I Didn't Expect This When I Sat Down to Watch YouTube With Lunch
Steve Morris is a legend. A true craftsman — the kind of subject matter expert that people expect AI to be. He builds billet blocks that make over 5,000 horsepower. Tolerances measured in hundredths of thousandths. He's been doing this for decades. By any honest measure, one of the best engine builders alive.
You know that "dad thing" some guys have where they can just diagnose shit? Oh, that sounds like it's the lower left thingamabobulator — and the MFer will be right. That's Steve with a big block Chevy. The kind of instinct that just demands respect.
In a recent video, Steve stops mid-build to talk about AI. He'd been seeing people walk into his shop — or message him online — with specs and clearances pulled from chatbots. Cam timing. Compression ratios. Bearing clearances. All generated by AI, all delivered with total confidence, and a disturbing amount of it just... wrong.
His quote:
"AI is basically just a giant aggregator of everyone else's opinions, and it doesn't know the difference between a guy who's built 1,000 engines and a guy who's just typed about it 1,000 times."
He compared it to someone who's read every book ever written about swimming but has never been in the water. They can describe the stroke perfectly. They'll still drown.
He's right.
Torque Wrench vs. Impact Hammer
Steve is describing the default use case. The person who opens a chatbot and asks "what should I run?" — no context, no constraints, no data. Just a question and a prayer.
That's not using AI as a tool. That's using it as a psychic hotline. And like Miss Cleo at 2am, it sounds incredibly confident about things it has no business being confident about. Don't call that 900 number.
The problem isn't the technology. It's how it's set up.
Imagine a different version of that same tool — one built with Steve's world in mind. Not a generic chatbot trained on internet forums. An LLM with a harness. Tested. Trained on verified data — actual material specs, actual dyno results, actual failure analysis. Connected to lookup tools that can cross-reference torque specs against manufacturer databases. Guardrails that flag when a clearance falls outside proven ranges instead of guessing.
That tool could answer the same questions Steve's customers are asking. Not because it "knows" engines, but because it's been given the right data, the right constraints, and the ability to verify before it speaks. And most importantly — the ability to say "I don't know" when it doesn't know.
A chatbot with none of that is Miss Cleo. A system built with all of it is a shop assistant who never forgets a spec sheet.
Why an Engine Builder Matters
I follow Steve Morris because he's the kind of person I respect: someone who builds things. Not because it's efficient. Not because the business case was strong. Because the block was sitting there and he wanted to see what it could do.
That's the same impulse I hope grows in more people — and the same impulse AI can help empower. Not replacing the struggle to build, but reducing the barriers that kept people from trying. If you have an idea but no starting point, AI can help you learn and get moving. The tool gets sharper as the knowledge behind it does — better data in, better answers out. That's not a shortcut. That's how every good tool works.
Steve and I agree on the fundamental point: expertise matters. Where we diverge is what happens when the expert picks up the tool. He's never seen that version. He's only seen the people who use AI to skip the work — who don't know what they don't know, and neither does the tool. That's a dangerous combination. But someone who already understands the domain, who knows which questions matter and which answers smell wrong, gets a fundamentally different experience from the same technology.
I think if Steve Morris ever sat down with a model trained on his own data — his failures, his tolerances, his decades of knowing which parts sing and which ones grenade — he'd stop calling it a shortcut and start calling it a shop assistant.
He wouldn't trust it to make the final call. He'd never let it touch the wrench. But he might let it sort the pile.
Thank you for your time. I hope you have an awesome day.
P.S. — Steve, if you ever want to talk about that shop assistant, hit me up.