Tool Attention Is All You Need
Tool Attention Is All You Need // Tool Attention Is All You Need // New research proposes a practical fix for the hidden 'MCP tax.' The work introduces a dynamic tool gating mechanism built on an Inte
Knowledge catalogue
Tool Attention Is All You Need // Tool Attention Is All You Need // New research proposes a practical fix for the hidden 'MCP tax.' The work introduces a dynamic tool gating mechanism built on an Inte
Build your own harness, folks. You won't regret it. These days, you just have to fix things yourself. It's doable, and it will set you up to easily deal with some of the madness that's happening in th
Cool paper on diversity collapse in AI agents. It's a common issue with all the deployed multi-agent systems. New paper shows that multi-agent LLM systems converge on near-identical outputs over time,
GPT-5.5 in Codex is a delight to work with: - Super sharp with responses - It understands intent better than any model - Great 'personality' - Gets lots of stuff done without pausing unnecessarily It
Things have been degrading super fast in Claude Code. I still use Claude Code, but my default is now Codex. I still prefer Opus models for coding, and so I will try again with the fixes. I appreciate
Worth a read if you are building production-grade AI agents. // Stateless Decision Memory for Enterprise AI Agents // Most of the interesting AI agent papers right now are about capability. This one i
How far are we from agents that can self-generate world knowledge? The work proposes an outcome-based reward that measures how much an agent's self-generated world knowledge actually improves its task
The post discusses the lack of AI models that work seamlessly across different agent frameworks and harnesses, suggesting this represents a significant opportunity for open-weight model development. T
Pay attention to this one, AI devs. This is particularly interesting if you work with long-horizon terminal agents that often drown in their own observations. TACO is a self-evolving framework that au
// Multi-Agent Synthesis RAG // Nice paper on improving RAG systems with multiple agents. (bookmark it) The paper introduces MASS-RAG, a multi-agent synthesis framework for retrieval-augmented generat
// Survey on Multi-Agent Systems // The paper traces the landscape from classical paradigms (consensus, distributed control, swarm intelligence, cooperative learning) to foundation-model-enabled MAS (
We are entering an extremely exciting era for open-weight models. Kimi K2.6 now feels like a top agentic model. I took it for a spin via @FireworksAI_HQ fast inference APIs. Kimi K2.6 has impressive a
NEW paper from NVIDIA. EDA tools like ABC have been hand-tuned by humans for decades. New research from NVIDIA shows they can evolve themselves. The work introduces the first self-evolving logic synth
Nice paper combining the strength of Skills and RAG. Most RAG systems retrieve on every query, whether the model needs help or not. This is wasteful when the model already knows the answer, and often
Great paper on self-improving agents. Why? We need to think more deeply about AI agent system design. The protocol specifies a framework for proposing, assessing, and committing improvements with audi
LLM Artifacts Connected to @karpathy's LLM Knowledge base idea, I've been building out a fun way to generate dynamic artifacts from these knowledge bases with the goal of discovering and revealing mea
The Top AI Papers of the Week (April 13 - 19) - AlphaEval - AiScientist - Auto-Diagnose - Nemotron 3 Super - Subliminal Learning - Automated W2S Researcher - Memory Transfer Learning Read on for more:
YT Podcast → LLM Artifact This is now my favorite way to consume podcasts. Knowledge artifacts generated by agents. The agent (Opus 4.7) spots important insights, does deep analysis, and generates tho
I prefer my design tool to be more closely integrated with where my agents work. I spent a few hours building my own design tool (inspired by Claude Design) inside my orchestrator. I can use this with
NEW paper from Apple. Interesting idea: 'Attention to Mamba'. The paper introduces a two-stage recipe for cross-architecture distillation from Transformers into Mamba. Naive distillation collapses tea
Nice paper from Google. And a great application of AI agents. Wearables capture a staggering amount of physiological signals every day. CoDaS is an AI co-data-scientist that turns raw wearable sensor
30K+ likes in the first hour. 🤯 That is crazy! Design is unsolved with agents. But lots of impactful work generated by agents is around design. Claude Design is Anthropic's way of saying that they are
cool new paper on self-improving agents // Self-Evolving Agent Protocol // One of the more interesting papers I read this week. (bookmark it if you are an AI dev) The paper introduces Autogenesis, a s
LLM agents loop, drift, and get stuck on hard reasoning tasks up to 30% of the time. Current fixes are either too blunt (hard step limits) or too expensive (LLM-as-judge adding 10-15% overhead per ste
// Skill Learning for Autonomous Web Agents // Web agents can navigate a page, but ask them to repeat a checkout flow they already completed, and they start from scratch every time. This work introduc
