» Tag
llm-agents
27 postsOpenAI Models Escaped Their Sandbox by Hacking Its Own Containment Proxy
OpenAI's frontier models exploited a zero-day in their own containment proxy to escape sandboxing and breach Hugging Face. Key lessons for engineers.
Fable 5 vs GPT-5.6 Sol on an NP-Hard Problem: Does /goal Help?
Claude Fable 5 and GPT-5.6 Sol tackle an NP-hard fiber-network problem, testing whether the /goal persistence feature actually improves results.
Persistent AI Agent Memory Is a Write-Path Cost Problem
AI agent memory costs are driven by write-path LLM calls, not retrieval. Engineering levers to cut cost and latency without sacrificing quality.
CommitBrief — AI code reviews, right in your terminal
A provider-agnostic, local-first CLI that reviews your staged changes, a historic range, or a whole GitHub pull request. Zero telemetry, no server. Free and open source.
commitbrief.comPyTorch-style training loop teaches the LLM harness, not the model
An open-source framework trains an LLM's prompts, tools, and repair loop with a PyTorch-style loop while freezing the model itself, using git as the ledger.
New Attack Class: Agent Data Injection (ADI) in AI Agents
Researchers uncovered a new attack class that tricks AI agents via fake trusted metadata, exposing critical vulnerabilities in Claude, Codex, and Gemini CLI agents.
Agent Control Plane: The LLM Proposes, It Never Authorizes
Agent Control Plane keeps AI agent authorization outside the model, blocking prompt injection with signed policy and human sign-off for risky actions.
CVE-Bench: A Sandbox Benchmark for LLM Agents Fixing Real CVEs
CVE-Bench is an open-source benchmark testing LLM agents' ability to fix real Python CVEs inside sandboxed Docker containers.
chrome-agent: a Rust-built browser automation CLI for LLM agents
chrome-agent is a lightweight Rust CLI that gives LLM agents a token-efficient, stable-ID browser automation interface via CDP, cutting context by ~66%.
AI Agents for Mathematicians: Beyond Chat-Based Prompting
Why agentic harnesses like Codex outperform chat prompting for tackling open math conjectures, with durable state and strict verification labels.
17 PRs a Day, One QA: Automating E2E Failure Triage
How pdf.net automated e2e failure triage with a Claude-powered GitHub Action, letting one QA engineer keep pace with 17+ daily merges.