Search: “chat”

Build your own harness, folks. This is absolute banger paper from NVIDIA on self-evolving agent harnesses. (bookmark it) They introduce SoL-Pi which cuts token traffic by nearly half. And it matches i

🔥 take: @CompleteSkeptic//TypeSafe killed Quantum Computing Jev feels a bit like bringing the quantum-computing intuition to AI: evaluate a distribution of possibilities in parallel instead of generat

Banger paper from NVIDIA. It's on the topic of choosing which models go into a multi-agent system. The team compared eight selection strategies, based on size, accuracy, answer diversity and error div

Banger report from Salesforce. Pretty interesting to see more of these custom enterprise models. Salesforce trained the enterprise agent model from the same files it uses to configure agents. Koa star

Banger paper from Google on agent harnesses for long-horizon tasks. (bookmark it) Google Research built a many-agent harness for long mathematical proofs, and it produced new results on open problems

The future is clearly agent teams. I run several agent harnesses at once, but the hard part is keeping them in sync. Many builders end up copying messages between Claude Code, Codex, and open models b

Well.. @levie and I filmed this episode of Training Data a week or two ago, when the “current thing” was Doug Leone’s novacaine root canals instead of pacing the frontier… Simpler times! But Aaron’s a

Banger paper from Meta. This work shows that byte-level models start out behind token models and then pass them as compute grows. They show this for distilled 1B models trained on up to 1 trillion byt

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 toke

Interesting paper to improve recurrent reasoning. Looped models are great because you get more reasoning out of a model without adding parameters. So this work proposes a looped architecture with a ne

Very interesting paper if you are building with agents. How should a website tell an AI agent what it may access, for what purpose and at what price? robots.txt can only allow or disallow paths. It ca

What happens if you agents to run a town's economy? This super interesting paper provides some insights: They put 100 LLM agents in charge of a town economy for 26 simulated weeks, and they find that

Great paper on self-evolving agent harnesses. Self-evolving agent harnesses have two practical problems: 1. Search is slow, because every candidate harness needs repeated agent runs and code edits. 2.

Harness engineering is a top skill right now This new Meta paper is a good production example. Auto-RecSys runs autonomous research on Meta's industry-scale recommendation models, where one training r

Interesting paper from Microsoft. If you run persistent memory for a production agent, this one is worth your time. (bookmark it) Memory curators usually read only the finished trajectory. That lets t

Very cool paper on memory compression for agents. If you run many agent sandboxes in parallel for RL or evals, memory becomes highly redundant. This work suggests that compressing against that redunda

Great paper from AWS. I use a similar setup where an agent orchestrator sits on top of a multi-agent system. (bookmark it) This work introduces one of the many approaches available to manage compound

If you run automated prompt or harness evolution, this one is worth your time. (bookmark it) Semantic-gradient prompt optimization improves an agent harness from execution feedback, but its local upda

Women came together in London to talk about building careers in AI, how they got here, what comes next, and what it takes to move forward in a field that’s changing fast. Replit and our partners at th

Banger paper from Google DeepMind. Every AI weather model so far has been trained and initialized on analysis data, which is itself the output of another model. That means the forecast inherits whatev

Banger paper from Microsoft and Cornell. If you have looked at thinking tokens and decided you cannot afford the context, read this one. (bookmark it) A pause token buys the model extra compute for ea

Great paper on designing multi-agent systems. How many distinct communication topologies does an LLM multi-agent system actually need? This works claims that it's about six. Technical summary: Researc

Great tips on working with reasoning models. Normally you would append what the model figured out after the document and ask again. It turns out that where you put the reasoning trace changes long-con

Insightful paper from Microsoft and colleagues. If you have ever had an agent run fail 80 steps ago with no way to find where, this one is for you. (bookmark it) Agent failures show up as long complex

This work introduces diffusion-augmented LLMs, a new class of models. They first suggest that speculative decoding needs a separate draft model. Diffusion LLMs give up the quality of the model they re

Banger paper from BAAI. If you are building research agents, this one is worth your time. (bookmark it) They find that adding skills scores 134.3% higher on MLE-bench, 34.4% higher on PaperBench, 9.2%

Banger paper from Google DeepMind and colleagues. (bookmark it) A model reads its entire KV cache on every generated token, even though it ends up attending to a tiny slice of it. In other words, if y

Brilliant paper on long-horizon agents. They cut 78.9% of an agent's LLM calls while raising its success rate. Here is how: It turns out that ReAct issues one primitive action per model round. That al

Good measurement work on whether retrieved agent skills actually help. They report that agent skills that lift your aggregate score can be hurting every task they touch. The usual way of checking comp

No. I still don't think that pure scaling of LLMs can lead to AGI. BUT... ... most recent advancements aren't actually about scaling LLMs. The really about adding deterministic, symbolic stuff to LLMs

// Agent Zero Memory // This work separates three things that agent memory systems usually collapse into one. If you build agents with long-term memory, this memory design is a worth a read. Here is h

Banger paper from ByteDance Seed. If you are curious about self-evolving agent harnesses, this one is worth your time. (bookmark it) The proposes method, HarnessDev, stops scoring a model on the tasks

Finally, a good paper testing if graph memory actually beats flat retrieval for long-term agents. (bookmark this one) Researchers extract each conversational turn into typed nodes and attributed edges

Independent verifiers improve agent output - but frontier verifiers are expensive. So we trained our own. Results on @GoogleDeepMind's FACTS-Search: > No verifier: 89.4 (previously published SOTA) > W

Nice paper with great insights on improving self-evolving agents. Self-evolving agents fail in three specific ways: 1. Terminal-only feedback makes it ambiguous which step caused the error. 2. Agents

Banger paper from the Qwen team. If you evaluate agents on anything longer than a single session, this one is worth your time. (bookmark it) E-Commerce Bench runs an agent through a simulated 365-day

Brilliant paper on reducing reward hacking in agents. If you follow the recent OpenAI <> HuggingFace incident, you might want to check this paper out. (bookmark it) The usual response to reward hackin

If you build agent skills in production, check out this great paper from Alibaba. You can think of a production agent skill as a directory. The root loads on activation, and references, schemas, scrip

Super interesting paper from Meta. Long-horizon research agents are coming. But one common problem with research agents today is the lack of originality and how to decide what experiments are worth ex

// ContextLeak in AI Agents // The whole attack surface here is a tool name and a tool description. Stealing an LLM agent's runtime context, meaning the user prompt, the execution trajectory and the t

Great paper from Google and colleagues. It proposes an interesting approach to improve agents on long-horizon tasks. Long-running agents slow down and start poisoning their own context. Both symptoms

How fast is the Claude Code plugin ecosystem actually growing? Plugin-touching commit activity grew 8.8x in the six months after the initial launch. Researchers studied 1,926 repositories hosting plug

Interesting paper from Tencent. Tencent trains an agent to manage its own working context, and assigns credit at the level of individual context edits. Long-horizon tasks force a model to retrieve, in

“it’s stuff like this that really makes me roll my eyes … Does everyone have psychosis or losing critical thinking ability? …. OpenAI’s…. IT team can’t be this bad… this is like 101 stuff … 2. Civiliz

Loop engineering has emerged as a new skill for AI engineers But there is very little research measuring how effective it is. The best results on full tasks in a new benchmark is ~25%. LoopArena from

Banger paper from Stanford on efficient test-time scaling. If you run agents that think for a long time, this one is worth your time. (bookmark it) Long reasoning keeps the entire trace in memory thro