregistered measurement of the trick', by: { name: 'r-ms', handle: null },
|
| 385 |
desc: 'Not another clone: a preregistered study of letter-logit versus grammar-JSON decoding on frozen Qwen3-4B and CLINC150. An actual measurement, no weights trained.',
|
| 386 |
links: { gh: 'https://github.com/r-ms/mini-jev' },
|
| 387 |
-
m: { stars:
|
| 388 |
|
| 389 |
{ cat: 'trained', promised: true, title: 'Archer Hume\'s open-weight Jev (promised)', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 390 |
desc: 'Announced alongside the architecture teardown: an open-weight version trained on the inferred design. Training in progress, nothing released yet.',
|
| 391 |
quote: 'I decided I\'d release an open-weight version. BUT training takes time, so while we all wait I decided I\'d drop the sauce.',
|
| 392 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 393 |
-
m: { likes:
|
| 394 |
|
| 395 |
/* ---------------- prior ---------------- */
|
| 396 |
{ cat: 'prior', title: 'GLiNER 2.5: "Weights are on Hugging Face."', by: { name: 'George (Fastino / GLiNER)', handle: 'george_onx' },
|
| 397 |
desc: 'Points the launch thread at the GLiNER 2.5 collection: a pretrained zero-shot encoder-classifier already doing single-pass classification over caller-supplied labels.',
|
| 398 |
quote: 'We\'ve been reading the comments. Weights are on Hugging Face.',
|
| 399 |
links: { tweet: 'https://x.com/george_onx/status/2100293114808119379', hf: 'https://huggingface.co/collections/fastino/gliner25-models' },
|
| 400 |
-
m: { likes:
|
| 401 |
{ cat: 'prior', title: 'DeepMost sales-conversion RL model + papers', by: { name: 'Nandakishor M', handle: 'Nandakishorm1' },
|
| 402 |
desc: 'Claims the idea shipped in 2025 papers (arXiv 2503.23303, 2510.01237) with a Hub model and dataset. PPO over embeddings; loosely related.',
|
| 403 |
quote: 'Also the second work published in September 2025 was exactly the same one jev proposed now.',
|
| 404 |
links: { tweet: 'https://x.com/Nandakishorm1/status/2100451360873611666', hf: 'https://huggingface.co/DeepMostInnovations/sales-conversion-model-reinf-learning', paper: 'https://arxiv.org/abs/2510.01237' },
|
| 405 |
-
m: { likes:
|
| 406 |
|
| 407 |
/* ---------------- arch ---------------- */
|
| 408 |
{ cat: 'explain', title: 'Jev\'s Architecture Unmasked', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 409 |
desc: 'Argues Jev is a causal transformer with shared-state encoding, per-question branches and direct probability readouts, likely on a sparse MoE backbone, inferred from latency probing.',
|
| 410 |
quote: 'All the grifters are completely wrong about Jev\'s architecture so I decided I\'d release an open-weight version. BUT training takes time…',
|
| 411 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 412 |
-
m: { likes:
|
| 413 |
|
| 414 |
/* ---------------- diffusion ---------------- */
|
| 415 |
{ cat: 'diffusion', title: 'DiffusionGemma-as-Jev + vLLM structured-generation patch', by: { name: 'Matt Mastracci', handle: 'mmastrac' },
|
| 416 |
desc: 'Adds a Jev-like structured mode for DiffusionGemma to vLLM (PR #57250) and runs live evals: not slower than Jev\'s API and roughly tied on quality.',
|
| 417 |
quote: 'Is Jev faster than DiffusionGemma? No ❌ Is Jev smarter than DiffusionGemma? No ❌ (they\'re roughly tied!)',
|
| 418 |
links: { tweet: 'https://x.com/mmastrac/status/2100626193943052784', pr: 'https://github.com/vllm-project/vllm/pull/57250' },
|
| 419 |
-
m: { likes:
|
| 420 |
{ cat: 'diffusion', title: 'razorback16/openjev + api.codiv.ai host', by: { name: 'razorback16', handle: null },
|
| 421 |
desc: 'The same vLLM PR #57250 route, packaged as a Jev-wire /v1/systemone service: a DiffusionGemma 26B-A4B canvas with only the label slots left as noise, one read-only denoise step reads out the distribution. Public host at api.codiv.ai.',
|
| 422 |
links: { gh: 'https://github.com/razorback16/openjev', site: 'https://api.codiv.ai/v1/systemone' },
|
| 423 |
-
m: { stars:
|
| 424 |
{ cat: 'diffusion', title: 'open-jev (JoshuaSP): typed JSON with DiffusionGemma', by: { name: 'Joshua SP', handle: null },
|
| 425 |
desc: 'Typed JSON inference with DiffusionGemma, benchmarked against Every and Jev.',
|
| 426 |
links: { gh: 'https://github.com/JoshuaSP/open-jev' },
|
| 427 |
-
m: { stars:
|
| 428 |
{ cat: 'diffusion', title: 'LocalJev (GitHub Next): oMLX Jev API + model bake-off', by: { name: 'GitHub Next', handle: 'GitHubNext' },
|
| 429 |
desc: 'Official GitHub Next bridge, since not everyone there has Jev access: a Bun server putting oMLX DiffusionGemma behind /v1/systemone. Prompts for JSON probabilities instead of reading logits: wire-compatible, not equivalent. Ships a 1,200-request bake-off.',
|
| 430 |
quote: 'Not everyone on the team has access to Jev yet. Spent a morning cobbling together a poor man\'s Jev on top of omlx for local use.',
|
| 431 |
links: { tweet: 'https://x.com/GitHubNext/status/2101193436816920798', gh: 'https://github.com/githubnext/localjev' },
|
| 432 |
-
m: { likes:
|
| 433 |
|
| 434 |
/* ---------------- explain ---------------- */
|
| 435 |
{ cat: 'explain', title: '45-second TL;DR on Jev (video)', by: { name: 'Matija Sosic', handle: 'MatijaSosic' },
|
| 436 |
desc: 'The most-shared explainer of the launch: a 45-second video that strips Jev down to its core idea, made because the official launch video was hard to follow.',
|
| 437 |
