When AI first started really cropping up a few years ago, I'll be honest, I was worried about what it meant for UVstocks. The work we do here is essentially using disciplined human judgment to comb through public data and find value that the crowd is missing. So my thinking at the time was something like this. If a model can now do all of that for free, then where does that leave us? That was the fear that sat with me for a while.
It took me a bit to come around to the other side of it.
A year ago we were sort of stuck. We had started putting together an investor pitch deck while trying to bootstrap as much proof of concept work as we could on our own. One of the questions I kept asking the mentors and advisors I'd been talking to was basically how much product maturity we needed to show before we'd be ready to walk into a room with investors. These were folks who'd been through capital raises and folks generally seasoned around startups. The honest answer was always pretty much the same. It was probably more than a very small team like ours could build in any reasonable amount of time. So we found ourselves in a kind of catch-22, where we'd need capital to hire engineers to get to the maturity bar that would have attracted the capital in the first place.
Then over the past few months, something remarkable happened. Claude Code advanced virtually overnight. I don't really know what happened over there internally, but the version we work in suddenly leapfrogged ChatGPT's Codex and a couple of other coding tools we'd been watching. And just like that, the math on our roadmap changed. Production retrofits we'd been kicking down the road for almost two years are now back in the pipeline. We've started running parallel AI-agent led work across functions. Sales, marketing, operations, R&D, you name it. The Russell 2000 Beta that I'd been promising "is coming soon" for way too long finally went out to early subscribers last week. And the Convergence on Sell list we talked about in last Sunday's note? That whole thing came together in just a few days.
When I think about the economics of it all, it's almost still hard to believe. The two Claude Max accounts we use cost about $200 a month total. And that's basically the entirety of the "engineering team" we'd been told we'd need to raise outside capital to hire. There's no onboarding. No ramp-up. No rework cycles. No vacation calendar to plan around. None of the coordination overhead that comes with building out any team, even a good one, really applies here.
So why try to raise outside capital to fund an engineering team we no longer need to hire?
Looking back at it now, what I was wrong about was how all of this was going to play out for us. AI hasn't replaced the disciplined human judgment that actually goes into our work. The screens, the rubrics, the editorial calls, the conviction we have on any particular name, all of that is still us. What AI has done is take away the friction between thinking "hey we should build that" and actually having a version of it in production a week or two later. We get to bootstrap a lot longer than we'd planned to, and on our own terms.
There's one more thing I missed in my original worry. They say data is the new oil. And the multi-source data we've been collecting and curating for our subscribers for a while now is something an AI model can reason over, but can't go back and actually manufacture from thin air. Our point-in-time archive (which is the underpinning of things like CSR and CoS, and probably more signals to come as we use AI to help surface them) is the kind of thing you have to start building well before you urgently need it. The good news is that we started building ours before AI was really part of the public conversation. And honestly, that's the part I'm probably most excited about as we look ahead.
It's not something we take for granted. But the more useful thing for this audience is probably to think about what all of this means for the businesses you own and the companies you work for. The same compression we've been experiencing is showing up at every small company on your watchlist. And probably in conversations happening at the company that signs your paycheck too. Engineering payroll lines that used to be a dominant operating cost are starting to turn into a software line item. Time to market for products that used to need a full team is collapsing to what feels like a long weekend in some cases. If you're screening for value the way we do, this is going to start showing up in the cash flow statements you read. The companies that absorb the savings first tend to hold those costs as margin for a while, until the competition catches up and pulls the prices back down.
There's a harder question that keeps creeping in for me, though. Hyperscalers are on track to spend over $700 billion on AI capacity this year. Demand is actually outpacing supply right now, so that capacity is getting absorbed. The part that bothers me is who's getting hired on the other side. Employment for software developers ages 22 to 25 is reportedly down close to 20% since 2024, concentrated in the routine coding work that AI now handles well. The engineer we ended up not needing to hire is the same kind of role getting squeezed out across the economy. New AI job categories are appearing, but they're going to a different cohort at what feels like the wrong pace.
That's a story still unfolding. We'll be looking out for the names where the data shows it landing first. More on this to come.
"You cannot endow even the best machine with initiative; the jolliest steamroller will not plant flowers."— Walter Lippmann
Not sure where to start or know someone that could use the help? Pass this article along and the UVstocks.io link to a friend or colleague who's thinking about what AI is doing to the businesses they own.
P.S. I'm sharing some investment information, but it's important to remember that what I'm providing is for informational purposes only and should not be construed as financial advice.
Happy Investing,
John
If someone shared this newsletter with you, you can subscribe for yourself here.
