🎙️ How I AI: Fable 5.2 Review + How I fight token fraud as a solo founder + Building a Claude Code family OS
- ID
- 47112
- Status
- summarized
- Published
- 17 Nov 2026, 12:02 AM
- Fetched
- 17 Nov 2026, 1:43 AM
- Provider
- Lenny's Newsletter
- Category
- product-startup
- Original URL
- https://www.lennysnewsletter.com/p/how-i-ai-fable-52-review-how-i-fight
- Source URL
- https://www.lennysnewsletter.com/feed
Summary
- Score
- 5.5
- Created
- 17 Nov 2026, 1:43 AM
- Tags
- Audience
- ai_agent_usersvibe_codersdevelopers
What happened
Lenny's Newsletter 'How I AI' episode features Scott Farrar, a math educator and parent who built a shared household system on Claude Code plus Obsidian, storing context in Markdown files he can inspect and correct rather than in long chat threads — his stated reason being 'memory drift,' not model intelligence. Concrete pieces: photographing every appliance to build a searchable inventory of model numbers, serial numbers and warranty details; a custom linter that stops verbose agents from bloating documents like the household to-do list; and a GitHub-for-shared-scripts plus Dropbox-for-everyday-files split chosen because GitHub added more complexity than his wife wanted. The episode page also headlines a 'Fable 5.2 review' and 'how I fight token fraud as a solo founder,' but the text provides no details on either.
Why it matters
The transferable pattern for anyone running agents is the linter, not the household story: an agent that writes to persistent files will pad them over time, so a check that rejects over-verbose writes is what keeps the files reusable across sessions. If you are choosing storage for a multi-person agent workflow, copy the Farrar compromise — shared scripts in git, everyday files in Dropbox — because the technically cleaner option (all-GitHub) was rejected for adding friction to the non-technical user. Do not plan anything around the Fable 5.2 or token-fraud segments from this text; there is no substance on either.
Discussion angle
Ask people who run agents against persistent files: what has actually stopped your agent-maintained docs from bloating — a linter, a size cap, a rewrite-every-time rule? And where have you had to downgrade to a 'worse' tool because a non-technical collaborator would not use the better one?