Ky-Cuong Huynh

This is also an example of how post-training enhancements (here, primarily scaffolding) can unlock success on longer time horizon tasks.

https://epoch.ai/blog/ai-capabilities-can-be-significantly-improved-without-expensive-retraining

If you're worried about losing your job to AI, these are great starting reads:

https://planforai.org/

https://80000hours.org/ai/guide/skills-ai-makes-valuable/

Then consider that economic transformation gone awry is one of the *less* dangerous outcomes from AI.

This is a once-in-a-species opportunity to help ensure better futures for all sentient beings: https://80000hours.org/ai/

For book-level takes: https://www.obsolete.pub/p/sam-altman-would-hate-it-my-book

https://www.amazon.com/dp/059398109X

Tired: building custom orchestrators for your legions of agents

Wired: just let Claude do it for you: https://claude.com/blog/introducing-dynamic-workflows-in-claude-code

If you've found yourself buried in tmux panes while managing multiple Claude sessions (not to mention the subagents and agent teams within each session), Agent View is a huge improvement: https://code.claude.com/docs/en/agent-view