2026
- 2026-11-03Let Agents Check Private Repositories Without Reading Them Into Public Replies
GitHub’s experimental enclaves constrain what crosses between agents. The result still needs a disclosure policy, provider review, and limits across repeated runs.
- 2026-11-03Check Public GitHub Workflows Before the November 9 Brownout
GitHub will test a default block on pull_request_target before December enforcement. Review privileged workflows instead of broadly allowing the trigger.
- 2026-10-31The Shared AI-Training Signal Is Coming. Keep Your Current Crawler Rules.
A proposed robots.txt token could simplify training preferences. Provider rollout is still ahead, and the signal does not control access.
- 2026-10-30Your Coding Agent Can Ignore a Preferred Tool. Enforce the Rule at the API.
Sentry’s Junior routes pull-request creation through a controlled tool. Its creator’s account shows why a shell changes enforcement.
- 2026-10-30Mistral Large 4 Weights Are Here. Check the License Before Hosting.
The release brings lower API rates and self-hosting, but model-service businesses face a separate license requirement.
- 2026-10-29Gemini 4 Carbon: Watch the Access Gate and the Regular Price
Google’s new frontier model starts with selected cyber defenders. Its announced $2/$10 introductory rates rise to $4/$20 per million input/output tokens afterward.
- 2026-10-29Copilot Fusion Picks the Models. Check the Whole Job’s Bill.
GitHub’s new routing system can use several models on one task. Billing follows the models used, while a temporary discount can obscure the longer-term comparison.
- 2026-10-28Opera’s Local AI Preview Is Worth a Small Browser Experiment
Opera One can expose local models to websites and extensions. Start with one optional feature, measure the user’s device, and keep a fallback.
- 2026-10-28Copilot Agents Can Use an App Identity. The Organization Pays.
GitHub’s new authentication preview removes the signed-in user dependency. Its read-only permission can still spend AI credits, and app-specific budgets are unavailable.
- 2026-10-27Keep Agent Pull Requests Blocked Until Every Required Check Passes
Kiro’s engineering account highlights a quiet failure mode: no red checks can mean nothing has run yet. Give agents a clear readiness signal.
- 2026-10-27GPT-6.1 Terra Cuts Token Rates for Routine Agent Work
OpenAI’s new model halves standard token rates versus GPT-5.6 Terra and GPT-6.1 Sol. Long prompts, cache writes and retries still matter.
- 2026-10-26Apple’s 20B Local Model: Test the Small Jobs First
A new single-Mac study shows a small memory footprint with important limits. Start with extraction or tagging and measure your own workload.
- 2026-10-23Update Claude Code: 2.1.309 Fixes Sandbox and Permission Gaps
Anthropic’s CLI release fixes shell-startup, symlink, MCP-configuration and teammate permission handling.
- 2026-10-23Claude Generated Checked Proofs for Software. Read the Guarantee Carefully.
Anthropic and Theorem report no-crash proofs for selected library entry points. The precise theorem matters more than the headline.
- 2026-10-23AI21’s 98% Code-Review Saving Is a Test to Reproduce
A small public-PR experiment cut token costs with a cheaper model and smaller prompts. Production savings remain unproven.
- 2026-10-22Railway’s Credential Warning Needs Action Beyond a New API Key
Railway’s new AWS and Google Cloud response guides make the next step clear: revoke exposed access, replace credentials, and inspect what they could reach.
- 2026-10-22METR Found Gaps in Agent Monitoring. Check During the Job, Not Just at the End.
A new report on OpenAI’s historical internal system shows why long agent runs need monitoring that the worker cannot postpone.
- 2026-10-22Flint Brings Coding Agents to Your Mac. Check the Memory Before the Speed.
Recursive’s open coding model is worth a local trial. Its optimized Mac engine needs 48 GB of memory, and vendor throughput is only the start of the test.
- 2026-10-03Apple Is Tightening Full Disk Access Because of AI Agents. Stop Asking for the Whole Mac.
Apple says autonomous agents make blanket disk access more dangerous. Give coding tools the project folder they need—not Mail, Messages, browser history, and everything else.
- 2026-10-02GitHub Copilot Can Click Desktop Apps. Use It Only When There Is No Better Tool.
Copilot can now click and type in desktop software. Use it for GUI-only work, with narrow permissions and a human before the final high-impact action.