quote: 'I find the core idea beautifully simple, but the video made it really hard to understand.',
|
| 438 |
links: { tweet: 'https://x.com/MatijaSosic/status/2100190746389135772' },
|
| 439 |
-
m: { likes:
|
| 440 |
{ cat: 'explain', title: 'typesafe-ai-benchmark: an "imposter Jev" on Cerebras', by: { name: 'Shannon Duncan', handle: 'iamMrDuncan' },
|
| 441 |
desc: 'An "imposter Jev" gateway on Cerebras with Qwen 3.8 27B, compared side by side. Similar quality and speed; TypeSafe far cheaper.',
|
| 442 |
quote: 'Built an alternative version of @typesafeai but on @cerebras with Qwen 3.8 27b. Similar quality, similar performance, but vastly different cost.',
|
| 443 |
links: { tweet: 'https://x.com/iamMrDuncan/status/2100467548298899918', gh: 'https://github.com/iammrduncan/typesafe-ai-benchmark' },
|
| 444 |
-
m: { likes:
|
| 445 |
{ cat: 'explain', title: 'awesome-jev lists (AnotiaWang · hellogumbo)', by: { name: 'AnotiaWang · hellogumbo', handle: null },
|
| 446 |
desc: 'Two curated indexes; AnotiaWang\'s "Research & Open Models" section is the tight one, hellogumbo\'s catches the long tail.',
|
| 447 |
links: { gh: 'https://github.com/AnotiaWang/awesome-jev', gh2: 'https://github.com/hellogumbo/awesome-jev' },
|
| 448 |
-
m: { stars:
|
| 449 |
{ cat: 'explain', title: 'jev-benchmarks: probability-aware eval harness', by: { name: 'AbdelStark', handle: null },
|
| 450 |
desc: 'Calibration, selective risk and latency evals for typed decision models, Jev or open.',
|
| 451 |
links: { gh: 'https://github.com/AbdelStark/jev-benchmarks' },
|
| 452 |
-
m: { stars:
|
| 453 |
{ cat: 'explain', title: 'Thread: System 1 / System 2 agents and the open recreations', by: { name: 'hhkkmon', handle: 'hhkkmon' },
|
| 454 |
desc: '14-part explainer framing agents as "expensive LLM calls pretending to be if statements", listing the open recreations at the end.',
|
| 455 |
quote: 'Big models think. Small decision models react.',
|
| 456 |
links: { tweet: 'https://x.com/hhkkmon/status/2100443314957038010' },
|
| 457 |
-
m: { likes: 2, rts: 0, views:
|
| 458 |
{ cat: 'explain', title: 'Long-form: "OpenJev is NOT Jev open-sourced"', by: { name: 'mygtmhire', handle: 'mygtmhire' },
|
| 459 |
desc: 'Most thorough single post on TheoLeeCJ/openjev, including the warning that its probabilities are not calibrated confidence.',
|
| 460 |
quote: 'State in → probabilities out → code acts. No conversation required.',
|
| 461 |
links: { tweet: 'https://x.com/mygtmhire/status/2100666488625734103' },
|
| 462 |
-
m: { likes:
|
| 463 |
{ cat: 'explain', title: 'jev-behavior-study: probing the closed model', by: { name: 'RINNECODER', handle: null },
|
| 464 |
desc: 'Black-box behavioural study of the closed Jev 1.13.0 itself. Not a replica — it measures the thing everyone is trying to reproduce.',
|
| 465 |
links: { gh: 'https://github.com/RINNECODER/jev-behavior-study' },
|
|
@@ -510,7 +511,7 @@ const likesOf = it => it.m?.likes ?? -1;
|
|
| 510 |
brand-new entry nobody has reacted to yet cannot climb on its date alone.
|
| 511 |
The credit therefore rewards "new AND real", never novelty by itself. */
|
| 512 |
const W = { likes: 1, stars: 5, hfLikes: 8, views: 1/500 };
|
| 513 |
-
const RECENCY = { credit:
|
| 514 |
const BASE_DATE = '2026-09-19'; /* the snapshot the metrics were last refreshed against */
|
| 515 |
const ageDays = it => Math.max(0, (new Date(BASE_DATE) - new Date(it.date)) / 864e5);
|
| 516 |
const recency = it => {
|
|
|
|
| 250 |
desc: 'Batches every JSON key at once and reads category probabilities from stock Qwen 2.5 1B logits, ~5× faster than generating JSON.',
|
| 251 |
quote: 'They were building in stealth for 2 years, I was building in stealth for 2 hours… No new training required.',
|
| 252 |
links: { tweet: 'https://x.com/harshagundal/status/2100044305536889015', hf: 'https://huggingface.co/harshatheg/Qwen-2.5-1B-RLCD', space: 'https://huggingface.co/spaces/drinkmoonshine/parallel-constrained-decoding' },
|
| 253 |
+
m: { likes: 5796, rts: 355, views: 1269064, hfLikes: 429 }, date: '2026-09-16', tags: 'qwen json mac m4 on-device forks epsilon3 shreyansh26 drinkmoonshine' },
|
| 254 |
{ cat: 'decode', title: 'openjev-sglang: Jev-compatible public API', by: { name: 'Eric Zhang', handle: 'ekzhang1' },
|
| 255 |
desc: 'A drop-in /v1/systemone endpoint over Qwen3.6-35B-A3B. SGLang\'s radix cache keeps the shared prefill hot, so 64 parallel tasks finish in under a second. Prefill-only, no generation.',
|
| 256 |
quote: 'It runs a comparable open model (Qwen3.6-35B-A3B), and just uses SGLang radix cache to preserve the prefill reuse — 64 tasks in <1s.',
|
| 257 |
links: { tweet: 'https://x.com/ekzhang1/status/2100651678110515383', gh: 'https://github.com/ekzhang/openjev-sglang' },
|
| 258 |
+
m: { likes: 977, rts: 60, views: 139265, stars: 196 }, date: '2026-09-17', tags: 'sglang qwen3.6 moe api systemone radix cache' },
|
| 259 |
{ cat: 'decode', title: 'openjev + openjev.com', by: { name: 'Theo Lee', handle: 'theoleecj' },
|
| 260 |
desc: 'Reads option logits on Qwen3.5-4B instead of generating text: 21 decisions in 1.02s vs 5.33s for JSON. CUDA plus a WebGPU browser demo.',
|
| 261 |
quote: 'On the Jev waitlist? You can use not Jev (but close enough) right here.',
|
| 262 |
links: { tweet: 'https://x.com/theoleecj/status/2100243609060856264', gh: 'https://github.com/TheoLeeCJ/openjev', demo: 'https://openjev.com' },
|
| 263 |
+
