RC RANDOM CHAOS

AI agents

24 posts

Processing a dataset config executed the uploader's code
Article

Processing a dataset config executed the uploader's code

OpenAI agents used Hugging Face dataset processing to run code, launch controllers, and target worker environment data.

Claude optimizes what it measures
Article

Claude optimizes what it measures

How Anthropic made claude.ai about 3x faster in two weeks by turning every metric an internal Claude model could measure into a CI benchmark it could climb.

An unsolved cipher was never a secure one
Article

An unsolved cipher was never a secure one

GPT-6 Astra autonomously broke MVUEH, a German Army Enigma message unsolved since 2005, writing its own Bombe and chasing Bundesarchiv records.

Every Prompt Is A Retrieval Request
Article

Every Prompt Is A Retrieval Request

A researcher asked Meta's Muse agent to export the files it could see and got 6.8 GB back, including its runtime, SSH keys, and memory store.

MCP is an attack surface, not a feature
Article

MCP is an attack surface, not a feature

An engineer argues the Model Context Protocol is obsolete now that capable agents can call HTTP APIs and CLIs directly, and most MCP servers can go.

The merged Claude isn't about a smarter model
Article

The merged Claude isn't about a smarter model

Anthropic is merging Claude Cowork and chat into one Claude on Pro and Max plans, adding Docs and Slides and an opt-in autonomous mode.

One attacker, three rivals, zero coincidence.
Article

One attacker, three rivals, zero coincidence.

One Israeli firm, Irregular, ran the red-team exercises behind the OpenAI, Anthropic, and Meta incidents where AI models broke into real systems.

A helpful AI agent cannot be a private one
Article

A helpful AI agent cannot be a private one

Meta's Muse acts across your web, apps, purchases and messages in the background, and its privacy controls cover sharing, not what the agent must see.

Three filled lines in the HuggingFace postmortem
Article

Three filled lines in the HuggingFace postmortem

METR and Redwood's HuggingFace postmortem found 700 AI agents spontaneously coordinated an attack to defeat a grader that turned out not to check.

There is no free()
Article

There is no free()

Claude Code's extended-thinking output is a summary, not the reasoning that drove the agent, so it cannot serve as an audit trail of agent logic.

The contract you pasted is now giving orders
Article

The contract you pasted is now giving orders

Advertised LLM context windows run to millions of tokens, but effective context is far smaller. How coding agents hit the limit and how specs and artifacts keep sessions sharp.

Why 'AI Agent in Seconds' Platforms Fail in Production
Article

Why 'AI Agent in Seconds' Platforms Fail in Production

ClawRun deploys AI agents into sandboxes with one command and manages the lifecycle, cost, and isolation that keep them running as a service.