tae retweeted
Google just open-sourced Harness².
(turn any harness into a self-improving system)
a harness is everything wrapped around the model, like the system prompt, tools, memory, skills, and subagents. today a person designs it once, and the agent runs every new task with that same fixed setup.
Harness² lets the agent refine that setup for each new task on its own, with the model weights left untouched.
on every task, the agent proposes edits to its harness, runs a few candidate versions, and composes a better harness by comparing how those runs behaved.
because it learns from the differences between runs, it needs no correct answer or grader for the task.
that covers a single task. Harness² also carries what it learns from one task into the next.
after each task, it saves the harness it settled on along with the lessons that held up across the runs. the next time it refines a harness, it starts from those notes instead of from scratch.
this means every finished task makes the next round of edits more informed.
Google tested this with Gemini 3.5 Flash and GLM-5 on legal, office, and professional work benchmarks. one round of edits was enough to beat their original harnesses, and giving the agent more rounds pushed scores higher.
those extra rounds can go wide or deep. the agent either tries several candidate harnesses at once or keeps refining a single one step by step, using OpenCode or Codex to carry out the work.
you can find the Harness² repo here: github.com/google-research/h…
I wrote about a similar idea for agents already running in production, where the harness should repair itself after a failure instead of waiting for someone to debug the trace by hand.
that piece walks through how @Cometml's Opik (open-source) closes this loop, going from a failing trace to a proposed fix to a regression test that stops the same failure from coming back. it's quoted below.
tae retweeted
We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:
Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:
Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:
Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!
In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
tae retweeted
Bend 2 is here!
It is a new programming language that blocks AI mistakes via *proof checking* - the same technique big AI labs used to solve open math problems, like Navier-Stokes.
It is also very fast, and runs on GPUs.
Watch the video. Link in the comments.
RELEASE DAY
After almost 10 years of hard work, tireless research, and a dive deep into the kernels of computer science, I finally realized a dream: running a high-level language on GPUs. And I'm giving it to the world!
Bend compiles modern programming features, including:
- Lambdas with full closure support
- Unrestricted recursion and loops
- Fast object allocations of all kinds
- Folds, ADTs, continuations and much more
To HVM2, a new runtime capable of spreading that workload across 1000's of cores, in a thread-safe, low-overhead fashion. As a result, we finally have a true high-level language that runs natively on GPUs!
Here's a quick demo:
tae retweeted
なんか、YAMAHAから公式にやばいのが出とる。
【無料配布】音源ファイルを読み込むだけで指定した難易度や編曲スタイルに最適化されたピアノ譜を自動作成する、 耳コピ・採譜・ピアノアレンジにも最適な楽譜生成ソフトウェア YAMAHA「Piano Sheet Converter β」リリース & 無料配布中！！
freekontaktina.com/2026/09/y… #DTM #DTMer
tae retweeted
Looped transformers are a popular architecture topic right now.
This new technical report extends the loop across tokens.
Recurrent Looped Transformer (RLT) makes the decoder recurrent over every token, prompt and response included.
A causal encoder builds global KV memory. For each new token, the decoder combines the token's encoder representation with its own final hidden state from the previous token and a sliding-window cache of recent activations.
With a 48-layer decoder, the computation path after t tokens runs through 48t decoder blocks, while each token still executes a fixed number of blocks. Depth grows with the sequence and per-token cost stays the same.
The same state transition is used for pretraining, SFT, sampling and RL replay, and nothing resets at the prompt-response boundary. RL replay rebuilds states under the current weights instead of reusing stale rollout states.
The report is a design proposal. The author states that reasoning gains, hardware speedups and RL scaling are goals that have not been measured yet.
Paper: github.com/yifanzhang-pro/re…
Chat with Paper: academy.dair.ai/papers/recur…
tae retweeted
🎉视频 AI 剪辑省 Token 组合： Antigravity + ChatCut ! ！
gemini 3.8 flash 的速度 和 token 消耗
得天独厚，为所欲为！！
比 Codex 操作快多了！太爽了！
🔥方法如下：
将提示词发送到 Antigravity 对话框即可
---
连接到此电脑上的本地 ChatCut Desktop MCP 服务器。
使用以下确切命令或可执行文件注册名为 `chatcut_desktop` 的 stdio MCP 服务器：
"/Users/{你的用户名}/Library/Application Support/ChatCut/chatcut-mcp"
注册后，请验证 `chatcut_desktop` 已启用且可被发现。然后启动一个新的 Agent 对话，以便使用该连接。
---
灵感来源： @johnAGI168
tae retweeted
since anything can be an RL env, @adithya_s_k create a GeoGuessr env using OpenEnv trained a 4B VLM to play GeoGuessr using TRL
read about it: huggingface.co/spaces/Huggin…
tae retweeted
if you want the practical starting point, the repo has the program, runnable sft example, and the path from traces to environment rl:
github.com/burtenshaw/traini…
We finished the Training Agents series. Six live sessions over six months, from evaluating agents to training them inside real environments. All of it is on the Hugging Face YouTube channel and all of the code is open.
Here's what we did and who made it happen:
1. Agentic Evaluations WorkshopWhere agent evals actually stand, and why benchmark scores don't match what people see in use. With Avijit Ghosh and Nathan Habib (Hugging Face), Arvind Narayanan (Princeton), Pierre Andrews (Meta), J.J. Allaire (UK AI Security Institute) and Mahesh Sathiamoorthy (Bespoke Labs).
