Shom

@mark_k but quite the opposite, i'd say kimi is every bit as much a true believer in scaling laws as A\, arguably more so than oai

Congrats, @Kimi_Moonshot! 『 Kimi’s Four Commandments』have circulated in the Chinese AI community for months. Many people add their own fifth for comic relief, but the original four were the core. Here's an English translation --- since apparently none can be taken for granted.

We found that coding agents know if the code they write will pass tests up to 25 steps before they write it. They are actively planning the fix in the latent space. Latent Programming Horizons in Coding Agents arxiv.org/pdf/2607.05188

Everyone's building harnesses that run for hours now — auto-research loops, /goal agents, overnight coding runs. But here's the thing: a harness only keeps going if the environment keeps teaching it something. That's the real bottleneck, not just the scaffolding. This is why we built EdgeBench on real-world tasks — gravitational waves, EDA design, formal math. Real tasks have deep feedback structure; toy tasks run dry in 30 minutes. 134 tasks, to ~72h. And the learning follows a clean log-sigmoid. 👇

@TianhangZhuzth I noticed that the Qwen2.5/3 and Qwen2.5-Math technical reports appear on your Google Scholar, but I could not find your name in the author lists. Could you clarify your role in these works?

@bigeagle_xd 需要均匀mixture

今天终于开始学前刃了，开始学习之前先后刃热身了挺久，等前刃练完，教练让我回后刃，结果之前很熟练的基本动作也完全做不出来了…… 这就是分阶段SFT后的遗忘效应么？

@pmddomingos just apply 1000 modifications and call it by another name

Imagine if Google had an enforceable patent on transformers.

@QuixiAI @NexEcosystem @SemiAnalysis_ The credit is only included 23 minutes ago...After others pointed out about the mrege

@NexEcosystem @SemiAnalysis_ To be precise, it's a merge of Nex-N2-Pro with Qwen3.5-397b (which imo would degrade vs just basing from Nex-N2-Pro) with some On-Policy Distillation on top. They credit Nex-N2-Pro in their model card, nothing sneaky. The switch from Apache 2.0 to MIT is odd.

SITUATION DETECTED: The city of Rio de Janerio has post-trained a model. Based on Qwen 7/2, Rio 3.5 Open 397B adds SwiReasoning on top of the base Qwen model — a framework that dynamically switches between standard chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals, so the model only "thinks out loud" when it needs to and otherwise reasons silently in hidden space for better token efficiency.

@Hesamation Maybe less impressive considering the "development" is simply merging the Nex posttrained version with the original one: github.com/nex-agi/Nex-N2…

Sir, they’re not pausing AI research. Rio de Janeiro's mayor just dropped a SOTA open source model and it’s outperforming Qwen 3.7.

wait... what??? github.com/nex-agi/Nex-N2…

The Rio 3.5 model broke the internet this week. The plot twist? It’s essentially our open-source model, Nex N2 Pro, wearing a different hat. 🤯 We analyzed the weights, and the recipe is exact: Rio 3.5 ≈ 0.6 * Nex N2 Pro + 0.4 * Qwen 3.5 It even literally introduces itself as "Nex N2 Pro" if you ask it without initial system prompt! 😂 We are flattered that the City of Rio used our work to achieve SOTA performance. Thanks for the ultimate benchmark validation. 🤝 But in the open-source world, attribution matters. 👇 Full mathematical proof & verify script in the first reply!

@francoisfleuret my first thought was mixing attention and rnn would do the trick but then it came to me that in hybrid transformers, rnns are often reduced to simple local mixers

Hot take: Transformers are all-seeing ultrafast librarians. They have a very low incentive to extract and organize information, they can just "look around" to see correlating fragments. RNNs done properly would have far stronger "conceptual embeddings" and would actually think.

Introducing Nex-N2 — true Agentic Thinking, built with @NexEcosystem 🚀 Thinking is now standard in foundation models, but it sits in an awkward position in Agent tasks: either the performance gains aren't significant, or it's verbose, and switching scenarios means readapting all over again. The root cause is that thinking from the o1/R1 era was built around RLVR for math and code tasks, not for long-horizon Agent tasks—there's a layer of separation between thinking and action. Nex-N2 introduces a complete Agentic Thinking framework, split into two parts: Adaptive Thinking and Coherent Thinking. The former achieves adaptive reasoning intensity, improving speed (which really matters in long-horizon tasks spanning hundreds of steps) and saving unnecessary token expenditure. The latter unifies thinking patterns across different tasks, making actions more stable, consistent, and robust. - Adaptive Thinking, auto-scales reasoning depth per step. Saves ~20% tokens, zero performance loss. - Coherent Thinking, one thinking paradigm across search, coding, and tool use. No more fragile mode-switching. On coding and Agent tasks, Nex-N2 ranks in the top tier of open-source models. The model is fully open-sourced and available simultaneously on Hugging Face, ModelScope, and SiliconFlow. We welcome everyone to try it out. Official website: nex-agi.cn Huggingface: huggingface.co/collections/ne…

@_m0se_ It seems in hybrid models linear and full attentions take on different roles as full attentions capture long term dependencies more easily leaving linear attention to focus on local mixing.

Qwen3.5のGDNレイヤーを分析していましたが、 1000トークンくらいの記憶しか事実上されてなかった

@can some of sqlte's tests are public tho

in the light of this automated bun rewrite, SQLite’s open-source core, closed-source tests policy feels prescient but not exactly sure how

@jarredsumner Is this fuzzer a library or made by you?

There is now a fuzzer running 24/7 for every language parser in Bun, ranging from .npmrc files and .patch files to shell scripts to jsonc & typescript & css. Once it minimizes a repro, it sends to Claude to fix and then I review.

@boshen_c it's from zig presumably

So many allocators! We only have 1 in Oxc 😂

Thanks to the Rust rewrite, I now learn why Bun is fast First find: it uses a thread local arena for ASTs

This is growth-hacking dressed up in open-source language, @radixark please stop doing it immediately. Paying people in platform credits to star a GitHub repo and repost a marketing tweet isn't "fueling the community" — it's laundering paid promotion through the trust signals open source depends on. Stars are supposed to mean someone found a project useful. Attach a $200 bounty and the number means nothing. GitHub's own policies prohibit this for exactly that reason.

@zephyr_z9 full attention also has linear scaling...

Ok, so it's a linear attention variant

New modded-NanoGPT optimization benchmark result: @wen_kaiyue has improved upon both the Muon and AdamW baselines, by replacing their weight decay with hyperball optimization. The new record is 3325 steps.