m: { likes: 10, rts: 0, views: 3291, stars: 1976 }, date: '2026-09-16', tags: 'webgpu browser rtx 3090 qwen3.5 logits mit' },
|
| 264 |
{ cat: 'decode', title: 'PocketJev: on-device iPhone visual decisions', by: { name: 'aqty', handle: 'aqty' },
|
| 265 |
desc: 'Qwen3-VL running locally on an iPhone via MLX. Camera frame + question + 3 options → direct option logits in ~1s, no photos saved.',
|
| 266 |
quote: 'iPhone上でQwen 2Bをローカル実行して、カメラ画像+質問+3択から文章生成なしで直接判定。',
|
| 267 |
links: { tweet: 'https://x.com/aqty/status/2100549517171175807', gh: 'https://github.com/NullPo-jp/PocketJev' },
|
| 268 |
+
m: { likes: 6, rts: 4, views: 424, stars: 0 }, date: '2026-09-17', tags: 'ios iphone mlx vision qwen3-vl camera japan' },
|
| 269 |
{ cat: 'decode', title: 'jevmlx: Jev-style decisions for any MLX model', by: { name: 'Beni', handle: 'beni_il_' },
|
| 270 |
desc: 'Typed schema fields broadcast across the KV cache, one batch row per field, all scored in a single forward pass on any MLX model.',
|
| 271 |
quote: 'If you can run Qwen on your machine, it can behave very similarly to Jev.',
|
| 272 |
links: { tweet: 'https://x.com/beni_il_/status/2100617387116568956', gh: 'https://github.com/bnsd55/jevmlx' },
|
| 273 |
+
m: { likes: 7, rts: 0, views: 1183, stars: 40 }, date: '2026-09-17', tags: 'mlx apple silicon kv cache hebrew' },
|
| 274 |
{ cat: 'decode', title: 'jev-on-a-laptop: the honest benchmark', by: { name: 'Richard Becker', handle: 'RichardBcker1' },
|
| 275 |
desc: 'Qwen 2.5 1.5B/7B and Qwen3 8B on a laptop vs Jev\'s published cases: 7B hits 73.8% agreement to Jev\'s 86.6%, 0.4–2s per decision. Finds confidence does not reliably flag errors.',
|
| 276 |
quote: 'All code, benchmarks, model-size comparisons and the full write-up (unofficial, no affiliation with TypeSafe).',
|
| 277 |
links: { tweet: 'https://x.com/RichardBcker1/status/2100195276107530393', gh: 'https://github.com/rorshopping/jev-on-a-laptop', space: 'https://huggingface.co/spaces/rorshopping/parallel-constrained-decisions' },
|
| 278 |
+
m: { likes: 1, rts: 0, views: 67, stars: 16 }, date: '2026-09-16', tags: 'benchmark study agreement fraud apple silicon m5 writeup explainer' },
|
| 279 |
{ cat: 'decode', title: 'jevfire: one context, many decisions on CUDA', by: { name: 'Enrique Alcázar', handle: 'enriquealcazarg' },
|
| 280 |
desc: 'Parallel decisions for CUDA LLMs behind a vLLM-style API, with game-agent examples. Claims 10×+ over regular inference, runs in the browser, accepts images.',
|
| 281 |
quote: 'Here you have an open-source approach to the same problem… uses the same principles as JEV, is type-safe, runs in your browser, and accepts images.',
|
| 282 |
links: { tweet: 'https://x.com/enriquealcazarg/status/2100267330022375858', gh: 'https://github.com/kikoncuo/jevfire', site: 'https://kikoncuo.github.io/jevfire/' },
|
| 283 |
+
m: { likes: 1, rts: 0, views: 250, stars: 8 }, date: '2026-09-16', tags: 'cuda vllm browser images games primeagen' },
|
| 284 |
{ cat: 'decode', title: 'LFM2.5-RLCD (350M & 2.6B): PCD on Liquid models', by: { name: 'notnotsamuel · monotykamary', handle: null },
|
| 285 |
desc: 'Parallel constrained decoding on unchanged Liquid LFM2.5 weights, branching attention and convolution state. Inference-only: 8–63× faster, field accuracy still ~60%.',
|
| 286 |
links: { hf: 'https://huggingface.co/notnotsamuel/LFM2.5-350M-RLCD', hf2: 'https://huggingface.co/monotykamary/LFM2.5-2.6B-RLCD' },
|
| 287 |
+
m: { hfLikes: 18 }, date: '2026-09-16', tags: 'liquid lfm hub weights renamed inference only pcd' },
|
| 288 |
{ cat: 'decode', title: 'parallelConstraintDecoding (Java + Python)', by: { name: 'Stephan Janssen', handle: 'Stephan007' },
|
| 289 |
desc: 'A Java port of the decode trick alongside a Python reference.',
|
| 290 |
quote: 'Here\'s a java version of Jev 🤩',
|
| 291 |
links: { tweet: 'https://x.com/Stephan007/status/2100673219967557906', gh: 'https://github.com/stephanj/parallelConstraintDecoding' },
|
| 292 |
+
m: { likes: 3, rts: 0, views: 103, stars: 10 }, date: '2026-09-17', tags: 'java jvm port' },
|
| 293 |
{ cat: 'decode', title: 'system-one (snellingio): "System One Lite"', by: { name: 'snellingio', handle: null },
|
| 294 |
desc: 'Choice / Score / Noul server over MLX Qwen3, mirroring TypeSafe\'s three primitives. No launch post found.',
|
| 295 |
links: { gh: 'https://github.com/snellingio/system-one' },
|
| 296 |
+
m: { stars: 38 }, date: '2026-09-16', tags: 'mlx qwen3 choice score noul server' },
|
| 297 |
{ cat: 'decode', title: 'open-jev (daseinlabs): Gemma 3 4B on MLX', by: { name: 'Dasein Labs', handle: null },
|
| 298 |
desc: 'Prefix-cache option scoring on a local Gemma 3 4B via MLX, a Doom demo, and a /v1/systemone shim. Only seen in roundups.',
|
| 299 |
links: { gh: 'https://github.com/daseinlabs/open-jev' },
|
| 300 |
+
m: { stars: 54 }, date: '2026-09-17', tags: 'gemma mlx doom systemone shim' },
|
| 301 |
{ cat: 'decode', title: 'openvons (open-Jev, Japanese)', by: { name: 'genai-craft', handle: null },
|
| 302 |
desc: 'Decision layer answering finite-option questions with probabilities, from text, images or Japanese voice commands.',
|
| 303 |
links: { gh: 'https://github.com/genai-craft/openvons' },
|
| 304 |
+
m: { stars: 9 }, date: '2026-09-17', tags: 'japanese voice image multimodal' },
|
| 305 |
{ cat: 'decode', title: 'openjev (zhihz): bilingual local decisions', by: { name: 'zhihz', handle: null },
|
| 306 |
desc: 'Local bilingual probability decisions from context, questions and candidate answers.',
