Agentic Coding Digest

What actually shipped.

A single daily edition of announcements from the AI coding tools you use, read from a curated source list, summarised, and linked back to the original. No infinite feed, no account, no advertising.

Leadmodel#

Copilot's model picker gained a model built for long-horizon autonomous work

GPT-6 Astra reached general availability in GitHub Copilot for Pro+, Max, Business and Enterprise, selectable from the model picker across Visual Studio Code, Visual Studio, Copilot CLI, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse. GitHub says internal testing showed it plans and validates as it goes, batches diagnosis with verification, and independently confirms its results before declaring a task done. Rollout is gradual and billed usage-based at provider list pricing, with Enterprise and Business administrators controlling access through model policies where new models are enabled by default unless disabled.

GitHub Changelog · Read the originalXHacker NewsLinkedIn
02tooling#

A running response accepts corrections over a WebSocket mid-turn

Mid-turn steering sends additional instructions while a response is still in progress, so the model can incorporate corrections or changing requirements without the turn being restarted. Async tool calling lets the model keep working while your application runs function or custom tools, with results returned as they become available. Reasoning effort can also be changed mid-conversation, raised for difficult work or reduced for routine follow-ups, while preserving the cached prompt prefix.

OpenAI API Changelog · Read the originalXHacker NewsLinkedIn
03research#

The side project you thought was yours may be the model's suggestion

Armin Ronacher describes hacking a Chinese CarPlay dongle with CatPlay, a Rust implementation of the CarPlay protocol that an LLM proposed rather than he found. An acquaintance independently arrived at a nearly identical CarPlay project after separate model conversations, which Ronacher reads as coordination that nobody deliberately arranged. His question is how much of what we build comes from eliciting the same latent capabilities out of the same models, rather than from training bias, social media or genuinely independent thought.

Armin Ronacher · Read the originalXHacker NewsLinkedIn