Daily Highlights
1. GitHub Copilot Shifts to Token-Based Billing: Starting today, June 1, 2026, GitHub has officially retired its flat-rate subscription model for Copilot, transitioning all plans to usage-based billing via 'GitHub AI Credits'. While standard in-editor code completions and Next Edit Suggestions remain unlimited, advanced features like Copilot Chat, CLI, and code reviews will now draw from a monthly credit pool ($10 for Pro, $39 for Pro+). This transition reflects a broader industry shift toward consumption-based pricing for heavy LLM workloads.
2. NVIDIA Releases Open-Source NemoClaw and Agent Toolkit: At GTC Taipei today, NVIDIA announced the open-source release of its Agent Toolkit. The centerpiece is NemoClaw, an orchestration framework that provides templates for planning, reasoning, and execution. The toolkit also introduces OpenShell, a secure runtime designed to enforce policy and privacy controls on local and cloud hardware, alongside the massive 550-billion-parameter Nemotron 3 Ultra mixture-of-experts model optimized for long-running workflows.
3. Cirrus CI Shuts Down, Relicenses Core Tools: Following its acquisition by OpenAI, Cirrus CI has officially shut down its continuous integration service today. In a parting gift to the developer community, Cirrus Labs has relicensed all of its popular source-available virtualization and orchestration utilities—including Tart, Vetu, and Orchard—under highly permissive open-source licenses, making them entirely free for public use.
Niche Project of the Day
Burn (tanrikuluozlem/burn): Trending on Hacker News today, burn is a zero-setup, agentless CLI tool written in Go that helps developers find exactly where their Kubernetes clusters are leaking money. By pulling node and pod telemetry directly from the Kubernetes API and Prometheus, and reconciling it with real-time AWS and Azure pricing APIs, burn generates instant cost breakdowns by namespace without requiring complex dashboards or persistent storage. It even features a Slack integration that suggests optimized kubectl commands to rightsize resources or transition workloads to spot instances.