AI was just the octane

I've spent most of my career in small startups. Even when they were midsize, it's because they grew to midsize while I was there. athena and PayPal have really been my only excursions, and athena barely counts, because it had that founder-feel the whole time I was there.

PayPal is a different animal. The structure. My god, it's full of orgs.

There's nothing inherently bad about that. It's just the nature of being a company that size. But for an autistic person? Oof. The number of teams, and the social cues on how to navigate everyone's scopes, could break your brain on their own. That's before the unwritten rules, and the ton of social work it takes to get an idea in place.

But I found a way to do good work here. I've learned to steer clear of multi-layer solutions that reach too deep into anyone else's territory. I've done well in this scope, but it was inevitable that my ideas were going to outgrow my safety cage.

This week they did. Naturally, I didn't wisely, slowly, and calmly peek my way out of it, looking around carefully and asking for lots of help along the way.

Nah. I Shawshanked that shit.

What I got wrong isn't only that I moved too fast, although I did. It's that the speed took something with it that I didn't know I'd been using.

The idea arrived whole

Like any good AuDHDer, I didn't build the concept layer by layer, carefully testing the boundaries and assumptions over weeks and months. That's not how my brain works. My work happens in the background. The idea sprang fully formed, and I just knew its shape, limitations, and details.

I'm not saying that to say it was impulsive. It wasn't. I legitimately did percolate this over months. It's just the way neurodivergent brains work. I take in all of this input and my brain churns away at it subconsciously over time until it spots the patterns, and weaves all of the inputs together into one crafted solution that covers all the angles. It's always been a strength and how I've built platforms over features my entire career.

When it all comes together, it's fast and exciting. I was, and still am, full of unbridled optimism about sharing it with others.

Here is the part that used to save me from myself, though, and I never used to think of it as protection. In the past I'd have been gated on bursting onto the scene with a proposal like this by the logistical work of drafting the documents and the charts, and talking to people about the things I did need to fill in. Weeks of that. I thought of it as overhead.

Now I had AI tools to put all of that together, and I can just start full tilt mumbling the red string lines at anyone who'd listen.

That approach isn't how it works with, well, anybody. But especially not how it works with big orgs.

The room

I presented the idea at a meeting with my eng team and architect. I had the perfect visual, and I spent the whole time describing the beauty of the solution.

The questions came, and every one of them took me down some rabbit hole of how it all would work. But coherence? Nah. Who's got time for that?

I don't think it's a trade secret to say that AI was involved in this solution, but I'll keep most of it confidential as it's PayPal's not YarBros. In the middle of introducing it, somebody asked an innocuous question that led me into eight minutes on the criticality of prompt-caching hygiene as foundational. Eight minutes. in a room that had not yet agreed that the platform should exist.

People were quite reasonably asking about whether this team or that org had weighed in or any number of logistical challenges. But mostly, they were just lost. One set of questions from the architect, a really smart guy whom I absolutely respect, literally had me dashing off a thirteen-page AI-drafted primer on a topic I figured I was explaining to him 10 minutes after the meeting.

But you can see how I misunderstood his question, right? I think it was something along the lines of "What the hell are you talking about?"

There was anger, frustration, and incomprehension. I was on an excitement-fueled high amped up on AI-octane.

And it showed.

What I thought was going to happen: they'd all get it and be immediately bought in. I figured stories would drop the next sprint, and we'd be off.

The reflection

It's also no secret here that I'm quite into AI as a tool. I've gotten really used to, if not addicted to, shipping fast in my personal repo, and when this opportunity came along, I wanted to ship fast here too.

That's where I hit the intersection. Of autism and rough social comprehension to begin with. Of AI-speed and the expectation that everyone was ready to move at it. And of the exuberance that comes with a new idea that has you bubbling over to share it.

I don't think the primer was wrong. I'm also not sure anyone could tell in that context if it were or not. It was impulsive and presumptuous, and its speed only brought more attention to a possible suspicion that this was me following an eagerly sycophantic AI into slop.

It wasn't. The solution came first. AI was just the octane.

But how will anyone else know that?

That's the thing I didn't see coming. All those weeks of building decks and chasing diagrams weren't only pacing. They were evidence. A document that took two weeks to assemble proved, by existing, that somebody had spent two weeks thinking. I never had to make that case. The artifact made it for me.

That inference is gone now. Thirteen pages that land in an afternoon prove nothing about the afternoon before them, and to a reasonably suspicious colleague they argue the other way.

And I don't have the fallback, either. My ideas arrive whole. There's no layer-by-layer trail behind them, no sequence of half-right drafts I can point at and say, look, here's where I tested that assumption. I've never had visible working. What I had was artifacts that looked expensive.

Both of my receipts went at once, and I found out about it in a room.

Slower

So things will slow down now, and it will be good. I'm smacked out of my coke-binge energy, and back to the honest work of socialization and consensus building, which is what execution requires in big orgs. Even if YarBros isn't quite there yet.

I had a good talk with the EM today, and there's another session next week with him, the architect, and my biggest doubter. Where I'll bring no primers, but I will bring a clearer set of value, a visible openness to collaboration, and a renewed respect for the methodical work of moving things forward.

I'm still talking about the vision in present and future tense, because I don't think there's anything wrong with the vision. There's just everything wrong with the delivery.

But I'm still me.

I had Claude book the meeting.