Note: none of this page was written by AI

My stack

  • Figma Make, Lovable

  • Cursor, Cline

  • ChatGPT, Claude, Perplexity, Gemini, NotebookLM

  • Ollama, LMStudio

  • n8n

  • writing good evals

Workflows I’ve automated

  • Pipeline to parse call segments for more detailed themes and multi-pass analysis

  • One-shot prompts to spin up detailed interactive prototypes

How AI fits into my process

Early

  • Help refining SQL and Python for data analysis

  • Text (call transcript, notes) analysis + theme aggregation

  • Deep Research

  • Iteration on problem framing

Messy middle

  • Iteration on problem<>solution alignment

  • Variations on messaging for alignment

  • Checking the codebase for behavior and patterns — for turning into components, refactoring work to be more in tune with standards, or just to answer my team’s questions

Late

  • Refinement of microcopy

  • double-checking framing, perspective, and marketing messaging

  • refactoring and cleaning up PRs

Yes:

  • Helping ideate is fine

  • Human has to check everything (put your damn name on it)

  • Make art by hand

  • Don’t be deceptive/manipulative

  • Show your thinking

Hard no:

  • No social-scoring, discriminatory systems, predictive policing, or surveillance

My Ethical Line

Emotional Support Tabs — who wants to read forever? Summarize, link, and bookmark to get them out of memory.

Asthma app (to be named) — early signals to help manage health and find community.

SpaceBack — Summarizes how many emails I have from different domains and sender, and how much total space they're taking up so i can free up storage space so i don't have to pay for more. Goodbye old mailing lists!

collater— using Gemini, break down my 5,000-item Watch Later list on YouTube into digestible playlists that i can binge listen to.

Experiments