|
| 307 |
links: { gh: 'https://github.com/zhihz/openjev' },
|
| 308 |
+
m: { stars: 14 }, date: '2026-09-16', tags: 'chinese bilingual local' },
|
| 309 |
{ cat: 'decode', title: 'system-one (sgoedecke): batched single-token choice', by: { name: 'Sean Goedecke', handle: null },
|
| 310 |
desc: 'Batched single-token choice inference for open models, wire-compatible with TypeSafe.',
|
| 311 |
links: { gh: 'https://github.com/sgoedecke/system-one' },
|
| 312 |
+
m: { stars: 14 }, date: '2026-09-17', tags: 'compatible batched single token' },
|
| 313 |
+
{ cat: 'decode', title: 'reflex: stock Qwen3.5 / Qwen3-VL behind /v1/systemone', by: { name: 'Kshetrajna Raghavan', handle: 'kshetrajna' },
|
| 314 |
+
desc: 'Runs Jev\'s shape on stock Qwen3.5: choice / score / noul in one batched pass, a /v1/systemone server, and a WebGPU demo on a 0.8B model. Temperature scaling gets Qwen3.5-4B to 0.039 ECE vs Jev\'s 0.031; a LoRA recipe trains honest probabilities on public data.',
|
| 315 |
+
quote: '"Wonder if we could build that?" is a pretty normal response to new tech at Shopify. Jev got me curious, so I built Reflex: a Qwen-based experiment in structured decisions + probabilities running on WebGPU.',
|
| 316 |
+
links: { tweet: 'https://x.com/kshetrajna/status/2100739853101195744', gh: 'https://github.com/kshetrajna12/reflex', demo: 'https://kshetrajna12.github.io/reflex/' },
|
| 317 |
+
m: { likes: 330, rts: 22, views: 138281, stars: 77 }, date: '2026-09-17', tags: 'systemone choice score noul qwen3-vl image shared state server webgpu browser demo lora calibration ece shopify mmlu' },
|
| 318 |
{ cat: 'decode', title: 'LitJev: Jev-schema decision layer for the Qwen family', by: { name: 'Zhengxu Yu', handle: 'yuzxfred' },
|
| 319 |
desc: 'Shared prefill, forked KV/recurrent state, one cached branch per question, option logits read from lm_head on any Qwen checkpoint (default Qwen3.8-27B), behind the exact Jev /v1/systemone schema. Playground, MMLU-Pro eval, optional temperature calibration.',
|
| 320 |
quote: 'Qwen3.8-27B,10 道 MMLU-Pro 一次完成,0.3 s。比让模型只生成答案快 2.9x。非官方,基于公开信息。',
|
| 321 |
links: { tweet: 'https://x.com/yuzxfred/status/2100652136878981337', gh: 'https://github.com/zhengxuyu/litjev', blog: 'https://x.com/yuzxfred/status/2100896621173047367' },
|
| 322 |
+
m: { likes: 2, rts: 0, views: 148, stars: 19 }, date: '2026-09-17', tags: 'qwen qwen3.8 27b systemone schema compatible calibration mmlu-pro doom chess vision transformers chinese explainer writeup limitations' },
|
| 323 |
|
| 324 |
/* ---------------- trained ---------------- */
|
| 325 |
+
{ cat: 'decode', title: 'snapjudge: TypeSafe-compatible decisions on a Mac, scored against Jev', by: { name: 'Michael Gross', handle: 'michagross' },
|
| 326 |
desc: 'Choice / Score / Noul behind /v1/systemone on MLX Qwen3.5-family models, probabilities read straight from the logits. On TypeSafe\'s 20 public cases (373 decisions) Qwen3.8-27B reaches 81.2% agreement vs Jev\'s 87.7%: level on short texts, behind on invoices.',
|
| 327 |
+
links: { tweet: 'https://x.com/michagross/status/2101062157290856604', gh: 'https://github.com/Micha0827/snapjudge' },
|
| 328 |
+
m: { stars: 3, likes: 1, rts: 0, views: 11 }, date: '2026-09-18', tags: 'mlx apple silicon qwen3.6 qwen3.8 systemone api benchmark agreement german runner game' },
|
| 329 |
{ cat: 'trained', title: 'Bespoke Nimble: open data, open model, open recipe', by: { name: 'Mahesh Sathiamoorthy', handle: 'madiator' },
|
| 330 |
desc: 'The first fully open recipe: LoRA on Qwen3.5-9B with the data curation published too. Contrastive curation flips one fact to make negatives, so training needs no probabilities. 90.1% agreement with Jev vs 66.4% for base Qwen.',
|
| 331 |
quote: 'Introducing Bespoke Nimble: an open data, open model, open recipe for an open Jev.',
|
| 332 |
links: { tweet: 'https://x.com/madiator/status/2100990591215783946', gh: 'https://github.com/bespokelabsai/nimble', hf: 'https://huggingface.co/bespokelabs/Bespoke-Nimble-9B' },
|
| 333 |
+
m: { likes: 1210, rts: 151, views: 415617, stars: 509, hfLikes: 59 }, date: '2026-09-18', tags: 'bespokelabs nimble qwen3.5 9b lora contrastive data curation open recipe synthetic distillation free parallel scorer 100ms h100 macbook' },
|
| 334 |
{ cat: 'trained', title: 'jevlike: train your own one-pass option scorer', by: { name: 'Vinny LaRouge', handle: 'vinnylarouge' },
|
| 335 |
desc: 'A tiny byte encoder (or frozen HF encoder) plus an option head maps context + N options to one probability each. Doom, chess and Wikispeedia demos; parent of several forks.',
|
| 336 |
quote: 'I reverse-engineered a jev-like architecture given its type. You can find the repo here to train your own jevlikes.',
|
| 337 |
links: { tweet: 'https://x.com/vinnylarouge/status/2100170846346097083', gh: 'https://github.com/vinnylarouge/jevlike', demo: 'https://x.com/vinnylarouge/status/2100281651930513460' },
|
| 338 |
+
m: { likes: 2018, rts: 168, views: 164627, stars: 999 }, date: '2026-09-16', tags: 'train scorer encoder doom chess wikispeedia rl' },
|
| 339 |
{ cat: 'trained', title: 'openjev (AlexWortega): Qwen3.5 + MLP head, real weights', by: { name: 'Alex Wortega', handle: 'justALEXWORTEGA' },
|
| 340 |
desc: 'The only trained "openjev" checkpoint on the Hub: Qwen3.5-4B as an NLI cross-encoder, argmax entailment is the move. Plays Doom zero-shot from text and pixels; 35B-A3B heads included.',
