I have ADHD. What that means for me is I'm constantly chasing novelty. I get bored fast. Repetitive work makes me miserable.
For years I thought that was the problem. Turns out the problem was the environment.
When I started my own company, I figured out that ADHD wasn't a liability in a job where every day is different, where every month there are new tools to learn and new problems to solve. Novelty-seeking is a superpower in a career built on doing things that haven't been done before.
The trick wasn't fixing my brain. It was building the right room.
The pattern I kept running into
A lot of what looks like a deficit in one setting is a strength in another. Your brain isn't broken. It's a different instrument. The question is whether you're playing the right song.
I will not commit every day to solving small problems, but rather commit every month to solving large problems. That's not a workaround — it's the actual shape of the work I'm built for.
My job is to do things that haven't been done before. New tools every quarter, new problems that don't have Stack Overflow answers yet, systems that need to be invented rather than maintained. That environment feeds the part of my brain that needs novelty to stay engaged.
Asked whether I'd ever want a regular job again: never. I don't work well in those environments, and they don't really let me be myself.
What the work actually looks like
Here's what measured looks like, pulled from an auto-generated Claude Code usage report covering about eight weeks in mid-2026:
- 855 messages over 262 hours.
- 246 messages in the morning, 343 in the afternoon, 266 in the evening, and zero between midnight and 6am.
- Peak hours around 10:00, 15:00 and 19:00.
Sessions usually open from a substantial pre-written brief — a PRD, a roadmap, a handoff file, or an "execute this prompt" pointer — then run for hours, sometimes overnight. I often watch running sessions from my phone.
That pattern doesn't look like traditional work, and it shouldn't. The environment is different, the tools are different, and the cadence is different. What matters is that it fits the way my brain works: long blocks of focused direction followed by the agent running autonomous work, punctuated by decisions and validations when they're needed.
The accessibility system
The Brain Vault is the system I built for this. In my own framing: this is largely designed as a support for my ADHD. This is an accessibility tool.
A notes app is built for retrieval: where did I put that? The Brain Vault is built for connection and action: what does an agent need to know right now to act on real context?
The architecture reflects how I actually work:
Two layers, on purpose. A raw layer holds imported source material and is never edited; a synthesized layer — concept pillars, notes, timelines — is written on top and links back. An import pipeline splits a batch into one file per item, removes near-duplicates at a 0.80 similarity threshold with every decision logged, then scores each item for signal strength before anything gets in.
Frontmatter as the agent's index. 96% of documents carry structured metadata. Undated material gets an inferred date and a stated confidence level, not a guess dressed as fact.
17 agent playbooks define named routines: imports, interlinking, health checks, weekly review. A routing layer tells an agent where to start and which sources are stale.
Operating documents that work as external scaffolding: a live priorities document, operating assumptions held as confidence-scored hypotheses, an energy map for scheduling against actual focus, and an accomplishment ledger built as a deliberate counterweight to undercounting my own output.
PageRank for importance. A standard-library Python script builds the link graph and ranks the synthesized layer. The same scores drive a 3D visualization plugin — "Cosmos" — that maps domains to stars, concept pillars to planets, and evidence notes to moons. The dense analytical graph is never rendered directly; the plugin renders a sparse version. The plugin is built with 31 tests, though one feedback round is still open.
Measured totals: 1,765 documents (869 synthesized, 822 raw imports), 7,075 links in the synthesized layer with a median of 5 per document, 50 concept pillars across 11 domains.
Each of those features answers a specific failure mode. The triage pipeline exists because I can't trust myself to filter signal from noise in real time. The energy map exists because pretending I have consistent focus across a day has never worked. The accomplishment ledger exists because my brain discounts finished work the moment it's done.
The system works. It's a highly effective accessibility tool, and it runs on the same recognition: instead of fighting how my brain works, build the room where that's an advantage.
Why this matters for builders
If you're building with AI agents, the question isn't whether your brain works differently. The question is whether you're building an environment that fits.
The same principle applies to tooling, to workflows, to the systems you run. Don't force a process that worked for someone else's brain, someone else's schedule, someone else's company. Build the room that fits the work you're actually doing, with the constraints and advantages you actually have.
Freedom didn't come first. The work came first. The right environment came from recognizing what the work needed and what my brain needed, then building toward the place where those two things overlap.
You might just need a different room.