|
I've never really shared this publicly, but AI ruined my business.
Many of you reading this know Parsity as a coding bootcamp. From 2017 through 2024, that's exactly what it was. When generative AI hit, I assumed we'd keep going, add more AI material to the curriculum, and keep helping people get hired.
That's not how it worked.
The slow-motion car crash
At our peak we were making multiple six figures a year. We were one of the most affordable coding bootcamps in the United States and still turned a healthy profit while offering (in my totally unbiased opinion) a better program.
Joining was an easy decision.
Then a few things happened, roughly in this order:
- AI showed up and junior jobs got a lot harder to land.
- Our 92% placement rate started to dip.
- Our loan partner decided to stop working with any and all coding bootcamps. They told us we were one of the best schools they worked with. So sad... too bad.
Put those together and the bootcamp side of Parsity was done.
Call me an optimist, call me naive, but I still think bootcamps have a place.
Or at least, they should.
Most adults trying to learn to code on their own with ChatGPT are not going to get the result they want. I'm seeing a flood of people who can generate code but will almost certainly bomb a technical interview, and that makes me nervous for their future.
But opinions don't pay the bills. As a business owner, I had to pivot, and fast.
The pivot
I tried a few versions of what Parsity could be before landing on building AI systems. That decision came straight from my own career. Learning to build RAG pipelines and agents at two AI startups opened up more opportunities for me than anything else I've done as an engineer, and it wasn't close.
So we built a program strictly about building AI systems, for engineers who already know how to code. I also gave it to existing Parsity students at no extra charge so I could get their feedback and make it better.
This year, that bet is starting to pay off.
Where I think the leverage is
After building this stuff, talking with a lot of developers, and watching what companies are hiring for, three skills keep showing up. None of them should surprise you. (If they do, you haven't been looking very closely.)
From easiest to hardest:
Agentic engineering is the easiest to learn and has the least leverage. This is using tools like Claude Code and Codex to ship faster. More and more developers will pick it up naturally, the same way everyone eventually learned git. Worth doing, but it won't set you apart for long.
Building agents is medium difficulty. There are plenty of courses and frameworks, and they're all racing to make this easier. It reminds me of AWS serverless around 2015: a little mysterious at first, and now it's standard, well documented, and pretty easy to set up. Agents will go the same way. For now there's still a lot of leverage here, because most people never go past the surface.
RAG and vector databases is the deep end, and I think it's where you can stand out the most. Anybody can write a SQL query. Very few people can tell you why that query takes nine seconds and how to get it under 100 milliseconds. Retrieval is going to work like that. Getting a demo to return something is easy. Getting the right results, fast, over messy data, at scale, is a specialty, and I think it turns into a new flavor of data engineering.
The other bet
I've also stumbled onto an opportunity I didn't see coming: working with founders who have clients but not the technical expertise.
I used to laugh at "vibe coding cleanup specialist" as a job title.
I'm not laughing anymore.
There are a lot of founders who built something themselves, held together with duct tape, regex, and a deployment setup nobody else can explain. What they need is basically a CTO to come in, modernize the stack, and make it work. I think that market is a lot bigger than I assumed, and I want Parsity to be part of it.
It also fits into a nice little flywheel: we teach engineers these skills, then partner with businesses that need people who have them.
TBD.
This year Parsity will make a little under $100,000. My goal for 2027 is somewhere between $400,000 and $500,000, which is what I'd consider a healthy company. Between the AI program, agentic engineering, and working with founders, I think we can get there.
If AI ruined your thing too
AI destroyed the business I had, and it also opened up a ton of new ones.
So if you feel like AI ruined your career, that's fair.
It might have. ngl. no cap.
And there might also be a pile of new opportunities staring you in the face that you're not taking.
Right now is a weird window. There are no rules yet, and the bar to be considered an expert is really low. Learn the basics well and you're already in a small pool of people.
The problem is that most people take the path of least resistance. If you're employed, not terminally online like me, and not looking for a job, you probably don't care to learn any of this. That's 90-something percent of software developers: employed, coasting, chilling (or super anxious more likely).
This is exactly why now, while you're not desperate, is the best time to learn. Why wait until you get laid off or go job hunting, only to find out the skills you've built are close to useless? Being a Ruby expert or a React expert in 2027 is a losing game IMO.
If you want help getting ahead of this, that's what we do at parsity.io. Or just hit reply and tell me what you're working on.
|