|
| 341 |
quote: 'Typesafe: pnewed 💨 Jev: liberated 🫡 I trained an MLP on top of qwen 4b and it works literally like JEV',
|
| 342 |
links: { tweet: 'https://x.com/justALEXWORTEGA/status/2100341039986798930', hf: 'https://huggingface.co/AlexWortega/openjev', gh: 'https://github.com/mikesmullin/openjev' },
|
| 343 |
+
m: { likes: 774, rts: 47, views: 103797, hfLikes: 213, stars: 2 }, date: '2026-09-16', tags: 'nli cross-encoder mlp qwen3.5 doom vision weights hub' },
|
| 344 |
{ cat: 'trained', title: 'decider-2b: full fine-tune of Qwen3.5-2B-Base', by: { name: 'Mapika', handle: null },
|
| 345 |
desc: 'Full fine-tune of Qwen3.5-2B-Base on 942k examples from 64 decision datasets, plus a vision variant. Weights on the Hub.',
|
| 346 |
links: { gh: 'https://github.com/Mapika/decider', hf: 'https://huggingface.co/Mapika/decider-2b', hf2: 'https://huggingface.co/Mapika/decider-2b-vision' },
|
| 347 |
+
m: { stars: 96, hfLikes: 49 }, date: '2026-09-16', tags: 'fine-tune qwen3.5 2b calibrated vision gh200 hub weights' },
|
| 348 |
{ cat: 'trained', title: 'rlcd-modernbert-151m ("OpenJev Verdict")', by: { name: 'heman10x', handle: null },
|
| 349 |
desc: 'Post-trained ModernBERT (151M) scoring 25 candidate slots, abstention included, in under 35ms. WebGPU playground, Apache-2.0.',
|
| 350 |
links: { hf: 'https://huggingface.co/heman10x/rlcd-modernbert-151m', gh: 'https://github.com/Heman10x-NGU/Verdict-open-jev' },
|
| 351 |
+
m: { hfLikes: 12, stars: 25 }, date: '2026-09-17', tags: 'modernbert encoder webgpu abstention gliclass hub weights' },
|
| 352 |
{ cat: 'trained', title: 'open-jev (pngwn): Qwen3.5-4B decision scorer + demo Space', by: { name: 'pngwn', handle: 'evilpingwin' },
|
| 353 |
desc: 'Qwen3.5-4B LoRA fine-tune that scores each candidate answer for a typed question, with temperature scaling so confidence tracks accuracy. Weights (v1, v2b) and a demo Space on the Hub.',
|
| 354 |
quote: 'Since everyone is open-jevving. I set ML Intern (on hugging chat) to work. After a few experiments and a little guidance and we did ok.',
|
| 355 |
links: { tweet: 'https://x.com/evilpingwin/status/2100721265107751121', space: 'https://huggingface.co/spaces/pngwn/open-jev', hf: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer-v2b', hf2: 'https://huggingface.co/pngwn/system-one-qwen3.5-4b-scorer' },
|
| 356 |
+
m: { likes: 230, rts: 36, views: 13184, hfLikes: 10 }, date: '2026-09-17', tags: 'qwen3.5 scoring head temperature scaling calibration ml intern huggingchat hub weights space demo' },
|
| 357 |
{ cat: 'trained', title: 'jevbetter: a stronger jevlike', by: { name: 'olanotolu', handle: null },
|
| 358 |
desc: 'jevlike with a fancier head: hashed n-gram encoder, rival-aware attention, temperature scaling. Benchmarked against jevlike.',
|
| 359 |
links: { gh: 'https://github.com/olanotolu/jevbetter' },
|
| 360 |
+
m: { stars: 12 }, date: '2026-09-16', tags: 'n-gram attention temperature scaling calibration' },
|
| 361 |
{ cat: 'trained', title: 'system-one-open: Gemma 4 E2B / Gemma 3 270M + head', by: { name: 'mithalouni', handle: null },
|
| 362 |
desc: 'Typed calibrated decisions in one pass on Gemma 4 E2B and Gemma 3 270M, trained on Modal.',
|
| 363 |
links: { gh: 'https://github.com/mithalouni/system-one-open' },
|
| 364 |
+
m: { stars: 11 }, date: '2026-09-17', tags: 'gemma modal scoring head' },
|
| 365 |
{ cat: 'trained', title: 'qwen-rlcd: Qwen3.5-0.8B Choice/Score/Noul', by: { name: 'shamazharikh', handle: null },
|
| 366 |
desc: 'Choice / Score / Noul on Qwen3.5-0.8B. Inference works; training marked on hold.',
|
| 367 |
links: { gh: 'https://github.com/shamazharikh/qwen-rlcd' },
|
|
|
|
| 369 |
{ cat: 'trained', title: 'system-one-gemma: Gemma 3 270M scoring head', by: { name: 'akash-kamat', handle: null },
|
| 370 |
desc: 'Gemma 3 270M with a scoring head; single forward pass, no text generation.',
|
| 371 |
links: { gh: 'https://github.com/akash-kamat/system-one-gemma' },
|
| 372 |
+
m: { stars: 2 }, date: '2026-09-17', tags: 'gemma 270m tiny' },
|
| 373 |
{ cat: 'trained', title: 'Laya: 421M RLCD decision model + demo Space', by: { name: 'Convai Innovations', handle: null },
|
| 374 |
desc: 'ModernBERT-large fine-tuned plus a decision head trained from scratch (421M total) with RLCD against proper scoring rules: choice/score/noul with calibrated probabilities, all questions in one pass. Weights and a 7-tab Gradio demo on the Hub.',
|
| 375 |
links: { gh: 'https://github.com/NandhaKishorM/laya', hf: 'https://huggingface.co/convaiinnovations/laya', space: 'https://huggingface.co/spaces/convaiinnovations/laya-demo', blog: 'https://dev.to/nandakishor_m_6cc0adfde9f/i-built-non-autoregressive-decision-models-a-year-ago-then-a-frontier-lab-called-it-a-18me', site: 'https://www.reddit.com/r/LocalLLaMA/comments/1wjieap/made_the_horizontal_opensource_model_for_jev_with/' },
|
| 376 |
+
m: { stars: 1410, hfLikes: 577 }, date: '2026-09-18', tags: 'modernbert rlcd calibrated choice score noul support triage guardrails rag moderation routing hub weights space demo github pypi reddit devto' },
|
| 377 |
{ cat: 'trained', title: 'System One Mini: 69M DistilBERT, five fixed decisions', by: { name: 'David Hatley', handle: null },
|
| 378 |
desc: '69M DistilBERT with five fixed typed heads over synthetic software-diagnosis summaries. Its card says: not a reproduction of Jev or RLCD.',
|
| 379 |
links: { hf: 'https://huggingface.co/DavidHatley/system-one-mini' },
