We’ve homeschooled for 9 years and are now at the point of thinking more career/post high school ed… and honestly the idea of getting my kids to just ID a market need, fill it with excellence, and bypass traditional ed is so appealing. Especially knowing I can help them, they won’t be yeeted from the house at 18, and them starting with a strong foundation and no debt would be incredible.
You’re missing the point. The point is that this stadiums not only exist, they’re quite abundant. My hometown in southern Indiana (pop 2500) has a big stadium and is still well known because of all the people who watch from the back road outside the fence because of how big a deal local sports are.
And 25 years ago, when I played, I visited 12-13 other small towns and not a single one had a dump of a field/stadium.
Just a few days ago I was thinking about starting a business doing this. The deeper down the AI rabbit hole I go, the more obvious it is that the vast majority of everyday folks are completely unawares of how powerful and ubiquitous these tools have become. And the crazy part is to do a job like this you really only need a few core skills and a bit of a willingness to walk into the unknown.
I’m a production engineer in oil & gas.
Not a developer. Not in tech.
Earlier this year,I started building an AI system for my actual job.
I started with virtually zero AI LLM experience.
Today I read an article naming what I’d intuitively built.
Here’s what I discovered before anyone told me it had a name. 🧵
The article says harness engineering “materialized in 90 days” starting February 2026.
I started building on March 16th. 84 days ago.
Not because I read about it. Because I kept losing context on real well decisions and got tired of it.
I was 6 days behind the people who named it. In oil & gas. Alone.
Domain pressure is a better teacher than any framework.
What’s your harness look like?
I’m a production engineer in oil & gas.
Not a developer. Not in tech.
Earlier this year,I started building an AI system for my actual job.
I started with virtually zero AI LLM experience.
Today I read an article naming what I’d intuitively built.
Here’s what I discovered before anyone told me it had a name. 🧵
Oh, and while I was building IRIS at work —
I also shipped my first iOS app.
And my first Android app.
In the same 12 weeks.
Same harness principles. Different domain. No coding background.
This isn’t a methodology for AI researchers.
It’s what happens when a domain expert stops waiting for someone to build the right tool and just builds it.
https://t.co/p7lD464GWB
@sairahul1 I’m a production engineer in oil & gas.
I read this and realized I’ve been doing this for 84 days in a field nobody was watching.
Built it alone. Discovered it from domain pressure, not from a framework.
Thread incoming.
At my company, we have an integrated GPT that allows us to create custom models starting from a base of GPT or Opus.
They team encourages me to keep trying gpt due to cost eff, and I do keep trying but it’s just so awful and off-putting. And the actual intuitive comprehension about very technical results just isn’t there with gpt. Claude is awesome.
Not perfect and can get so amazingly tunnel-visioned, but it’s “knowable” and I’m able to actually learn its personality and adjust.
And also, taking the lives of soldiers, who almost never have the choice to be anywhere except fighting where they are, isn’t to be taken lightly.
We sacrificed that many of our own, but also took on the soul-dragging burden of taking millions of lives to achieve these goals. Many were innocent.
Dang I have an intuition for this. Been working on a transformation of my job, which is now leading to a transformation of how everyone does the job I do, and I have worked my way into patterns that match this article probably 90%. Certainly is good validation I’m on the right track.