SwiftLee author on shipping 3 apps in a year with AI agents: hooks beat instructions
avanderlee · reddit · 2026-11-13
Antoine (SwiftLee, 70K+ installs of his agent skills) shares lessons from a year of agentic Swift/Xcode development, during which he shipped 3 apps after his blog traffic and course sales declined:
- Focused task context: minimal context per task; long noisy sessions degrade quality and burn tokens
- Agent instructions: short AGENTS.md + reusable skills, but hooks are far more reliable than instructions for enforcement
- Validation loops: linter+formatter+tests script enforced via hooks; building CLIs just so agents can verify functionality is worth it
- Tool independence: keep workflows in instructions/habits so switching Cursor/Codex/Claude costs little
- Orchestration: smart model orchestrating cheaper subagents beats using the most expensive model everywhere
More from coding & agent
- Nexus launches an AI context engine that turns deployment knowledge into agent-queryable graphs — rohanpaul_ai · 2026-11-14
- Developer now runs 100% of his agents locally on a Mac Mini, up from 25% earlier this year — josh_bickett · 2026-11-14
- Mediabunny ships 400 fixes from an LLM-assisted bug hunt across its 20M-download codebase — Vjeux · 2026-11-14
- Perplexity Teammates screenshot shows agents on Datadog duty, reading PRs overnight — testingcatalog · 2026-11-14
- Nexus launches as an agentic workforce for AI deployment teams — dr_cintas · 2026-11-14
- Nous Research's Hermes Agent Desktop gets a live aurora landscape theme — Teknium · 2026-11-14