|
|
|
|
| 381 |
{ cat: 'trained', title: 'NanoJev: 0.6B Qwen3 with a real decision head', by: { name: 'TianyuCodings', handle: null },
|
| 382 |
desc: 'A 0.6B Qwen3 backbone with an actual trained head emitting Choice / Boolean / Score across 2–255 candidates in one forward pass. Full training pipeline, maze and snake runs, weights and data on the Hub. The one Chinese and Korean roundups are pushing hardest.',
|
| 383 |
links: { gh: 'https://github.com/TianyuCodings/NanoJev', hf: 'https://huggingface.co/C-Tianyu/NanoJev' },
|
| 384 |
+
m: { stars: 975, hfLikes: 35 }, date: '2026-09-17', tags: 'nanojev 0.6b qwen3 decision head choice boolean score maze snake training pipeline weights hub china korea' },
|
| 385 |
{ cat: 'trained', title: 'mini-jev: a preregistered measurement of the trick', by: { name: 'r-ms', handle: null },
|
| 386 |
desc: 'Not another clone: a preregistered study of letter-logit versus grammar-JSON decoding on frozen Qwen3-4B and CLINC150. An actual measurement, no weights trained.',
|
| 387 |
links: { gh: 'https://github.com/r-ms/mini-jev' },
|
| 388 |
+
m: { stars: 24 }, date: '2026-09-17', tags: 'preregistered study letter logits grammar json clinc150 qwen3-4b measurement' },
|
| 389 |
|
| 390 |
{ cat: 'trained', promised: true, title: 'Archer Hume\'s open-weight Jev (promised)', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 391 |
desc: 'Announced alongside the architecture teardown: an open-weight version trained on the inferred design. Training in progress, nothing released yet.',
|
| 392 |
quote: 'I decided I\'d release an open-weight version. BUT training takes time, so while we all wait I decided I\'d drop the sauce.',
|
| 393 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 394 |
+
m: { likes: 2242, rts: 212, views: 376500 }, date: '2026-09-17', tags: 'promised coming soon open-weight training archer' },
|
| 395 |
|
| 396 |
/* ---------------- prior ---------------- */
|
| 397 |
{ cat: 'prior', title: 'GLiNER 2.5: "Weights are on Hugging Face."', by: { name: 'George (Fastino / GLiNER)', handle: 'george_onx' },
|
| 398 |
desc: 'Points the launch thread at the GLiNER 2.5 collection: a pretrained zero-shot encoder-classifier already doing single-pass classification over caller-supplied labels.',
|
| 399 |
quote: 'We\'ve been reading the comments. Weights are on Hugging Face.',
|
| 400 |
links: { tweet: 'https://x.com/george_onx/status/2100293114808119379', hf: 'https://huggingface.co/collections/fastino/gliner25-models' },
|
| 401 |
+
m: { likes: 1584, rts: 94, views: 224123 }, date: '2026-09-16', tags: 'gliner fastino zero-shot encoder classifier collection' },
|
| 402 |
{ cat: 'prior', title: 'DeepMost sales-conversion RL model + papers', by: { name: 'Nandakishor M', handle: 'Nandakishorm1' },
|
| 403 |
desc: 'Claims the idea shipped in 2025 papers (arXiv 2503.23303, 2510.01237) with a Hub model and dataset. PPO over embeddings; loosely related.',
|
| 404 |
quote: 'Also the second work published in September 2025 was exactly the same one jev proposed now.',
|
| 405 |
links: { tweet: 'https://x.com/Nandakishorm1/status/2100451360873611666', hf: 'https://huggingface.co/DeepMostInnovations/sales-conversion-model-reinf-learning', paper: 'https://arxiv.org/abs/2510.01237' },
|
| 406 |
+
m: { likes: 1, rts: 0, views: 84, hfLikes: 132 }, date: '2026-09-17', tags: 'ppo embeddings arxiv prior art sales' },
|
| 407 |
|
| 408 |
/* ---------------- arch ---------------- */
|
| 409 |
{ cat: 'explain', title: 'Jev\'s Architecture Unmasked', by: { name: 'Archer Hume', handle: '4rcherhume' },
|
| 410 |
desc: 'Argues Jev is a causal transformer with shared-state encoding, per-question branches and direct probability readouts, likely on a sparse MoE backbone, inferred from latency probing.',
|
| 411 |
quote: 'All the grifters are completely wrong about Jev\'s architecture so I decided I\'d release an open-weight version. BUT training takes time…',
|
| 412 |
links: { tweet: 'https://x.com/4rcherhume/status/2100555442061820286', blog: 'https://archerhume.com/posts/jevs-architecture-unmasked/' },
|
| 413 |
+
m: { likes: 2242, rts: 212, views: 376501 }, date: '2026-09-17', tags: 'moe teardown speculation open-weight promise blog' },
|
| 414 |
|
| 415 |
/* ---------------- diffusion ---------------- */
|
| 416 |
{ cat: 'diffusion', title: 'DiffusionGemma-as-Jev + vLLM structured-generation patch', by: { name: 'Matt Mastracci', handle: 'mmastrac' },
|
| 417 |
desc: 'Adds a Jev-like structured mode for DiffusionGemma to vLLM (PR #57250) and runs live evals: not slower than Jev\'s API and roughly tied on quality.',
|
| 418 |
quote: 'Is Jev faster than DiffusionGemma? No ❌ Is Jev smarter than DiffusionGemma? No ❌ (they\'re roughly tied!)',
|
| 419 |
links: { tweet: 'https://x.com/mmastrac/status/2100626193943052784', pr: 'https://github.com/vllm-project/vllm/pull/57250' },
|
| 420 |
+
m: { likes: 1003, rts: 75, views: 326616 }, date: '2026-09-17', tags: 'diffusion gemma vllm dgx spark evals' },
|
| 421 |
{ cat: 'diffusion', title: 'razorback16/openjev + api.codiv.ai host', by: { name: 'razorback16', handle: null },
|
| 422 |
desc: 'The same vLLM PR #57250 route, packaged as a Jev-wire /v1/systemone service: a DiffusionGemma 26B-A4B canvas with only the label slots left as noise, one read-only denoise step reads out the distribution. Public host at api.codiv.ai.',
|
| 423 |
