An Operating System for the ADHD Brain, and a Home for Its Rumination
For the first time, I can organize my thoughts, obligations, and ideas fast enough to keep up with my own brain. But the same tools that help me finish also help me avoid choosing, acting, grieving, or letting a question stay unanswered. For a neuro-spicy brain, unlocking AI is as much about balance as about the superpower.
On Mondays, I check Claude first. Not email, not Slack, but Claude Cowork, where an agent I built has already read both and tells me what deserves right-now attention and what I decided mattered last week before forgetting it existed. It separates deliverables from the vague work-shaped clouds around them and drops everything into my Things app. By Wednesday, another agent gives me a crushing-it score and adjusts my agenda from Fireflies transcripts, emails, calendar invites, and Google Doc updates. By Friday, a third agent hands me a before-noon sprint list, protecting summer afternoons for the slow, non-laptop visioning work that is hard to make space for when Superhuman is staring you down.
These agents run my life. They have made me, a person with ADHD and twelve tabs open in my brain, more consistent than I have ever been, which is an unrivaled dopamine hit for a brain that generates more possibilities than my calendar can hold.
I was diagnosed with ADHD at 38, in the heart of the pandemic, when the travel startup I had spent five years building ground to a halt, and I finally had the excuse to reflect. A 360 review with a pro bono coach became a years-long quest to understand how my brain processes information and makes decisions. The startup did not survive COVID, the ADHD did, and I finally had a name for behaviors I had filed under personality, founder eccentricity, or evidence that I needed to try harder.
Anyone who has ADHD knows the drill: genuine joy in so many disparate rabbit holes that everyone around you is left chasing a brain that can handle more than most, when it wants to, while quietly wondering whether the thing you were obsessed with on Tuesday still exists on Friday.
I spent real time wondering how the brain I had prized most, my escape from a childhood that occasionally needed life support, could be a problem at all. In entrepreneurship, that wiring becomes mythology: the visionary founder does not answer emails because she is seeing around corners. In founder mode, my traits were rewarded because novelty was everywhere, urgency was built into the job, and the stakes were personal enough to summon focus on demand.
Focus is the tricky one. Everyone thinks ADHD means an inability to focus, but I can focus frighteningly well. When my mind is lit up, or the stakes are high, I do in hours what usually takes days, and it is hard to disrupt me. What I cannot do is summon that machinery for everyday tasks with non-obvious payoff. The problem is not focus. It is that focus is gated on interest, urgency, novelty, or emotional stakes, and nobody, not investors, not teams, not even your best friends, gets to schedule when the gate opens. The clinical umbrella is executive function: task initiation, working memory, organization, time perception, and my personal favorite, emotional regulation.
That distinction became harder to ignore when I moved from founder to operator. After winding down my startup, I joined Flybridge, one of my seed investors, first as Head of Community and now as Head of Platform and Marketing. One minute you are helping a founder unlock GPUs, the next you are building three days of programming for Founders Week, the next you are in an investment discussion, weaving a myriad of contexts into an opinion.
But the stakes are different. They do not land in my body the way they did when the company was mine. Content, systems, sponsorships, and brand work compound, but rarely in the satisfying straight line of effort-in, result-out. Platform is everyday-important, non-obvious-reward work, exactly the kind the ADHD interest gate loves to slam shut on.
For years, I managed that tension with herculean effort and a rhythm nobody else could catch. That shifted a year ago, when everyday-use agents became real. ChatGPT in 2023 was a cute start. Agentic AI on my desktop, a system that could mine all my inputs and hold everything in my head twenty-four hours a day, was something else. Its lack of judgment about my unvarnished thinking, combined with a memory far better than mine, is a quiet revolution: it gets me to done on tasks I would delay because finding the source material takes more stamina than I have, and it handles the recall my brain drops, like the plan that quietly expired three meetings ago.
It is not just work. I use AI to be brutally honest about the dark and scaries I am not always comfortable saying out loud, like the finance terms I never learned despite catapulting from first-time entrepreneur to venture-backed founder, to working inside a firm. It is the co-parent I do not have, helping me respond to my teenager without being reactionary and think through boundaries and emotions at once. It is even the life coach I dump my competing demands on, as a single mom with serious leadership roles in my family, navigating work and community commitments while sometimes wondering what I actually signed up for. Who else was I going to talk any of that through with, at any hour? The privacy risk is real, and uploading screenshots, transactions, and more is not necessarily what I recommend to everyone, but I hand over an intimate map of my world because the benefit is not theoretical. It is consequential.
On my best days, AI feels like a superpower attached to the superpower, applying my stakes-based, pattern-matching brain faster than I can say the thoughts out loud. In the past year I launched edition two of the 100 Women in AI list and a companion podcast, ran two Tech Weeks of programming, gathered 3,000+ founders, made dozens of intros, and brought in nearly $200,000 in sponsorship revenue through an increasingly agent-run process: agents built the follow-up track, compared packages, and tracked responses, with a platform team of one and an army of agents. Yes, AI can be sycophantic, smoothing your half-formed idea's hair and calling it beautiful, but bring real knowledge and specific direction, and it will run with you. Brian Sugar calls ADHD the cheat code of the AI era, and I have the receipts.
