Actually Shipped: building real software with AI, judged by what actually ships
Latest articles
Article / 9 min read
Claude Code skills: how they load, what a big catalog costs, and which ones to keep
How skills load in Claude Code and Codex, what a big skill catalog costs on every turn, and a trigger test you can repeat on your own skill in ten minutes.
Article / 5 min read
My AI agents over-checked a nine-page site. Now I set the check budget
In a review loop, agents answer findings with more checks and nobody prices them. Why the owner, not the agent, should set the verification budget, what I keep, and the rule I gave them.
Article / 9 min read
Claude Code hooks: recipes we ran, and the cases where they do not block
Eight Claude Code hook recipes with settings and scripts, each run in a real headless session on version 2.1.314, plus the setups that look safe and are not.
Article / 7 min read
CLAUDE.md and AGENTS.md: an annotated example and what each tool loads
An annotated AGENTS.md and CLAUDE.md pair you can copy, plus a load test of which instruction files Claude Code 2.1.314 reads, with the Codex, Copilot and Cursor rules from their docs.
Latest news
News / github.blog
Copilot code review picks its own effort level and flags leaked secrets, in preview
A GitHub update lets Copilot choose Lite, Balanced or Max review depth per pull request, scans for exposed secrets before merge, and takes Azure DevOps review to general availability.
News / learn.chatgpt.com
GPT-6.2 Astra and Sol reach Codex and Copilot, with a 4.5x faster Ultrafast mode
OpenAI's two new coding models are live in ChatGPT Work, Codex, the API and GitHub Copilot as of November 17. What each is for, who gets the Ultrafast speed option, and what the sources leave open.
News / developers.openai.com
ChatGPT Pro and Business now include up to $500 a month in OpenAI API credits
Eligible Pro and Business subscribers can link an API organization they own and receive monthly API credits that expire with each billing cycle. The amounts per plan, what the credits cover and the Codex exception.
News / github.com
Claude Code 2.1.314 and 2.1.316: fast mode for Opus 4.8 ends November 30
Two Claude Code releases from October 30 and 31 deprecate fast mode on Opus 4.8, add a setting that blocks file-tool writes outside the working directories and fix the exit code of a failed claude -p start. The changes that matter if you build with it.
News / mistral.ai
Mistral releases the weights of Mistral Large 4, its 1-trillion-parameter model
Mistral Large 4, nicknamed Le Chonk, is now open weight under a modified MIT license that asks only model-as-a-service businesses for a separate deal. Where you can run it, what the API costs and which claims are Mistral's own.
News / anthropic.com
Anthropic hands the Agent Skills standard to the Agentic AI Foundation
SKILL.md now sits under the Linux Foundation's Agentic AI Foundation, next to MCP and AGENTS.md. What changes for people who write skills, and what does not.
News / cursor.com
Cursor's Agent SDK is generally available
Cursor now lets you define an agent as a folder, test it with evals and deploy it to Cursor or your own infrastructure, triggered from GitHub, Slack, webhooks or a schedule.
News / github.blog
GitHub Agentic Workflows is generally available on Free, Team and Enterprise
Repository automations written in Markdown now compile to GitHub Actions and run a coding agent with read-only permissions by default. Who can use it and what it costs to run.