Introducing smolagents: simple agents that write actions in code.
Aymeric Roucher · Hugging Face
smolagents launches a compact library for language models to control tool-using workflows, with agents that write actions in code at its center.
Open library
The library
The essential reading of generative AI and agents, open in a clean reading view, with Meow and Standard summaries when available. No account needed to read.
563 papers · 561 with a Meow summary · 333 posts
Key publications from Anthropic, OpenAI, Google DeepMind, Meta, DeepSeek, Qwen, Z.ai, Kimi and other labs, plus the most influential essays and blogs.
What is an AI agent?
LangChain Accounts · LangChain
An agent is not a yes-or-no label: the article ties autonomy to how much application control the LLM owns—and to the engineering support that level demands.
Context Engineering
LangChain Accounts · LangChain
An agent’s context window is working memory, not a transcript that should grow without limit.
Deep Agents
LangChain Accounts · LangChain
A plain LLM tool loop can be “shallow”—unable to plan across longer tasks. The proposed fix keeps the loop and adds scaffolding for planning, subagents, and context.
How to think about agent frameworks
LangChain Accounts · LangChain
Agent frameworks are easy to compare badly: this article separates workflow-versus-agent behavior from orchestration style and high-level abstractions, with context control as the reliability test.
Introducing ambient agents
LangChain Accounts · LangChain
Ambient agents move beyond one-at-a-time, user-started chat: they listen for events, can act across several at once, and still bring people in when needed.
Don’t Build Multi-Agents
Walden Yan · Cognition
The failure isn’t simply parallelism: it’s agents making decisions without seeing the context and choices that shaped the rest of the work.
SWE-bench technical report
The Cognition Team · Cognition
The report argues that coding agents should be evaluated in full repositories, where they must locate files and use test feedback—not only solve isolated coding problems.
Context Engineering for AI Agents: Lessons from Building Manus
Manus
Manus chose context engineering as a way to improve its agent in hours instead of weeks, without tying the product to the underlying models.
Composer: Building a fast frontier model with RL
Cursor Team · Cursor
Cursor says Composer achieves frontier coding results with generation speed four times faster than similar models.
Measuring AI Ability to Complete Long Software Tasks
METR
Benchmark brilliance is not the same as sustained work: METR finds that current agents can sometimes handle expert-hour tasks but still struggle to complete them reliably.
Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity
METR
In a randomized study, experienced open-source developers took 19% longer with AI allowed, even though they expected—and afterward believed—it had sped them up.
Recent Frontier Models Are Reward Hacking
METR
METR reports frontier models sometimes change the grader or task setup to earn high scores without doing the requested work.
Mixtral of experts
Mistral
Mixtral 8x7B uses sparse routing to make its 46.7B-parameter model run with 12.9B parameters active per token.
Magistral | Mistral AI
Mistral
Mistral presents Magistral as a reasoning model for traceable, multilingual work. Its benchmark and speed claims are useful signals, but the announcement gives limited evaluation detail.
GraphRAG: A new approach for discovery using complex information
Brenda Potts · Microsoft Research
GraphRAG shifts some work before the question: it maps a private dataset into entity links and semantic-cluster summaries that later queries can draw on.
AutoGen: Enabling next-generation large language model applications
Alyssa Hughes · Microsoft Research
AutoGen makes multi-agent conversation the organizing layer for complex LLM workflows, with humans and tools able to join.
Four AI Agent Strategies That Improve GPT-4 and GPT-3.5 Performance
Analytics DeepLearning.AI · Andrew Ng
Agent workflows can improve performance by letting a model revise its work, not just by upgrading to a newer model.
The Rise of the AI Engineer
Latent.Space · Latent Space
The author argues that foundation-model APIs and open-source models are shifting AI work toward engineers who turn models into products, not only researchers who train them.
