GitHub is putting Claude Opus 5 directly into Copilot, shifting the new model from a standalone launch into a daily developer surface. The rollout covers Pro+, Max, Business, and Enterprise users across VS Code, Copilot CLI, cloud agent, JetBrains, and more.
The new Claude default for high-end daily work shifts the model race toward performance per dollar. Anthropic says Opus 5 approaches Claude Fable 5 on coding and knowledge work while keeping API pricing at $5/M input and $25/M output tokens.
The post pushed back on the idea that more loops can remove human judgment from production coding workflows.
Open coding agents are maturing around harnesses, protocols, and SDK compatibility. OpenInterpreter says its Kimi K3 native harness is written in Rust, Apache licensed, and compatible with ACP and the Codex SDK.
HN’s interest landed on the tradeoff Bionic represents: local models, cloud fallback, coding workflows, and a closed-source desktop app all in one package.
HN treated Grok Build less as a feature drop and more as a test of control: an open Rust coding agent is useful, but telemetry, forks, and provider lock-in shaped the discussion.
Coding agents need real execution rights to be useful, but handing them a local machine is uncomfortable. Clawk drew HN interest by moving that work into a disposable, network-restricted VM.
OpenAI says SWE-Bench Pro no longer reliably measures frontier coding capability after finding 30% of its public tasks broken. The cited issues include hidden requirements, contradictory instructions, strict tests and incomplete grading criteria.
The HN debate centered on measurement: equal pass rates do not mean equal agent cost or navigation behavior.
AI coding is shifting from picking one assistant to orchestrating several agents. Omnigent is an open-source meta-harness with shared sessions, guardrails, and human-in-the-loop workflows.
Copilot now has its first selectable open-weight model. GitHub says Kimi K2.7 Code starts in VS Code for Pro tiers, with Business and Enterprise admins required to enable it by policy.
HN interest centered on whether the model feels useful in real coding loops, not just on the benchmark table.