Agents

GitHub at Ignite: Copilot Evals, “Adaptive” code review and Copilot in the browser; Foundry adds Baseten; TensorWave commits 457 MW to OpenAI

At Ignite, GitHub puts Copilot Evals in public preview to test your skills, agents and MCP tools, lets Copilot pick the effort for each code review and lets it drive Chrome. Microsoft Foundry adds Baseten and deploys models and agents onto your own AKS clusters. TensorWave will supply 457 MW of AMD capacity to OpenAI. Also: Claude Code 2.1.327, Codex CLI 0.170.0, OpenCode 2.3.2.

By ·

Microsoft Ignite opened on November 17, 2026, and GitHub shipped about fifteen Copilot updates that day (GitHub changelog). Three matter directly to anyone who customizes their agents: repeatable evaluations, a code review that sizes its own effort, and browser use. On the infrastructure side, Microsoft Foundry opens up to Baseten and to your own Kubernetes clusters, and TensorWave signed a 457 MW contract with OpenAI.

Copilot Evals: test your skills, agents and MCP tools

GitHub Copilot Evals has been in public preview since November 17 for Copilot Individual, Business and Enterprise plans (GitHub). An “eval” is a repeatable test: you give an agent a realistic task, describe what a good result looks like, and check both the result and how the agent got there. With the /eval command or plain language you can:

The evaluation engine is Vally, an extensible local tool. Copilot Evals installs as a VS Code extension, as a Copilot CLI plugin, or from the Customize menu of the GitHub Copilot app.

Code review: Copilot picks the effort and looks for secrets

According to GitHub, Copilot code review gets four updates, open to the Pro, Pro+, Max, Business and Enterprise plans:

The same day, the “security review” specialist in Copilot CLI and the Copilot app gained the ability to run GitHub secret scanning on a developer's or an agent's changes before merge. It is off by default and consumes AI credits under the Secret Protection AI Credit SKU (GitHub).

Copilot drives the browser and signs in with Entra

Copilot Fusion and harness comparisons

GitHub says Copilot Fusion Max reaches a 71.2% resolution rate on Terminal-Bench 4.1, 5.4 points ahead of Opus 5.5 and Fable 5.2, at a cost 88% lower than Fable 5.2 and 57% lower than Opus 5.5 (GitHub). These are GitHub's figures. Fusion comes to github.com, Visual Studio and the GitHub Mobile beta apps, and adds Claude Sonnet 5.5 (Max level), GPT-6.1 Terra (High) and MAI-Code-2 (Standard). The Copilot benchmarks site adds a view comparing Copilot CLI with Claude Code and Codex CLI on the same model at the same (medium) reasoning effort, plus Terminal-Bench-Science 0.1 (GitHub).

Microsoft Foundry: Baseten, model router and AKS

Infrastructure: 457 MW for OpenAI at TensorWave

TensorWave, a GPU cloud backed by AMD Ventures, announced on November 16 a six-year agreement to provide OpenAI with 457 MW of IT capacity across several US sites, on AMD Instinct MI455X GPUs in Helios racks, as dedicated capacity for OpenAI (TensorWave, Business Wire). First capacity is due in the first half of 2027 and the full deployment by early 2028. Neither the value nor the sites were disclosed (The Next Web). TensorWave says it has secured more than 2 GW of long-term data center capacity. On November 17, Argentum AI and Kyndryl also signed a 20-year, $630 million agreement for 10 MW in Boulder, Colorado, where Argentum is deploying Nvidia B300 systems for inference (Kyndryl).

Claude Code 2.1.327

Released overnight from November 17 to 18, Paris time (release notes): claude mcp add-json <name> - reads the server's JSON from standard input, so tokens stay off the command line; /install-github-app asks for confirmation before changing a repository; /status lists broken agent, skill and command files, and hook matchers that never match. Several fixes protect data: stopping the Remote Control server in worktree mode no longer deletes git-ignored files such as .env, claude mcp add and remove no longer delete other keys in .mcp.json, and /export asks before overwriting a file. WebFetch now refuses URLs over 2,000 characters.

Codex CLI 0.170.0

Released on November 18 (release notes): WebRTC negotiation for realtime voice calls in the TUI, automatically refreshed authentication headers for OpenTelemetry HTTP exporters, and experimental support for identically named MCP servers across environments. Fixes cover MCP OAuth credential races, Windows sandbox and file replacement, and child threads losing sandbox restrictions.

OpenCode 2.3.1 and 2.3.2

Two releases on November 18. 2.3.1 moves a shell command that hits its timeout to the background instead of killing it, writes the compaction summary in the background (turn it off with compaction.background: false; an unused summary still costs tokens), reads CLAUDE.md again where there is no AGENTS.md, adds GPT 6.2 Sol and Astra for ChatGPT sign-in, and turns experimental schedules into “automations” (time, interval, cron or MCP event). 2.3.2 fixes system prompt caching for GPT-6.2 models: on gpt-6.2-sol through GitHub Copilot, uncached input per step drops from about 14,000 tokens to 3, according to OpenCode.

What we updated

Sources

Spotted an error in this article? Report it · Our corrections policy

OrigoTech logo

New: the app iPhone & iPad

OrigoTech is now an iPhone and iPad app

Read the articles natively, with “New” markers on unread ones, local search and offline reading of what you have already loaded. Free, no account, iOS 17 or later; ads appear between articles.

Rekha icon: a house drawn in a single stroke

From the same author Self-promotion

An Eulerian path, drawn with your finger

Rekha, the iPhone game by OrigoTech's author: connect every dot in a single stroke without tracing the same link twice, across 500 drawings that are all solvable.

The OrigoTech newsletter

A sourced weekly digest of AI tool news.

Read next

Agents

Eclipse Enclave sandboxes coding agents, OVHcloud launches 55 AI products with B300 and Rubin GPUs, OpenAI shows its Codex-run Performance Factory; Claude Code 2.1.329 and Codex 0.171.0

The Eclipse Foundation releases Enclave 1.0, an MIT-licensed sandbox that runs Claude Code, Codex or OpenCode in isolated containers; Kong launches Volcano; the GitHub Copilot app reaches Azure DevOps repos. OpenAI says Codex agents ran more than 1,700 performance investigations in about three months, and ships GPT-Live-1.1 in its API. OVHcloud unveils 55 products and up to 60,000 B300 GPUs, Tencent publishes the Hy4 weights, and Claude Code 2.1.329, Codex 0.171.0 and Zed 1.29 ship.

· 6 min read

Agents

JetBrains runs Codex and Claude Code inside TeamCity builds; Hugging Face hands its inference routing to OpenRouter; ngrok and Semgrep ship MCP servers

JetBrains released an experimental TeamCity plugin that installs Codex or Claude Code on the build agent, lends it credentials only for the build and records tokens and trajectories. Hugging Face now routes chat requests through OpenRouter, with failover and no markup. Also: ngrok's MCP server, Semgrep's agentic platform, Claude Code 2.1.326 and OpenCode 2.3.0.

· 5 min read

Models

Gemini 4 Hydrogen: price, benchmarks and Antigravity's new Orchestrator; Claude Code adds Code Mode, Codex CLI 0.168.0

Google released Gemini 4 Hydrogen on November 12, a fast model for everyday coding and agents at $1.50/$7.50 per million tokens, with a free API tier and a new Orchestrator in Antigravity that pairs it with Carbon or Argon. Also: Code Mode in Claude Code 2.1.323, Codex CLI 0.168.0, the agent-built AWS Emulator and Nebius's $40-45 billion capex plan for 2027.

· 5 min read