Prism, a scaffold for automating science-of-evals research developed by Louis Thomson during MATS 9.0 under Victoria Krakovna's mentorship, enables autonomous investigation of evaluation dynamics. In …
A guide explains how to safely run coding agents like Claude Code in YOLO mode locally by executing them inside a Docker container, protecting the host machine from malicious actions while allowing th…
A small software company's AI coding spend tripled as engineers adopted Claude Code and Codex, mirroring cost overruns at Uber and Meta. The engineering lead identified three blind spots—broken prompt…
Silica, an open-source agent for notes and codebases, builds a live model of a folder's contents and rolls back any edit that breaks coherence, according to its developer. The local-first tool, compat…
A developer tracked Claude Code and Codex pass rates for 95 days and calculated the per-session cost of Anthropic's agentic coding tool. A typical mid-size session of 40 model calls with a 90K token w…
A developer shipped a second pack of 10 Claude Code slash commands, 4 subagents, 3 hook recipes, and 2 templates, built with a discipline of no fix without a reproduction, no refactor without a test h…
Clawk, a new open-source tool, gives coding agents a disposable Linux VM instead of direct host access, using hypervisor isolation to let agents install packages, run servers, and access the network w…
A developer detailed how they configured Claude Code as a testing toolkit, creating custom hooks and a CLAUDE.md file to enforce security and coding conventions. They emphasized building personalized …
Claire explains that a harness is code written around an AI agent to make it more effective at a specific job, and she walks through the custom Claude Agent SDK harness she built to automate Sentry bu…
Tadashi Shigeoka published a checklist on July 13, 2026, identifying 15 production-readiness checks that non-developers should apply before shipping AI-written code, citing common failures such as dou…
Shell history and AI coding assistant transcripts are a major source of leaked credentials, according to developer Panayotis Vrypanis, who created shg (the shell guard) v0.2.6 to scan shell/REPL histo…
Jerrycan, an AI-native Rust backend framework and generation platform, has been released in version 0.2.0, allowing AI agents to design, generate, verify, and package complete backends without human c…
Clawk, a new open-source tool, gives coding agents like Claude Code and Codex a disposable Linux VM instead of direct access to the user's laptop, blocking unauthorized network connections and file ac…
On July 8, 2026, Wiz Security published GhostApproval, a vulnerability in six major AI coding assistants — Claude Code, Cursor, Amazon Q Developer, Windsurf, Augment, and Google Antigravity — that use…
Researchers at the University of Missouri-Kansas City disclosed the Ghostcommit attack, which hides malicious prompt-injection instructions inside a PNG image referenced by an AGENTS.md file to turn A…
Large language models have advanced to the point where they can solve open mathematical problems, generate accurate QR codes, and handle real-time customer service calls, according to a compilation of…
A developer argues that the biggest problem with AI coding assistants is not the AI itself but how code is written. By adding metadata like file paths, usage locations, and feature READMEs, developers…
A developer describes using git worktrees to run parallel Claude Code agents without file conflicts. Each agent gets its own worktree and branch, isolating file changes and preventing mid-run overwrit…
A newsletter post explains how to use loops in Claude Code, describing them as recursive goals where an orchestrator agent iterates until a purpose is met, potentially spinning off subagents for paral…
Anthropic has extended the deadline for Claude Fable 5 access to July 19 at 11:59:59 PM PT, also extending Claude Code's availability. The move affects users of Anthropic's AI products.…