Stuart Kim
Field NotesPL. 02 · April 2026 · Stuart Kim

My second brain, sort of

A practical setup for using AI to handle the executive-function tax that quietly burns out the brain doing the actual work.

Why I built this

I'm a late-diagnosed, high-functioning ADHD adult. The way I usually describe it: the brain is fast and pattern-hungry, which is the good part. The cost is the executive function — the prioritizing, sequencing, remembering, switching, updating. That part is a tax I pay every day on the cognition I actually want to be using.

I'd been using Sunsama as my daily planner for years. It matches how my brain wants to work: one focused day at a time, intentional pulls from the backlog, no infinite scroll of overdue tasks. Then AI got good enough to wire into it. So I did.

The result isn't "I save time." It's more of my brain spent on problem-solving and less on administrative overhead. Same hours. Different distribution.

The stack

Two surfaces. Native connectors. About thirty minutes of setup.

C
Claude
The thinking layer. Brain dumps, prioritization, and synthesis happen here.
Cowork
The doing layer. Anything that needs an actual action picked up and run.
S
Sunsama
The execution layer. Where the day actually happens.
G
Gmail
Scanned for anything that should become a task.
📅
Calendar
Read for context and conflicts.

All of it connects natively now. Sunsama added an MCP server. Gmail and Calendar are connectors in Claude. Cowork picks up anything that needs to be done.

A year ago this required Zapier as a bridge. Worth noting only because the setup tax just dropped considerably.

What it actually does

Two main rituals. One in the morning. One at the end of the day.

Morning

Brain dump → prioritized day

I dump whatever's in my head into Claude. Half-thoughts, follow-ups I remembered in the shower, things I owe people, dread items. Claude scans my inbox and calendar, runs an Eisenhower-matrix-style prioritization across what I dumped plus what it found, and pushes the result into Sunsama as a planned day.

What used to be twenty minutes of staring at lists is now five minutes of typing.

End of day

Sweep → catch what I missed

Claude scans the day's emails and any docs I worked on, surfaces anything that needs follow-up, and either updates Sunsama or queues it for tomorrow. The point isn't completeness. The point is that I don't carry the "did I forget something" loop into the evening.

What a session looks like

Approximately, with the boring parts compressed:

morning planning · claude
stu morning. brain dump:
- finish the lab writeup, been circling 2 days
- coaching student follow-up notes, said i'd send by friday
- portfolio site has 2 broken images i keep forgetting
- reading for school, falling behind
- need to schedule a dentist appt for my son
- that linkedin post i half-drafted last week
what does today actually look like? clean me up.
claude scanning sunsama, inbox, calendar... (2 min)
found 4 overdue tasks already in sunsama —
pulled them up so we can decide
2 emails need replies before EOD
calendar's clear after 3pm
today (high-impact, finishable):
1. lab writeup — momentum item, finish it
2. coaching notes — committed timeline, ~30 min
3. broken images on site — small, satisfying, unblock
moved to thursday:
- school reading (block 90 min, morning energy)
- linkedin post (needs more thinking, not today)
handled:
- created task: schedule son's dentist (this week)
- flagged 2 emails to reply by 5pm
want me to push this to sunsama as today's plan?

Most of the cognitive work happens before I touch the keyboard. The dump itself is 30 seconds. The output is the day, planned, in Sunsama, with the right things at the top.

Why this works for me specifically

The neurospicy brain isn't broken. It's just running on infrastructure that wasn't built for it. AI is starting to be that infrastructure.

What I'd flag if you're trying this

Stuart Kim · April 2026 · part of the Lab at stuartkim.design
Tools mentioned: Claude, Cowork, Sunsama.