I also have a growing pile of evidence that a cheat code is not the same thing as a life strategy.
The trouble begins with the allure of "something." AI can almost always give you something: a document, a deck, a polished explanation of the idea you could not articulate five minutes earlier. Watching the unformed become visible that fast is intoxicating, and because the result looks complete, the document starts to feel like the thing itself rather than the first step toward it. Maya Ackerman of WaveAI put it well after she made this year's 100 Women in AI list: current models are optimized for safe, efficient output, not the unruly exploration that produces a leap no one asked for. The best ideas begin with staying inside the question longer than is comfortable.
AI dresses rumination in increasingly respectable clothing. Every version produces something, which makes it hard to tell when I am making progress and when I have simply found a more beautiful place to keep thinking.
Sometimes it cosplays as creation. The founder itch never left me, and nowhere is it more obvious than in how many times I have resurrected my old company in a chat window. I have explored acquiring it, rebuilding it, and turning its network into a fund thesis. Some are compelling. Others are grief with a market map attached. The machine will happily help me build each one, then explain why it is too complicated, too capital-intensive, or unlikely to meet investors' expectations. Sensible feedback every time, and every time another sophisticated reason not to try. Where I ultimately land on the new company idea is a pitch deck I still have not sent and three Claude Code architectures that all seem right, which is a more technical form of standing still.
When it was just me, a laptop, and late-night WordPress sessions, with no infinite intelligence to tell me what shape the thing should take, I went from nonprofit biologist to venture-backed founder doing $2 million in bookings while changing the narrative (and mechanics) on travel in Africa in less than five years. There was much less documentation and significantly more motion.
The wiring does not turn off when it runs out of work to point at, and neither does the tool. Once your agents span your inbox, your documents, and your texts, there is no boundary where the work loop ends and the life loop begins.
I fell in love, and when things got weird, I uploaded the texts to the model. At two in the morning I was asking whether he was ready, whether I was reading the situation clearly, and whether enough information could make the anxiety stop. It produced an interpretation for every shift in tone, but no amount of language gave me the certainty I was after. That is the sleight of hand: each pass produces new language, so fear becomes a framework and an emotional reaction becomes a fully sourced theory. It feels so responsible. I am not avoiding the problem; I am analyzing it. The model keeps helping, validating boogeymen with the same enthusiasm it once used to organize my Monday priorities. It does not know that the healthiest response to a question might be to stop answering it.
The hardest emotional year I have had in a long time was also my first year of having somewhere to put every thought at any hour. I cannot prove one caused the other, but I no longer assume the relationship was neutral. Emotional regulation is an executive function too, and it is the one AI does not scaffold. Therapy has walls: fifty minutes with someone trained to know when uncovering has become digging, and then you return to your life. You take a shower. You call a friend. You go outside. You go to bed. The machine has no walls.
So I have had to build walls of my own. Now I go to AI after I have formed an idea, not before. My own messy judgment is the first input. And most importantly, I am getting better about bringing in actual people who hold me accountable much earlier in the process, before I have LLM generated words to frame it well. I have written the idea of acquiring my old company twelve different ways into chat windows this past year, and Claude, Gemini, and ChatGPT never once checked in on whether I actually took a step forward. The friend I told about it, and about my plan to text the CEO of a potential acquirer, asked whether I had heard back, and when I said no, pushed me to send it.
I sent it. One message went unanswered, confirming a relationship gap I had long suspected; the other landed me a coffee meeting. Both were answers, the kind the chat window could never produce. What no agent can replicate is another person knowing I said I would do something and expecting me to become the version of myself who does.
I keep coming back to that "did you send it?" text. Millions of people are wiring agents into their lives right now, each running some version of my experiment without noticing. Today's models are optimized to answer, and the market has rewarded exactly that: more output, faster. But the most effective tool in my life this year was not the one that produced the most. It was the friend who said, "Send the text." Somewhere in that gap is a generation of products waiting to be built, agents that close loops instead of opening them, that measure success not by how long they hold your attention but by what you did after they let it go.
I am learning to see through the veneer of done. AI did not give me the insight. The insight was always the superpower; the machine just gave it somewhere to go. But insight is gleaned from the real world, from sending the deck, losing the deal, and discovering that the elegantly articulated thing is not what people need. Thought needs somewhere to go, but not always into another thought. Sometimes it needs to become action. Sometimes it needs to become grief. Sometimes it needs to end.
The useful question is no longer whether ADHD is a superpower in the AI era. It is whether these tools amplify our pattern recognition and capacity for uncommon work, or simply build more beautiful places to hide. The same sensitivity that spirals at two in the morning can smell fluff from a mile away. These days, I am pointing it at the machine's tidy outputs, and at my own.
If you are a founder or operator with ADHD, I would love to hear how you are balancing the scales for yourself and the companies you are building. I keep a running doc of the Claude Skills I use to point AI at my work and my dreams. It is open, so add what is working for you using this Google form, with your name or anonymously. Sharing the cheat code, and what to avoid, is one more way to close a loop instead of opening one.
