Software Defined Talk Episode 566: The code is actually kinda useless
8 snips
Apr 3, 2026 They dig into the Claude Code leak and whether leaked TypeScript changes the game. The conversation covers running and sandboxing coding agents safely with Docker micro VMs. They unpack the Axios NPM supply chain compromise and how malicious packages can target developer tooling. There is also a thread about email identity and whether long-held Gmail addresses should be retired.
AI Snips
Chapters
Transcript
Episode notes
Long Lived Gmail Address Story
- Coté has used the same quirky Gmail address (based on a misremembered movie name) for ~22 years and rarely publishes it, illustrating long-lived digital identities.
- He considers using Gmail’s new ability to change addresses every 12 months to pick something closer to his real name for others' clarity.
Leaked Source Code Doesn't Stop Competitors
- The leaked Claude Code revealed AI-generated, repetitive TypeScript and prompted ports (including a Python rewrite) that sidestep copyright claims.
- That shows leaked source no longer prevents competitors from cloning features since AI can re-generate or reimplement behavior quickly.
Respond Pragmatically To Code Leaks
- Anthropic should acknowledge limits of copyright on AI-generated outputs and pursue a pragmatic 'clean room and iterate' strategy rather than only legal takedowns.
- Own the product roadmap and keep shipping features to stay ahead of clones.