AI Terminal
BetaLLM selection matrix, AI infra leaderboard, and production warnings — sourced from newsletters.
LLM Model Selection Matrix
| Model | Provider | Score | Speed | $/1M Tokens | Context | Best For | Type |
|---|---|---|---|---|---|---|---|
| Qwen3.5 | Alibaba | — | — | — | — | Open | |
| Composer 2 | Cursor | 72 | fast | $2.50 output / $0.50 input (standard); $7.50 output / $1.50 input (fast mode) | — | multi-file coding workflowsrefactoringdebugging | Closed |
AI Infra Leaderboard
Open vs Closed Radar
Open Source (1)
Closed Source (1)
Agentic Architecture Playbooks
No AI playbooks extracted yet. Frameworks appear when newsletters teach agentic architectures.
🚨 Production Warnings
- ⚠️fast mode is 3x more expensive than standard mode
What Authors Think
How to use AI-generated code
Developers should stop reading AI output and instead steer it — treat AI as a fast typist executing your will, and if it doesn't work, just ask it to rewrite to your exact specifications.
AI and cognitive labor
AI will eventually exceed human capability across most cognitive domains, making intelligence a commodity on tap and displacing cognitive labor at scale.
AI commercialization and design ethics
The shift from academic AI research to rapid commercialization is a reckless arms race that puts designers in moral tension, as the tools they're pressured to use originate from systems designed to harm rather than help people.