I frame the whole problem
& design the way through.

Problem:
My partner and I were running our family on 9 calendars and 2 inboxes, and things kept slipping.

Solution:
I built a custom AI agent to improve family coordination. Every night a 5-stage workflow reads everything and emails us one synthesized brief: who’s driving, what the kids need for school, what to reply to. It carries deadlines forward across runs, so nothing slips between nights. In production daily.

Problem:
To prep my team’s daily standup, I was manually scanning Drive, Figma, 9 chat spaces, Gmail, and meeting notes, often missing details that mattered.

Solution:
I built a custom AI agent to draft our standup agenda. Every weekday morning, a two-pass pipeline reads everything and writes 3-5 ranked items into our running doc: what’s blocked, what’s converging, what’s waiting on a decision. It tracks decision debt across runs, so unresolved calls resurface instead of vanishing.

Problem:
We over-trust confident AI, accepting wrong answers we’d catch if we slowed down.

Solution:
I’m designing and building Second Opinion to push back. Paste any AI response and a four-stage pass flags only the claims worth doubting and gives you what you need to question each: the hidden assumption, a counterargument, and searches to run yourself. You stay the one deciding what to trust.

I built this site from scratch: no template, no page builder. The design system lives in Figma as bound variables and maps one-to-one to the site’s CSS tokens, so design and code stay one source of truth. I built it with Claude Code, syncing Figma and code in both directions as I iterated.