Welcome to the daily AI news brief for vibe coders. It is Tuesday, June 2, 2026, and the last 24 hours hit two things every builder cares about: the cost of your tools and the protocols they speak. GitHub Copilot’s token billing went live yesterday, Firebase Genkit reached a major milestone with native MCP support, and Microsoft Build opens today. If you ship on Shopify, Claude, Gemini, or Firebase, these moves change your budget and your wiring this week.
TL;DR
- GitHub Copilot’s token billing is live; power users report 10x to 50x cost jumps.
- Firebase Genkit 2.0 reaches GA with native MCP, streaming, and Cloud Trace.
- Shopify Scripts editing is frozen; all Scripts stop working June 30.
- Claude Agent SDK now connects MCP servers in the background and uses Task tools.
- Microsoft Build 2026 opens today with Windows Agent Runtime and A2A updates.
- Anthropic raised $65B at a $965B valuation, now the top private AI company.
Firebase Genkit 2.0 reaches general availability
What shipped. Genkit, Google’s open-source AI app framework, reached its 2.0 general-availability milestone. The release adds streaming responses, deeper observability with traces wired into Cloud Trace, and native Model Context Protocol server integration, alongside Python and expanded Go support.
Why it matters for vibe coders. If your stack leans on Firebase, native MCP lets Genkit flows share the same tool servers your Claude and Cowork agents use. One protocol across backend and agents cuts glue code.
What to do today. Spin up a throwaway Genkit 2.0 project, register an MCP server you already run, and confirm a flow can call a tool through it. Note any breaking changes before touching production. Firebase Genkit
Shopify Scripts retire June 30 — the edit freeze is already on
What shipped. Shopify Scripts are on a hard deprecation path. Editing and publishing new Scripts has been blocked since April 15, and every existing Script stops running after June 30. Functions, compiled to WebAssembly and available on all plans, are the replacement.
Why it matters for vibe coders. Scripts powered checkout, discount, and shipping logic on Plus stores for years. After the cutover there is no extension path, so any custom cart behavior you rely on simply stops.
What to do today. Inventory every active Script, map each to a Function API equivalent, and rebuild and test discount and shipping logic in a development store. Use Claude Code to scaffold the boilerplate. Shopify
Claude Agent SDK connects MCP servers in the background
What shipped. Recent Claude Agent SDK builds changed two defaults. MCP servers now connect in the background, so sessions start immediately and a slow server reports a pending status during init instead of blocking startup. Headless and SDK sessions also use the Task tools, TaskCreate, TaskUpdate, TaskGet, and TaskList, for multi-step tracking.
Why it matters for vibe coders. The Task tools give multi-step agents a structured, inspectable to-do model, which makes long autonomous runs easier to debug.
What to do today. Update your SDK, restart an agent that loads several MCP servers, and confirm it no longer stalls on the slowest one. Mark June 15: programmatic Claude usage moves to a separate monthly credit pool that day. Claude release notes
GitHub Copilot’s token billing is live
What shipped. As of June 1, GitHub Copilot bills by tokens consumed rather than a flat monthly rate. One credit equals a cent, and input, output, and cached tokens all draw it down at each model’s published rate. Developers running agentic sessions report bills jumping 10x to 50x, and the reaction on GitHub’s own thread has been overwhelmingly negative.
Why it matters for vibe coders. If you lean on Copilot for long agent runs, your cost is now variable and tied to how much context you push, and unmonitored loops can run up a bill fast.
What to do today. Check your usage dashboard, set a hard credit cap if your plan allows, and trim oversized context windows in your most-used flows. Compare projected monthly spend against raw API rates before you commit. TechCrunch
Microsoft Build 2026 opens today
What shipped. Microsoft Build 2026 begins today, June 2, in San Francisco and runs through June 3. The developer preview points to a Windows Agent Runtime, a multi-model Copilot, and updates to the Agent-to-Agent protocol for coordinating agents across frameworks.
Why it matters for vibe coders. Protocol changes that land at Build often reach indie tools within weeks, and A2A updates sit close to the MCP wiring many of us already use.
