Automation
Developer automation, CI/CD, platform and workflow tooling.
-
GitHub Made Its Actions Run Counts Less Precise on 25 September 2026 — and Called That More Accurate
GitHub announced on 25 September 2026 that queries for workflow runs in the GitHub Actions API and UI now return a less precise but more accurate count of records when searched by workflow, event, status,…
-
Eighth Place at 59, and the Cheapest of the Three: What the Record Around Gemini 3.8 Flash Actually Contains
Google shipped Gemini 3.8 Flash and a Cyber variant. The official record is a model card, an API listing and a web rollout. The unofficial record is larger.
-
GitHub Is Retiring Selected Copilot Models on October 19, 2026 — And Code Completions Are on the List
GitHub published a changelog entry announcing the upcoming deprecation of selected GitHub Copilot models in mid-October.
-
Inside playagit’s GitHub/CodeQL Research Pipeline: Evidence Gaps and Draft Outcomes
playagit’s GitHub/CodeQL research covered 36 runs, 446 claims, and 24 reviewed drafts. Examine the recorded evidence gaps, fixes, and publication outcomes.
-
Agentic Delivery Gates and the Publication Count Our Records Do Not Explain
Agentic frameworks document gates and reviews. Playagit’s 166 pipeline runs show recorded decisions and a publication breakdown that does not reconcile.
-
A Passing Eval Is Not a Safe Agent: What the Sandbox Decides, the 11-Point Behaviour Shift One Remote Swap Produced, and the Production Telemetry That Names Every MCP Server
An agent evaluation produces a number, and the number is usually read as a property of the model.
-
Four Projects Pay AI Agents in USDC, and the Only One Where an Agent Pays a Human Was Built for a Hackathon
Four projects surfaced in close succession, all of them paying software agents rather than people, and all of them settling in USDC — but on four different settlement layers.
-
Cloudflare Managed Defense: AI Alert Analysis with Separate Evidence Collection
Cloudflare Managed Defense uses Workers-based AI agents and global telemetry, separating evidence collection from inference to support security analysts.
-
GPT-6 Astra in Copilot: What Our 152 Codex Runs Can Explain
GPT-6 Astra is available in Copilot. Our 152 Codex CLI runs document timing, changes, and 59 errors, with three missing from the listed breakdown.
-
Counting Claude’s Failures: 52 Errors Across 594 Runs, and the Judge That Read Them Wrong
The basic description of the product — that Claude is a large language model and AI chatbot developed by Anthropic — comes from an encyclopedia entry and a survey listing of large language models, and…