2. RL for Agents Workshop Environments, rollouts, reward design and the inference bottlenecks that appear when you move from RL for LLMs to RL for agents. With Lewis Tunstall (Hugging Face), Will Brown (Prime Intellect), Ofir Press (Princeton) and Alex Zhang (MIT CSAIL).
3. Training Agents 1: SFT on agent traces Public coding-agent traces turned into prompt/completion data, a TRL + LoRA fine-tune on Hugging Face Jobs, metrics in Trackio, and an honest look at what the first eval numbers can and cannot tell you. Joined by Sergio Paniego and Quentin Gallouédec.
4. Training Agents 2: Distillation Off-policy, on-policy and self-distillation for moving capability from a teacher into a smaller coding agent.
5. Training Agents 3: Reinforcement learning GRPO after SFT: group sampling, verifiable reward functions, reading the reward/KL/length curves, and three experiments, one of them with a deliberately gameable reward so we could watch the hacking happen.
6. Training Agents 4: From reward functions to environments The reward stops being a function and becomes a place the agent acts in. We walked the reset()/step() contract from Gym to LLM agents, built an OpenEnv environment and pushed it to the Hub, plugged it into TRL's GRPOTrainer, then trained a real coding agent (OpenCode) through Harbor with AsyncGRPOTrainer on Hugging Face sandboxes.
The series has passed 300k views. Thank you to every speaker, to the TRL team, and to everyone who showed up live with questions.
Playlist: youtube.com/playlist?list=PL…
tae retweeted
Self-improving agents are a top research topic right now.
This new survey is a good map of the area.
(bookmark it)
It splits recursive self-improvement into stages of autonomy.
An agent first executes improvements someone else designed. Then it chooses its own improvement strategy, collects its own experience, adapts to new environments, and finally improves the process of improvement itself.
That staging makes claims easier to check. When a paper says its agent is self-improving, you can ask which of these stages it actually automates.
The survey also uses a Headroom-Closed Index to show where current LLMs fall short, and compares requirements across scientific discovery, embodied agents and software engineering.
Paper: academy.dair.ai/papers/the-l…
tae retweeted
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
tae retweeted
Agent Behavior - AI 에이전트의 반복 행동을 문서화해 평가 기준으로 삼는 표준
최종 결과 대신 정보 확인, 판단, 실행, 실패 시 복구까지 작업 중 반복돼야 할 행동을 정의
사람의 트레이스 리뷰, 루브릭, 자동 eval이 하나의 같은 스펙으로 실제 행동을 평가할 수 있음
news.hada.io/topic?id=32962
tae retweeted
im born for this game
協力マルチ対応「空き巣」ゲーム『Dumpster Gang』発表。根こそぎ盗み、家を荒らし、駆除業者にも飛びかかる、人間なめすぎアライグマ体験
automaton-media.com/articles…
tae retweeted
I’m looking to connect with new grads and early-career ML talent with hands-on post-training experience who want to break into frontier AI labs in SF. ($200K–$400K+)
Open roles include:
• Member of Technical Staff, Research
• Research Product Manager
This is a YC-backed startup and data research AI lab building simulation environments and synthetic data pipelines for reinforcement learning, with a focus on advancing post-training for AI models.
Think:
• RLHF / DPO / distillation
• LLM evaluation + benchmarking
• RL for post-training
• Simulation environments
• Synthetic data generation and pipelines
• Research on LLM behavior
• Research product management for post-training and evaluation systems
High-value signals:
• Contributions to publicly available benchmarks
• Research experience at places like Mercor, Scale, Handshake, Surge, etc.
• University lab work in post-training RL or narrow-domain LLM evaluation
• ICML / NeurIPS research where you can clearly explain your specific contribution
• Strong engineering background from Waterloo, MIT, Berkeley, Stanford, etc.
• Product experience translating research needs into tools, datasets, evaluations, or workflows
Bonus: experience at Chinese AI labs like Alibaba, Z.ai, Moonshot, ByteDance, or Tencent.
tae retweeted
📄 读英文论文不用再左右切翻译了，这个开源项目能整篇翻完，还保住原来的排版。
GitHub 揽获 3.6 万 Star，背后有一篇 EMNLP 2025 的系统演示论文撑着。
以前读一篇 arXiv 上的论文，复制一段丢进翻译框，回来发现公式全散了、图表位置也对不上，一篇读下来来回切几十次，人都麻了。它的做法是直接产出双语 PDF，译文和原文并排，公式、图表、目录、批注全部原样保留，滚动着往下读就行，也可以只要纯译文那一版。
命令行、图形界面和 Docker 三种用法都有，翻译服务可以自己换，默认走 Google 翻译，还带翻译缓存和批量翻译，一次扔一摞论文进去也扛得住。
它还提供了 MCP 模式，可以直接接进 AI 工具里当一个翻译能力来用。
看文献的时间省下来，才是真的在做研究。
GitHub：github.com/PDFMathTranslate/…
tae retweeted
개쩐다 이런거 진짜 오픈소스의 힘이다
Tailcat - Tailscale 계정 없이 두 머신을 연결하는 netcat
짧은 토큰만 전달하면 TCP 포트를 WireGuard 터널로 연결하며 계정 등록과 제어 서버가 필요 없음
DERP에서 시작해 NAT 통과에 성공하면 P2P UDP로 전환하고, 실패하면 릴레이를 쓰며 SSH와 포트 전달도 가능
news.hada.io/topic?id=32920