Agent evals are drifting away from production reality. Most benchmarks use clean tasks, well-specified requirements, deterministic metrics, and retrospective curation. Production work is messier, with
Coding agents learn from experience, but that knowledge stays locked in silos. Solve a thousand SWE tasks, and none of that wisdom helps with competitive coding. What if memories could transfer across
NEW Research from Google. Integration test failures are painful because the signal is buried in messy logs. Massive output, heterogeneous systems, low signal-to-noise ratio, and unclear root causes. T
What you need to know about Opus 4.7 * Takes instructions literally * Better vision means improved computer use and producing slides and other visual artifacts * Optimized for large-scale real-world a
Banger paper from NVIDIA. Agentic reasoning needs models that are not just capable, but efficient at long-context inference. The agent model layer is moving toward open, long-context, high-throughput
Long-horizon AI research agents are mostly a state-management problem. It is not enough for an agent to reason well in the next turn. ML research requires task setup, implementation, experiments, debu
Small models are cheap to run, but expensive to adapt. The hard part is not only fine-tuning. It is the surrounding loop that involves collecting data, diagnosing failures, building evals, avoiding re
// Artifacts as Memory Beyond the Agent Boundary // An agent doesn't always need a bigger memory buffer. Sometimes the environment itself remembers on the agent's behalf. New research formalizes this
It looks like everyone is finally catching up with the fact that agent sessions in CLI mode can only get you so far. It makes sense that the new Codex app, Cursor, and Claude Code (desktop) feel and l
LLM Knowledge Base → Slides When @karpathy shared his LLM Knowledge Base setup, many were wondering how to generate more visual forms of the wiki. There are many options, but I think @GammaApp is one
Most AI assistants wait for you to ask. But a truly useful agent should notice you need help before you say anything. New research takes a serious shot at building proactive agents that work in real t
I like this view from the brilliant Terence Tao. I have always believed that AI could help to augment us. We built and train agents like that today. But all forms of intelligence can and should be pos
DAIR.AI (Democratizing AI Research) is an organization focused on AI education and research democratization, frequently sharing updates on their X (formerly Twitter) account about prompt engineering,
The Top AI Papers of the Week (April 6 - 12) - Memento - Neural Computers - The Universal Verifier - Agent Skills in the Wild - Memory Intelligence Agent (MIA) - Single-Agent vs Multi-Agent LLMs - Sca
One great outcome of PaperWiki is personalized surveys. Survey papers continue to be one of the best ways to track a field. My agents are now generating personalized surveys on topics using my paper L
What if a model became the computer itself? NEW paper from Meta. (bookmark this one) What if the model wasn't just using the computer, but became the computer? New research from Meta AI and KAUST make
Another banger paper from Microsoft. Why it's a big deal: It teaches reasoning models to compress their own chain-of-thought mid-generation. The most interesting finding isn't the 2-3x memory savings
DAIR.AI's Claude Code Cheat Sheet, published at academy.dair.ai, is a consolidated quick-reference resource covering Claude Code commands, keyboard shortcuts, and best practices in one place. It ac...
I just built my own wiki generator plugin for my agents. My agents can now generate wikis for anything I ask. One of my favorite wikis is called PaperWiki. This is a great example of what @karpathy de
The DAIR.AI post references a Claude Code Cheat Sheet consolidating key commands, shortcuts, and best practices for Anthropic's terminal-based agentic coding tool. Claude Code is Anthropic's AI-po...
NEW paper from Microsoft Every agent benchmark has the same hidden problem: how do you know the agent actually succeeded? Microsoft researchers introduce the Universal Verifier, which discusses lesson
// Scaling Coding Agents via Atomic Skills // Most coding agents train end-to-end on full tasks like resolving GitHub issues. But complex software engineering is really a composition of simpler skills
Claude Code Cheat Sheet Here is a list of top commands, shortcuts, and patterns you need to know when using Claude Code. We will update this as new updates for Claude Code come out. The idea is to cur
Great paper on improving memory for AI agents. NEW paper: Memory Intelligence Agent (MIA) MIA boosts GPT-5.4 by up to 9% on LiveVQA. Quick summary: Most memory-augmented agents treat memory as a stati
NEW: Meta announces Muse Spark. All you need to know: * It's their new multi-modal reasoning model. * Strong at multi-agent orchestration and multi-modal reasoning. * Contemplating mode orchestrates m
NEW paper from Google on multi-agent research agents. It's one of the first systems that handles end-to-end LaTeX generation, targeted literature reviews, and conceptual diagrams as a decoupled, stand