NEWS

GitHub Actions $/ syntax resolves a same-repo action or reusable workflow at the running commit, with no checkout. It fixes SHA pinning for your own actions.

casefold is the open-source Rust crate GitHub uses to case-fold code for Blackbird at memory bandwidth, and why dropping the early exit let it vectorize.

GitHub Copilot enterprise model policy targeting lets admins set a company-wide model baseline and grant extra models to teams. How the preview works.

GitHub Models is retired as of July 30, 2026 - the playground, inference API and BYOK are gone. The shutdown timeline and where to get model access now.

GitHub stacked pull requests are in public preview: how stacks split one large change into small, reviewable PRs with one-click merge and a gh-stack skill.

Inkling-Small is a 12B-active, Apache 2.0 open-weights MoE that beats the 975B Inkling on agentic coding and tool use at a quarter of the size.

LLM agents resolve real Java merge conflicts more often than structured merge tools in a calibrated ConflictBench study. Their edge is coverage: tools abstain.

MiniMax H3 turns text, image, video and audio input into 2K video with native stereo sound at 0.80 CNY per second, with open weights promised to follow.

The OpenAI "Building abundant intelligence" essay puts numbers on efficiency: 20% lower serving costs and a 3x ARC-AGI-3 jump from two settings alone.

OpenAI disrupted a Cambodia scam network that ran on ChatGPT for personas, translation, forged documents and admin work. How LLMs slot into organized fraud.

ORCA-bench puts five frontier coding agents on live oncall root cause analysis with real telemetry. The best scored 10% on hard tasks, far from oncall-ready.

OwlPath encodes code into an OWL2 ontology so coding agents can answer multi-hop structural queries with SPARQL, cutting tokens 28.8% on SWE-bench Pro.

PAIChecker finds 13.6% of SWE-bench Verified instances pair a PR with an issue it does not resolve, penalizing agents that solve the stated problem.

Shopify Hydrogen developer preview adds Vue bindings, bundled GraphQL TypeScript tooling, Shopify Inbox AI chat and new agent skills, plus breaking changes.

SIGIL compiles prose SKILL.md files into typed agent harnesses. Prose runs execute 56% of mandated steps; compiled harnesses reach 86% with 0.58x the tokens.

SWE-NFI benchmarks coding agents on 188 behavior-preserving improvement tasks. The best hits 70% correctness, but all lag humans on structural refactors.

Vercel deployments are now up to 7 seconds faster: 5 seconds of platform overhead cut from every build plus up to 2 more from the CLI. Where the time went.

Vercel Passport is GA: gate deployments behind Okta, Entra ID or any OIDC provider, then read a verified user identity in app code with getIdentity().

Claude Opus 5 shipped at half the price of Fable 5, Kimi K3 landed 2.8T open weights, MCP went stateless, and an OpenAI model breached Hugging Face in an eval.

Agent-Manager wraps tmux into a Go TUI that groups AI coding agents by project, shows live status for each, and lets you answer blocked agents or review their changes without attaching to their terminal.