I'm a self-taught technical founder with an English lit degree. I shipped a live AI product and tracked exactly how much I grew in 10 days.
Founders are unreliable narrators of their own growth. So I built a system to fact-check myself.
It started as daily logs.
Every night, I voice-record my progress: a brain dump of what I did, what moved and what I was accountable for. Talking makes it honest, even at 2am.
My AI Skill Auditor processes it, structures it, and logs it against actual work.
Self-reporting doesn't paint an objective picture. So I built in receipts - three sources cross-checked:
→ My Notion - startup-wide contributions
→ Maibel workspace - what actually shipped
→ Team chat - how I showed up under pressure
I wanted a visual to map it all. Each node on my radar chart is mapped to receipts: specific bug, the date, the delta.
I'm not just tracking technical skills, but how I think, debug, and lead under pressure. I also wanted insights into my blindspots - the ones I'd never have caught.
Of course, AI roasts went straight for the jugular.
"You are still a central dependency in too many decision paths."
Guilty as charged.
"Some quality judgments still depend on founder detection rather than robust eval loops."
To be fair - I helped design the evals. I know what good looks like before it flags it. The Auditor just thinks I should trust it more, my gut slightly less.
Probably right. But the wins feel earned too:
"From executing with AI tools to orchestrating agentic systems."
Two moments stood out from launch chaos:
1. My mental model shift: I stopped seeing AI as just tools and started thinking how to give better instructions. Every part of the system got a specific job (agents and humans alike):
ChatGPT for strategy, Codex for execution, Notion for memory, contributors for judgment.
At one point I stopped feeling like I was putting out fires but running an emergency response station.
2. A leadership win: early in the launch, when something broke, I'd jump right in to fix it.
Then I realised: when you're the fastest way to solve it, THAT becomes the problem. Every time I stepped in, I was training my team to wait for me. That doesn't scale.
So I stopped. Triaged instead: Which priority level? Whose lane is this? What do they need to own it?
Logged it, wrote a brief and handed it off.
Bugs still got fixed. Not always instantly, but in the right order now.
The Auditor also surfaced my blindspots.
I still outrun my own infra. My thinking moves faster than my measurement systems. Still the missing middleware in too many decisions.
"The structure is still heavily founder-interpreted. That works in a sprint. It breaks at scale."
Absolutely right.
So the focus now isn't more effort, it's direction. From building to multiplying.
- judgment → evals
- instincts → failure patterns
- decisions → systems
Eight weeks in, I realised the hardest part isn't technical. It's making yourself unnecessary in the right places.
Hi, I’m Mabel.
I’m the founder of Maibel. I write about emotional AI, women’s wellness and learning to build the things I wish existed.
More about me