links: { gh: 'https://github.com/razorback16/openjev', site: 'https://api.codiv.ai/v1/systemone' },
|
| 424 |
+
m: { stars: 134 }, date: '2026-09-17', tags: 'diffusiongemma 26b a4b nvfp4 vllm 57250 hosted api codiv systemone denoise canvas' },
|
| 425 |
{ cat: 'diffusion', title: 'open-jev (JoshuaSP): typed JSON with DiffusionGemma', by: { name: 'Joshua SP', handle: null },
|
| 426 |
desc: 'Typed JSON inference with DiffusionGemma, benchmarked against Every and Jev.',
|
| 427 |
links: { gh: 'https://github.com/JoshuaSP/open-jev' },
|
| 428 |
+
m: { stars: 16 }, date: '2026-09-16', tags: 'diffusion gemma benchmark every' },
|
| 429 |
{ cat: 'diffusion', title: 'LocalJev (GitHub Next): oMLX Jev API + model bake-off', by: { name: 'GitHub Next', handle: 'GitHubNext' },
|
| 430 |
desc: 'Official GitHub Next bridge, since not everyone there has Jev access: a Bun server putting oMLX DiffusionGemma behind /v1/systemone. Prompts for JSON probabilities instead of reading logits: wire-compatible, not equivalent. Ships a 1,200-request bake-off.',
|
| 431 |
quote: 'Not everyone on the team has access to Jev yet. Spent a morning cobbling together a poor man\'s Jev on top of omlx for local use.',
|
| 432 |
links: { tweet: 'https://x.com/GitHubNext/status/2101193436816920798', gh: 'https://github.com/githubnext/localjev' },
|
| 433 |
+
m: { likes: 1327, rts: 168, views: 185193, stars: 449 }, date: '2026-09-19', tags: 'github next localjev omlx bun typescript diffusiongemma 26b a4b json probabilities bakeoff m5 max benchmark qwen gemma moe wire compatible promptcoding' },
|
| 434 |
|
| 435 |
/* ---------------- explain ---------------- */
|
| 436 |
{ cat: 'explain', title: '45-second TL;DR on Jev (video)', by: { name: 'Matija Sosic', handle: 'MatijaSosic' },
|
| 437 |
desc: 'The most-shared explainer of the launch: a 45-second video that strips Jev down to its core idea, made because the official launch video was hard to follow.',
|
| 438 |
quote: 'I find the core idea beautifully simple, but the video made it really hard to understand.',
|
| 439 |
links: { tweet: 'https://x.com/MatijaSosic/status/2100190746389135772' },
|
| 440 |
+
m: { likes: 9277, rts: 668, views: 1584798 }, date: '2026-09-16', tags: 'video tldr explainer launch' },
|
| 441 |
{ cat: 'explain', title: 'typesafe-ai-benchmark: an "imposter Jev" on Cerebras', by: { name: 'Shannon Duncan', handle: 'iamMrDuncan' },
|
| 442 |
desc: 'An "imposter Jev" gateway on Cerebras with Qwen 3.8 27B, compared side by side. Similar quality and speed; TypeSafe far cheaper.',
|
| 443 |
quote: 'Built an alternative version of @typesafeai but on @cerebras with Qwen 3.8 27b. Similar quality, similar performance, but vastly different cost.',
|
| 444 |
links: { tweet: 'https://x.com/iamMrDuncan/status/2100467548298899918', gh: 'https://github.com/iammrduncan/typesafe-ai-benchmark' },
|
| 445 |
+
m: { likes: 209, rts: 25, views: 25998, stars: 32 }, date: '2026-09-17', tags: 'cerebras qwen gateway benchmark comparison cost' },
|
| 446 |
{ cat: 'explain', title: 'awesome-jev lists (AnotiaWang · hellogumbo)', by: { name: 'AnotiaWang · hellogumbo', handle: null },
|
| 447 |
desc: 'Two curated indexes; AnotiaWang\'s "Research & Open Models" section is the tight one, hellogumbo\'s catches the long tail.',
|
| 448 |
links: { gh: 'https://github.com/AnotiaWang/awesome-jev', gh2: 'https://github.com/hellogumbo/awesome-jev' },
|
| 449 |
+
m: { stars: 155 }, date: '2026-09-16', tags: 'awesome list index curated roundup' },
|
| 450 |
{ cat: 'explain', title: 'jev-benchmarks: probability-aware eval harness', by: { name: 'AbdelStark', handle: null },
|
| 451 |
desc: 'Calibration, selective risk and latency evals for typed decision models, Jev or open.',
|
| 452 |
links: { gh: 'https://github.com/AbdelStark/jev-benchmarks' },
|
| 453 |
+
m: { stars: 11 }, date: '2026-09-17', tags: 'calibration selective risk eval harness' },
|
| 454 |
{ cat: 'explain', title: 'Thread: System 1 / System 2 agents and the open recreations', by: { name: 'hhkkmon', handle: 'hhkkmon' },
|
| 455 |
desc: '14-part explainer framing agents as "expensive LLM calls pretending to be if statements", listing the open recreations at the end.',
|
| 456 |
quote: 'Big models think. Small decision models react.',
|
| 457 |
links: { tweet: 'https://x.com/hhkkmon/status/2100443314957038010' },
|
| 458 |
+
m: { likes: 2, rts: 0, views: 119 }, date: '2026-09-17', tags: 'thread explainer system 1 system 2 agents' },
|
| 459 |
{ cat: 'explain', title: 'Long-form: "OpenJev is NOT Jev open-sourced"', by: { name: 'mygtmhire', handle: 'mygtmhire' },
|
| 460 |
desc: 'Most thorough single post on TheoLeeCJ/openjev, including the warning that its probabilities are not calibrated confidence.',
|
| 461 |
quote: 'State in → probabilities out → code acts. No conversation required.',
|
| 462 |
links: { tweet: 'https://x.com/mygtmhire/status/2100666488625734103' },
|
| 463 |
+
m: { likes: 2, rts: 0, views: 123 }, date: '2026-09-17', tags: 'explainer openjev routing calibration' },
|
| 464 |
{ cat: 'explain', title: 'jev-behavior-study: probing the closed model', by: { name: 'RINNECODER', handle: null },
|
| 465 |
desc: 'Black-box behavioural study of the closed Jev 1.13.0 itself. Not a replica — it measures the thing everyone is trying to reproduce.',
|
| 466 |
links: { gh: 'https://github.com/RINNECODER/jev-behavior-study' },
|
|
|
|
| 511 |
brand-new entry nobody has reacted to yet cannot climb on its date alone.
