Claude/Code/Reddit/26/01/Boris shares his setup with 13 detailed steps
Blog/mgranda/Claude Code On-The-Go imported - blog post about running Claude Code agents from a phone using Termius, Vultr VM, and Tailscale
Person/Mehul Gupta/Blog/25/12/Andrej Karpathy warns Programmers imported - Medium article discussing Karpathy's perspective on the evolving programming landscape and the shift to AI-assisted development
Updated Person/Boris Cherny with comprehensive information including professional background, education, publications, and online presence

Linked references — 1 page

2026-01-02 Fri

2025-12-28 Sun

LangChain/Ecosystem
- Moved the LangGraph assistant configuration notes (Annotated[Literal[...]], json_schema_extra, __template_metadata__) to LangSmith/Docs/Observability Studio/LangGraph assistant configuration.
DevContainer/Report/25/12/Multi-Env DevContainer Setup
- This morning I was about to start on 🐛 fix(devcontainer): docker-compose.override.yml not picked up by JetBrains Gateway rebuilds · Issue #718 · codekiln/langstar with the plan to make multiple devcontainer configurations in Langstar, and I opened a devcontainer for JetBrains/RustRover inside of a git/worktree that was at <repo-root>/wip/<branch>. When I got it open, I had problems with accessing git, which makes sense; there was no git/.git directory.
  - This made me wonder ... what's the fastest, best, most secure way to spin up a different devcontainer for each AI/Coding/Agent while keeping them isolated? Sure, git worktrees let you parallelize development, but the ability to do git work in that environment is conditioned on your ability to access the .git directory.
  - But if I really am making a clone from VCS of the entire repo, including the .git dir, and running it in its own unique space with Docker, then why are git/worktrees even necessary?

Linked references — 1 page