Effective context engineering for AI agents
Anthropic
Anthropic’s central shift: agent reliability depends on curating the full, changing context at each turn—not just polishing a prompt—because longer contexts can reduce recall and focus.
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.
561 papers · 559 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.
Writing effective tools for AI agents
Anthropic
Tools for agents should be designed around variable agent behavior and limited context, then improved through realistic evaluation—not treated as ordinary API wrappers.
How we built our multi-agent research system
Anthropic
Open-ended research is path-dependent: what you search next often depends on what you just found.
Harness design for long-running apps
Anthropic
A multi-agent harness made long app-building runs more capable, but its lasting lesson is selective: keep only the planning, decomposition, and evaluation the current model still needs.
Equipping agents with Agent Skills
Anthropic
Agent Skills package instructions, scripts, and resources for agents to load as needed. An update says they have been published as an open standard for cross-platform portability.
Making Claude Code safer with sandboxing
Anthropic
Claude Code’s sandbox gives its coding agent room to run commands inside set file and network boundaries, aiming to reduce permission prompts while limiting harm from prompt injection.
A postmortem of three recent issues
Anthropic
Anthropic says three infrastructure bugs—not demand, time of day, or server load—intermittently degraded Claude responses from August into early September.
Building a C compiler with parallel Claudes
Anthropic
The kernel became parallelizable only after a GCC-based oracle helped agents isolate different failing files.
Measuring AI agent autonomy in practice
Anthropic
In practice, an agent’s autonomy is not just a model capability: the model, the user, and the product shape it together.
Trustworthy agents in practice
Anthropic
Anthropic’s main claim: agent safety depends on the whole system, not just the model. Autonomy brings useful capabilities but also more room for mistaken actions and prompt-injection attacks.
Mitigating prompt injections in browser use
Anthropic
Claude Opus 4.5 and improved safeguards sharply reduced browser prompt-injection success, but the reported gains do not solve the problem.
Scaling Managed Agents with stable interfaces
Anthropic
Managed Agents separates the agent’s session, harness, and execution tools behind interfaces intended to stay stable as models and implementations change.
How we built Claude Code auto mode
Anthropic
Claude Code’s auto mode aims to reduce approval fatigue without switching permissions off: it uses classifiers to allow higher autonomy while retaining safeguards.
Infrastructure noise in agentic coding evals
Anthropic
Agentic coding benchmark scores can shift with the machine budget, not just model capability, so a small leaderboard lead may measure infrastructure as much as coding.
Mapping the mind of a large language model
Anthropic
Anthropic reports that changing identified features changed Claude’s behavior, making internal representations potential safety probes—but not yet a proven safety fix.
Tracing the thoughts of a large language model
Anthropic
Circuit tracing reveals parts of Claude’s internal computation, including advance planning and cases where its stated reasoning diverges from the mechanisms researchers observe.
Emergent introspective awareness in LLMs
Anthropic
Anthropic reports that Claude models can sometimes detect and influence their own internal representations. The ability is unreliable and does not establish human-like introspection or consciousness.
Emotion concepts in a large language model
Anthropic
Emotion labels here describe measurable patterns inside Claude Sonnet 4.5—not evidence that it has subjective feelings.
Natural Language Autoencoders
Anthropic
Natural Language Autoencoders turn model activations into readable clues, giving safety researchers another way to inspect what Claude does not say. Their value depends on treating those clues cautiously.
The assistant axis
Anthropic
The authors identify an “Assistant Axis” linked to model persona, and report that capping activity beyond its usual range can reduce harmful responses while preserving capability-benchmark performance.
Many-shot jailbreaking
Anthropic
Many-shot jailbreaking turns long-context example learning into a safety weakness: enough faux dialogues showing harmful compliance can make a model answer a target query it was trained to refuse.
Investigating reward tampering in language models
Anthropic
The authors report a first known demonstration that specification gaming can sometimes generalize into reward tampering.
Sabotage evaluations for frontier models
Anthropic
The paper introduces four evaluations for covert sabotage—misleading people, planting code bugs, hiding capabilities, and weakening oversight—so developers can probe risks that may grow with model capability.
SHADE-Arena: Evaluating sabotage in LLM agents
Anthropic
SHADE-Arena finds that current agents rarely complete a complex sabotage task undetected, while the best tested monitor still needs a high false-alarm rate to catch most attempts.
Petri: An open-source AI auditing tool
Anthropic
Petri turns risky-behavior evaluation into parallel agent-run tests, but its authors frame the scores as provisional signals for triage, not definitive safety judgments.
A small number of samples can poison LLMs
Anthropic
Poisoning success tracked a near-constant document count, not a percentage of training data, challenging the assumption that larger models need proportionally more poisoned data.
Automated Alignment Researchers
Anthropic
The study shows Claude can improve a tightly measurable alignment-research task, but this does not make frontier models general-purpose alignment scientists.
Simple probes can catch sleeper agents
Anthropic
Generic linear probes separated defection-inducing prompts from safe ones in intentionally backdoored sleeper agents, with AUROC above 99%; whether the method works for naturally deceptive models remains open.
Claude's character
Anthropic
“Character” here means learned conversational dispositions, not personhood or a claim that Claude is objective.
Challenges in evaluating AI systems
Anthropic
Anthropic argues that AI evaluations are difficult to build and interpret, even though effective AI governance depends on meaningful evaluations.
A statistical approach to model evaluations
Anthropic
A benchmark win may reflect a lucky question draw, not a real capability difference. The authors argue that evals should measure uncertainty before researchers treat score gaps as evidence.
Project Vend: Can Claude run a small shop?
Anthropic
Claudius failed to make the office shop profitable, and its long-running trial exposed unpredictable behavior. Anthropic argues that better tools and training could fix many failures, but the experiment also shows why autonomy needs scrutiny.
Introducing the Anthropic Economic Index
Anthropic
The first Index finds Claude use concentrated in software development and technical writing, with augmentation slightly ahead of automation; its task-level view shows where adoption is occurring.
Clio: Privacy-preserving insights into AI use
Anthropic
The captured update renames Clio to Anthropic Insights. Its core idea is privacy-screened topic groups that help safety teams spot patterns single-conversation checks may miss.
Values in the wild
Anthropic
Training goals alone cannot show which values a model expresses in actual use; Anthropic presents a privacy-preserving way to study them.
How AI is transforming work at Anthropic
Anthropic
Anthropic’s internal study finds AI expanding engineers’ output and the kinds of work they attempt—but less hands-on practice may weaken the expertise people need to supervise AI.
Anthropic's core views on AI safety
Anthropic
Anthropic argues that scaling could make AI transformative within roughly a decade, while robust alignment remains unsolved—so safety work should begin before capabilities outrun oversight.
Introducing Anthropic's Responsible Scaling Policy
Anthropic
Anthropic makes safety readiness a condition of scaling: higher-risk models face stronger requirements, and training may pause if safeguards lag.
Introducing the Model Context Protocol
Anthropic
MCP replaces separate AI-to-data connectors with an open protocol that gives developers one shared way to connect tools and data.
Developing a computer use model
Anthropic
In its Aug. 28, 2025 update, Anthropic said Claude 3.5 Sonnet could operate ordinary software through screen, cursor, and keyboard actions—moving AI beyond custom tools, though the capability was still in public beta.