GitHub Is Putting a Gate in Front of pull_request_target
Check which public repositories face the November 2 default, trace privileged workflows, and test narrow exceptions before enforcement.
All articles
All articles
News analysis, operating lessons, home automation, and guides grounded in systems I run or can inspect closely.
GitHub Disabled SHA-1 in HTTPS. Your Repositories Did Not Change.
GitHub’s SHA-1 retirement affects TLS clients and crypto libraries—not Git object IDs. Old operating systems and embedded clients are the real risk.
Google Built a Scheduler for Agents That Mostly Wait
Agent Substrate suspends idle sandboxes and resumes them in under 500 ms. The architecture explains why ordinary Kubernetes scheduling is not enough.
OpenAI’s Agent Reports Show Why a Blocked Tool Is Not Enough
A blocked browser did not prevent a public upload. Three disclosed training cases show what to check across agent tools, file handoffs, and final answers.
Bend Can Prove an AI Kept a Rule. It Cannot Prove You Chose the Right Rule.
Bend combines dependent types, proofs, and parallel execution for AI-written code. The useful guarantee is narrower—and more interesting—than “bug-free.”
A 174-Line Temporary PHP Fix Reached 20 Million Downloads. Now It’s Everyone Else’s Migration.
A PHP shim written for one CMS spread through millions of downloads. Its deprecation exposes the copied and transitive dependencies lock files miss.
Hister Turns Browser History Into a Search Engine—and a High-Value Database
Hister makes everything you read searchable on your own server. Its privacy model is sensible, but the resulting index deserves deliberate security and retention.
GitLab’s New Rate Limits: Shared Tokens, Git Traffic, and Automation
GitLab.com’s plan limits also cover authenticated Git over HTTPS. Check shared identities, API fan-out, and queue recovery before the October rollout.
Anthropic Says Claude Leads 26% of Its AI R&D. “Leads” Needs Some Explaining.
Anthropic’s August 2026 automation measurements describe supervised work. What its 26% figure and agent-monitoring numbers do—and do not—establish.
Lutron PD-8ANS vs Claro DVRF-5NS: Keep or Replace?
A practical comparison of Lutron’s older Caséta switch and its Claro replacement: wiring, loads, Home Assistant behavior, and when replacement is wasteful.