|
| 512 |
The credit therefore rewards "new AND real", never novelty by itself. */
|
| 513 |
const W = { likes: 1, stars: 5, hfLikes: 8, views: 1/500 };
|
| 514 |
+
const RECENCY = { credit: 10000, halfLifeDays: 1, ref: 400 };
|
| 515 |
const BASE_DATE = '2026-09-19'; /* the snapshot the metrics were last refreshed against */
|
| 516 |
const ageDays = it => Math.max(0, (new Date(BASE_DATE) - new Date(it.date)) / 864e5);
|
| 517 |
const recency = it => {
|
|
Git LFS Details
|
|
Git LFS Details
|
|
@@ -1,4 +1,4 @@
|
|
| 1 |
-
const fs=require('fs');
|
| 2 |
const P='index.html'; let src=fs.readFileSync(P,'utf8');
|
| 3 |
const js=src.slice(src.indexOf('<script>')+8,src.indexOf('</script>'));
|
| 4 |
const head=js.slice(0,js.indexOf('/* ---- state ---- */'));
|
|
@@ -6,6 +6,9 @@ const ITEMS=eval(head+';ITEMS');
|
|
| 6 |
const sleep=ms=>new Promise(r=>setTimeout(r,ms));
|
| 7 |
const clean=t=>t.split('').filter(c=>{const k=c.charCodeAt(0);return k>=32||c==='\n'||c==='\t'||c==='\r';}).join('');
|
| 8 |
async function j(url,opt){const r=await fetch(url,opt);if(!r.ok)throw new Error(r.status+' '+url);return JSON.parse(clean(await r.text()));}
|
|
|
|
|
|
|
|
|
|
| 9 |
const esc=t=>t.replace(/\\/g,'\\\\').replace(/'/g,"\\'");
|
| 10 |
const changes=[];
|
| 11 |
(async()=>{
|
|
@@ -14,18 +17,18 @@ const changes=[];
|
|
| 14 |
if(L.tweet){const id=L.tweet.match(/status\/(\d+)/)[1];
|
| 15 |
try{const d=await j('https://api.fxtwitter.com/status/'+id);const t=d.tweet;m.likes=t.likes;m.rts=t.retweets;m.views=t.views;}catch(e){notes.push('tweet:'+e.message);}}
|
| 16 |
const ghs=[L.gh,L.gh2].filter(Boolean).map(u=>u.match(/github\.com\/([^\/]+\/[^\/#?]+)/)?.[1]).filter(Boolean);
|
| 17 |
-
if(ghs.length){let sum=0,ok=true;for(const r of ghs){
|
| 18 |
const hfs=[L.hf,L.hf2].filter(Boolean).filter(u=>!u.includes('/collections/')&&!u.includes('/spaces/')).map(u=>u.match(/huggingface\.co\/([^\/]+\/[^\/#?]+)/)?.[1]).filter(Boolean);
|
| 19 |
-
if(hfs.length){let sum=0,ok=true;for(const r of hfs){try{const d=await j('https://huggingface.co/api/models/'+r);sum+=d.likes||0;}catch(e){ok=false;notes.push('hf:'+
|
| 20 |
const before=JSON.stringify(it.m||{}),after=JSON.stringify(m);
|
| 21 |
-
if(before!==after||notes.length)changes.push([it.title.slice(0,44),before,after,notes.join(' ')]);
|
| 22 |
const ti=src.indexOf("title: '"+esc(it.title)+"'"); if(ti<0){console.log('TITLE NOT FOUND',it.title);continue;}
|
| 23 |
const mi=src.indexOf('m: {',ti); const me=src.indexOf('}',mi)+1;
|
| 24 |
const lit='m: { '+Object.entries(m).map(([k,v])=>k+': '+v).join(', ')+' }';
|
| 25 |
src=src.slice(0,mi)+lit+src.slice(me);
|
| 26 |
-
await sleep(
|
| 27 |
}
|
| 28 |
fs.writeFileSync(P,src);
|
| 29 |
-
for(const c of changes)console.log(c.join('
|
| 30 |
console.log('changed',changes.length,'of',ITEMS.length);
|
| 31 |
})();
|
|
|
|
| 1 |
+
const fs=require('fs'),{execFileSync}=require('child_process');
|
| 2 |
const P='index.html'; let src=fs.readFileSync(P,'utf8');
|
| 3 |
const js=src.slice(src.indexOf('<script>')+8,src.indexOf('</script>'));
|
| 4 |
const head=js.slice(0,js.indexOf('/* ---- state ---- */'));
|
|
|
|
| 6 |
const sleep=ms=>new Promise(r=>setTimeout(r,ms));
|
| 7 |
const clean=t=>t.split('').filter(c=>{const k=c.charCodeAt(0);return k>=32||c==='\n'||c==='\t'||c==='\r';}).join('');
|
| 8 |
async function j(url,opt){const r=await fetch(url,opt);if(!r.ok)throw new Error(r.status+' '+url);return JSON.parse(clean(await r.text()));}
|
| 9 |
+
/* GitHub stars go through the gh CLI so the whole run fits in one authenticated
|
| 10 |
+
hour (5,000 requests); the unauthenticated API allows only 60. */
|
| 11 |
+
const ghStars=repo=>{try{return parseInt(execFileSync('gh',['api','repos/'+repo,'--jq','.stargazers_count'],{encoding:'utf8',stdio:['ignore','pipe','ignore']}).trim(),10);}catch(e){return null;}};
|
| 12 |
const esc=t=>t.replace(/\\/g,'\\\\').replace(/'/g,"\\'");
|
| 13 |
const changes=[];
|
| 14 |
(async()=>{
|
|
|
|
| 17 |
if(L.tweet){const id=L.tweet.match(/status\/(\d+)/)[1];
|
| 18 |
try{const d=await j('https://api.fxtwitter.com/status/'+id);const t=d.tweet;m.likes=t.likes;m.rts=t.retweets;m.views=t.views;}catch(e){notes.push('tweet:'+e.message);}}
|
| 19 |
const ghs=[L.gh,L.gh2].filter(Boolean).map(u=>u.match(/github\.com\/([^\/]+\/[^\/#?]+)/)?.[1]).filter(Boolean);
|
| 20 |
+
if(ghs.length){let sum=0,ok=true;for(const r of ghs){const s=ghStars(r);if(s===null){ok=false;notes.push('gh:'+r);}else sum+=s;}if(ok)m.stars=sum;}
|
| 21 |
const hfs=[L.hf,L.hf2].filter(Boolean).filter(u=>!u.includes('/collections/')&&!u.includes('/spaces/')).map(u=>u.match(/huggingface\.co\/([^\/]+\/[^\/#?]+)/)?.[1]).filter(Boolean);
|
| 22 |
+
if(hfs.length){let sum=0,ok=true;for(const r of hfs){try{const d=await j('https://huggingface.co/api/models/'+r);sum+=d.likes||0;}catch(e){ok=false;notes.push('hf:'+r);}}if(ok)m.hfLikes=sum;}
|
| 23 |
const before=JSON.stringify(it.m||{}),after=JSON.stringify(m);
|
| 24 |
+
if(before!==after||notes.length)changes.push([it.title.slice(0,44).padEnd(44),before,'->',after,notes.join(' ')]);
|
| 25 |
const ti=src.indexOf("title: '"+esc(it.title)+"'"); if(ti<0){console.log('TITLE NOT FOUND',it.title);continue;}
|
| 26 |
const mi=src.indexOf('m: {',ti); const me=src.indexOf('}',mi)+1;
|
| 27 |
const lit='m: { '+Object.entries(m).map(([k,v])=>k+': '+v).join(', ')+' }';
|
| 28 |
src=src.slice(0,mi)+lit+src.slice(me);
|
| 29 |
+
await sleep(60);
|
| 30 |
}
|
| 31 |
fs.writeFileSync(P,src);
|
| 32 |
+
for(const c of changes)console.log(c.join(' '));
|
| 33 |
console.log('changed',changes.length,'of',ITEMS.length);
|
| 34 |
})();
|