AI
Notes and interactive projects on LLMs, agents, and the stack that runs them.
Writing
Introducing Repodify: turn a stretch of a podcast into one tailored digest
Search a show, pick a stretch of episodes & it writes u one shorter, tailored digest — on ur GPU, or w/ ur own keys. Not a SaaS that keeps the audio.
- Repodify
- Podcasts
- Local AI
Introducing claude-code-py-ts-pg: a full-stack template w/ Claude built in
Why & what of my new GitHub template (Python/FastAPI + TypeScript/React + Postgres) w/ a committed Claude Code harness: rules, skills, subagents & least-privilege permissions, treated as production infrastructure.
- Claude Code
- Template
- Full-Stack
Classifiers in Harnesses
A field guide to the classifiers wrapped around modern LLMs — security & jailbreak screens, prompt-injection probes, tool-risk & authorization gates, and the probabilistic classifier in front of Anthropic's Fable model.
- Classifiers
- Harnesses
- AI Safety
The Past Three Months in AI (March–May 2026)
A post-blackout recap of the AI frontier: the model wave, the architecture pivot away from raw scale, agentic coding's convergence, the MCP+A2A protocol stack & unprecedented capital — translated from my Farsi original.
- Frontier Models
- Agents
- Translation
How I Actually Use AI Every Day as a Software Engineer
My field-tested setup for AI-assisted engineering: context management, skills, sub-agents, MCP, hooks & plan mode, spec-first development, dynamic workflows — and why verification is the skill that actually 10x's you.
- Claude Code
- Agents
- Workflow