In 2025, the world didn’t get kinder, calmer, or more reasonable.
But I got better.
Not in a vague “self-care” way. In a systems way. I stopped waiting for permission and started building the thing I needed. For the first time, I felt in control.
I stopped treating my limitations as personal failures and started treating them as design constraints.
I stopped hoping someone would notice my potential and started shipping evidence.
This is the story of how I became the CEO of a tiny one-person company of one. And how AI—controversial, messy, imperfect AI—became both my unfair advantage and my accessibility prosthetic.
From 2020 through 2024, life happened at me: the blur of the pandemic, my vision fading until I reached for a cane, and mental health settling in like heavy gravity. I felt like a passenger—reacting, trying to keep up, hoping momentum would count as direction.
But I didn’t have the time or the energy to act on the feeling.
The company I started 2025 with had banned AI usage, which felt like a box. The tools that could have helped me do my best work were off-limits.
Then in February, I got laid off.
Along with the panic, there was relief—and a practical factor that kept me from spiraling: a generous severance package (3–6 months).
That cushion bought me time.
I resisted the urge to panic-apply.
I slept. I let my nervous system unclench.
Then I shipped the first win of the year, more important than any job offer: I finished writing a book in March. I started it in January. I shipped it in March. Proof-of-life. I can still complete hard things.
After that, I made a decision that sounds blunt but changed everything.
I’m a bad engineer, so I will use AI to become a good one, broadcasting the mess I make in the process.
Not to fake competence—I can’t. The bet was that AI would let me build competence faster.
So I wrote weekly field notes about frontier tech and my builds.
And I built small software projects as proof.
I also adopted a rule that saved me from perfectionism:
If it’s too hard for AI to do well, I shelve it. In a month, AI will be capable of helping me.
That rule kept me shipping and kept me pointed at the frontier.
My top 3 projects of 2025:
Pico Panic https://www.lexaloffle.com/bbs/?tid=149351#comments — a tiny game inspired by inscrutable 8-bit computer games, shipped in one week. I tried to make the Pico-8 fantasy console deliver a vector-like experience. I had a lot of fun making it, but honestly, it needs a little more tuning before game night.
Here’s the scene I keep coming back to.
I sat down to make Pico Panic, staring at a blank lua file—a language I didn’t really know. My wife was on the couch playing Yakuza. The room had that cozy, ordinary glow of a weeknight.
I opened an LLM and started pair programming without shame. I asked dumb questions. It answered. I asked better ones. While I sketched the game loop, the agent spun code in Lua. While it spun code, I wrote the music.
For a few hours, I wasn’t “one human, one brain”—I was a tiny studio.
I felt like Prometheus with a coupon code, paying pennies to borrow as much brainpower as the work required.
Eventually, the broadcast worked.
I landed at ActiveCampaign, a workplace where AI wasn’t treated like contraband, but as the transformative technology that will change the way we all work.
I shifted from being an “inconsistent star performer” (bursts of brilliance, then burnout) to being consistently solid and trustworthy.
Not because I became a different person.
Because I built a better operating system.
Here’s what unlocked the whole year:
I’m the CEO of a micro-company.
AI agents are my employees.
They don’t replace my taste, judgment, or ethical compass.
They replace the parts of work that chew up my executive function: first drafts, scaffolding, repetitive transforms, formatting, and the inertia of a blank page.
There’s a layer people miss when they talk about “AI productivity.”
My vision loss makes certain kinds of reading genuinely difficult—dense docs, sprawling spreadsheets, anything that demands constant scanning.
My vision loss makes certain kinds of reading genuinely difficult—dense docs, sprawling spreadsheets that swim before my eyes, anything that demands constant scanning.
So when I say AI is a force multiplier, I also mean:
AI is a prosthetic for information processing.
Instead of hunting, I can ask:
- “Summarize the decisions.”
- “Pull out the risks and dependencies.”
- “Find where we defined the metric.”
Things that used to take me twice as long as a sighted person now take minutes—most of that time spent waiting for the LLM to respond.
Using these tools strained relationships.
I lost friends.
I watched people I respect draw hard moral lines that didn’t account for accessibility or reality, a steady background radiation of judgment—an implicit demand that I apologize for using the very tool helping me survive.
I can’t tie a neat moral bow on that.
I have a hope:
2026 normalizes these tools enough that the stigma fades, and we can argue about ethics like adults, not a mob.
Here’s what I learned in the simplest terms:
- Don’t play someone else’s game. You play best when chasing your own goals and ambitions.
- Leverage your weirdness openly.
- Build the proof. Don’t wait for the verdict.
- Be curious about the things that annoy you – they’ll become successful projects later.
AI isn’t just a better Google.
The people who master it will go far, but there is no roadmap yet. We’re still bushwhacking the trail.
So don’t just take a class. Build things. Then build some more until the muscle becomes second nature.
Become your own CEO.
Leave a Reply