
Changelog & Friends Finale & Friends
Mar 2, 2026
A farewell conversation about retirement and a two-decade tech partnership. They break down Ladybird’s move to Rust and how AI helped port huge C++ codebases. Rust tooling and high-performance JavaScript tools come up. A lively debate on self-hosting, Tailscale identity, and building self-hosted CI. They speculate on AI reshaping code review, software lifecycles, and future engineer roles.
AI Snips
Chapters
Transcript
Episode notes
Start Migrations By Translating Tested Code
- If you have a well-tested C++ codebase, try translating it with AI rather than rewriting from scratch to minimize risk and get quick validation.
- Adam recommends giving tools existing code + tests so agents can target feature parity and preserve architecture during a gradual migration.
Funding Shapes AI Access And Pricing
- Large AI model providers must monetize access; market dynamics favor deep pockets like Google who can subsidize services.
- Adam explains Anthropic/OpenAI need revenue while Google can operate at thin margins, shaping the ecosystem and pricing moves.
Using Cloud Code Inside Tidewave Then Hitting Limits
- Jerod described using Cloud Code subscriptions inside Tidewave via the Agent Client Protocol to piggyback existing quotas.
- That made agent use frictionless until Anthropic changed rules restricting third-party use, forcing separate API tokens.
