I'm a computer science student at the University of Guelph. Over the last three years I worked co-ops at three companies that could not be more different from each other:
CGI
One of the largest IT consultancies in the world. I worked on enterprise systems for clients you'd recognize.
Scotiabank
A Canadian top-five bank. Heavily regulated, heavily process-driven, three layers of approval for everything.
Bombardier
Trains and business jets. The kind of company where a single document can take a year to update because lives depend on it.
Three different industries. Three different cultures. One pattern, unmistakable in every single one:
The company runs on knowledge that lives in people's heads, in old emails, in Slack threads from two years ago, in Confluence pages last edited by someone who left in 2023.
Onboarding a new engineer at any of these companies took months, not because the work was hard, but because the documented playbook was thirty percent of what you actually needed to know. The other seventy percent was “ask Sarah,” or “there's a thread from last March, let me find it,” or “here's a guy I should introduce you to.”
That's the operating system of every real company. Tribal knowledge. It works because humans are good at vague pattern recognition over years. It breaks the moment you try to put an AI agent into the same workflow.
The wall
In early 2026, I started using Claude Code seriously. Like, every day. The agent is brilliant. Drop it into a fresh project and it's a senior engineer. Drop it into a real, ongoing, opinionated project and it's — well, a brilliant intern who needs context.
I'd ask it to do something against an internal-feeling workflow and it would give me a generic, textbook answer because it didn't know our refund matrix, our discount ladder, ourincident runbook. The model wasn't the bottleneck. The knowledge was.
I started writing Claude Skills by hand for my own projects. They worked instantly. The agent went from “intern” to “senior who knows our codebase.” But each skill took me hours to compile from scattered notes, threads, and old PRs. I was the compiler.
That's when it clicked. The compiler is the product.
Why solo, why now
I'm applying to YC alone. I've been deliberate about it. The right co-founder shows up after months of shipping, not at the application deadline. I'd rather YC see a product I built alone than a team I assembled for the application.
I built the live Skillforge MVP in five days, by myself. Real ingestion pipeline. Real Claude Agent Skills generator. Real streaming side-by-side demo. Real live compiler. The codebase is shipped, the demo is on, anyone can run it before reading my application.
That's the velocity I'll bring to a YC batch. No coordination overhead. One person owning every decision. Every day shipping.
Why this RFS
Tom Blomfield put the “Company Brain” problem on YC's Summer 2026 RFS. He's right that every company in the world is going to need one. He's right that the model isn't the bottleneck anymore.
And I'm not just going to build it because YC asked. I was building it before I read the RFS. The RFS is signal that the market is here. The product is what I was already going to ship.
What you can do
If you're a YC partner reading this: please run the demo before reading my application. The product is the strongest part of my pitch.
If you're a builder hitting the same wall: get on the waitlist. We're onboarding the first design partners now and I'd love to talk to you.
If you're someone who'd be a great engineering hire: email me. We'll need help fast.
— Ahmed Tariq, Guelph, Ontario · April 2026
ahmedtariqcs@gmail.com