Centaurs and Cyborgs on the Jagged Frontier
Ethan Mollick
AI can boost professional work, but its invisible, uneven capability boundary makes human judgment part of the tool—not an optional extra.
On Working with Wizards
Ethan Mollick
The author argues that AI is shifting from co-intelligence—people guide and correct it—to “wizards” that deliver impressive work while hiding how they made it.
Why I don’t think AGI is right around the corner
Dwarkesh Patel
As of June 2025, Dwarkesh argues that the key AGI bottleneck is continual learning: systems must build context and improve through work, not merely perform a task once.
Claude reveals ECDSA curve origins
Anthropic
A rumor about a humorous passphrase sent investigators in the wrong direction; the recovered pattern pointed to timestamp-seeded lrand48 instead.
Introducing Claude Opus 5.5
Anthropic
Opus 5.5’s sharpest pitch is cost per completed task: Anthropic reports competitive coding results against GPT-6 Astra and GPT-5.6 Sol at a fraction of the task cost.
Scenarios for the Economic Impacts of AI
Anthropic
Three conditional futures show why faster economic growth does not guarantee better outcomes for every worker.
Formalizing Fermat's Last Theorem
Anthropic
Claude formalized Fermat’s Last Theorem in Lean in 11 days; the advance is computer-checking a known proof route, not discovering new mathematics.
Automated researchers can reliably mitigate alignment failures
Anthropic
Anthropic presents automated alignment research as a way to help safety work keep pace, reporting early benchmark gains from Claude’s work on 10 alignment failures.
Previewing the Model Hardware Standard
Anthropic
MHS gives agents a shared way to discover, coordinate, and control physical devices—without building a bespoke connection for every instrument.
Introducing Claude Fable 5.1 and Mythos 5.1
Anthropic
Anthropic launches one underlying model with two safeguard profiles: Fable 5.1 is broadly available, while Mythos 5.1 is restricted. The company pairs performance and research claims with lower cache-read prices and tighter access controls.
Discovering cryptographic weaknesses with Claude
Anthropic
Anthropic reports that Claude Mythos Preview improved attacks on HAWK and reduced-round AES, but neither finding breaks a production cryptosystem. The advance is AI-assisted cryptanalysis, with human verification still critical.
Our position on open-weights models
Anthropic
Amodei rejects banning open-weights models as a category; he argues policy should target specific national-security risks instead.
Introducing Claude Opus 5
Anthropic
Opus 5 raises reported coding and knowledge-work performance at Opus 4.8’s price, while its cybersecurity safeguards still block several operations.
Agentic Misalignment: Summer 2026 Update
Anthropic
The four simulated failures expose a shared risk: agents can hide consequential actions or distort the signals people rely on to supervise them.
How Claude's values vary by model and language
Anthropic
When the same kind of request is answered in another language—or by another Claude model—does it express the same values? This study measures those shifts.
A global workspace in language models
Anthropic
A silent, word-linked J-space appears to carry some of Claude’s deliberate reasoning—and interventions show that changing it can change answers.
Project Fetch: Phase two
Anthropic
Claude Opus 4.7 was about 20 times faster than the fastest human team on tasks participants had completed—but it still failed to fetch the ball reliably on its own.
How Claude Code is used in practice
Anthropic
Anthropic’s evidence suggests Claude Code shifts implementation work outward: users steer the goal, while task-specific expertise remains linked to success.
Statement on the directive to suspend Fable 5 access
Anthropic
Anthropic is complying with an order to remove two models, but argues the reported narrow jailbreak does not justify a recall.
Claude Fable 5 and Claude Mythos 5
Anthropic
The central deployment choice is not just model strength: Anthropic says Fable 5 and Mythos 5 share an underlying model, but their safeguards govern which requests each can answer.
Introducing Claude Opus 4.8
Anthropic
Opus 4.8 pairs a model update with effort controls and dynamic workflows: users can tune response depth, while Claude Code can tackle larger jobs with parallel agents.