What to do today. Block thirty minutes for the keynote recap, and flag any agent-interop or protocol news that could affect how your agents talk to each other or to external tools. DEV Community
Also worth noting
- Anthropic raised $65 billion at a $965 billion post-money valuation, making it the most valuable private AI company and surpassing OpenAI’s private valuation. Build Fast with AI
- SoftBank committed €75 billion to build roughly 5 gigawatts of AI infrastructure in France. Build Fast with AI
- Gemini 2.0 Flash and Flash-Lite shut down June 1; apps on Firebase AI Logic should migrate to gemini-3.1-flash-lite. Gemini API
- Cognition, maker of the Devin agent, raised $1 billion at a $26 billion valuation as agentic coding demand climbs. LLM-Stats
Build of the day
Ship a token-budget guard for your agent runs in under two hours. With Copilot now metered and Claude’s programmatic usage moving to a credit pool on June 15, an unmonitored loop is a financial risk. Write a small Python or TypeScript wrapper that logs tokens per call, sums them per session, and stops the run when it crosses a ceiling you set. Print a per-session cost estimate from each model’s published rate, and wire it into a Claude Code subagent so every long task reports its spend. The first time it halts a runaway loop, it pays for the build.
FAQ
What changed with GitHub Copilot billing on June 1, 2026?
Copilot moved from a flat subscription to token-based billing. One credit equals a cent, and input, output, and cached tokens all draw it down at each model’s published rate. Developers running agentic sessions report cost increases of 10x to 50x, with heavy backlash on GitHub’s thread. Check your usage dashboard and set caps before running long loops. TechCrunch
What does Firebase Genkit 2.0 add for app builders?
Genkit 2.0 reached general availability with streaming responses, observability through traces integrated with Cloud Trace, and native Model Context Protocol server integration, plus broader Python and Go support. For builders, native MCP means Genkit flows can share the same tool servers your agents already use, which reduces glue code. Firebase Genkit
When do Shopify Scripts stop working?
Editing and publishing Scripts has been blocked since April 15, 2026, and all existing Scripts stop running after June 30, 2026. There is no extension path. Map each active Script to a Shopify Functions equivalent and test discount and shipping logic in a development store before the cutover, since custom cart behavior breaks once Scripts are gone. Shopify
How does the Claude Agent SDK handle MCP servers now?
Recent SDK builds connect MCP servers in the background, so sessions start immediately and slow servers report a pending status during init instead of blocking startup. Headless and SDK sessions also use the Task tools for multi-step tracking. Update your SDK and restart an agent that loads several servers to confirm it no longer stalls on the slowest one. Claude release notes
What should I watch for at Microsoft Build 2026?
Build runs June 2 to 3 in San Francisco, with a Windows Agent Runtime, a multi-model Copilot, and Agent-to-Agent protocol updates in focus. Announcements there often shape agent tooling and protocols the broader ecosystem adopts within weeks. For vibe coders, the A2A and agent-interop news is most relevant, since it sits close to the MCP wiring many agents already rely on. DEV Community
Why did Anthropic’s valuation jump to $965 billion?
Anthropic raised $65 billion at a $965 billion post-money valuation, making it the most valuable private AI company and pushing it past OpenAI’s private valuation. For builders, heavy capital in frontier labs signals continued investment in Claude and MCP tooling, though funding alone is not a reason to change your stack. Build Fast with AI
Sources
- Firebase Genkit: https://firebase.google.com/products/genkit
- Shopify — Scripts deprecation: https://shopify.dev/changelog/shopify-scripts-will-be-deprecated-on-june-30-2026
- Claude release notes: https://platform.claude.com/docs/en/release-notes/overview
- TechCrunch — GitHub Copilot token billing: https://techcrunch.com/2026/05/30/what-a-joke-github-copilots-new-token-based-billing-spurs-consternation-among-devs/
- DEV Community — Microsoft Build 2026 preview: https://dev.to/akaranjkar08/microsoft-build-2026-developer-preview-windows-agent-runtime-multi-model-copilot-and-the-agentic-5pi
- Build Fast with AI — AI news, June 1, 2026: https://www.buildfastwithai.com/blogs/ai-news-today-june-1-2026
- Gemini API — changelog: https://ai.google.dev/gemini-api/docs/changelog
- LLM-Stats — LLM updates: https://llm-stats.com/llm-updates
About the author
Robert McCullock runs Design Delight Studio, building AI-assisted Shopify storefronts and multi-agent tooling for sustainable DTC brands. See selected work in the professional portfolio.