Welcome to Slow Ventures’ Snailmail, where we slow down and share what’s been on our partners’ and founders’ minds.
TL;DR:
Others:
AI Is Not The Opportunity. The Economic Disruption AI Creates Is The Opportunity.
Will Quist
Context
The way I think about AI starts with a pretty simple question: Is there some rip in the universe that’s going to change how economics flow?
With AI, the answer is obviously yes.
The mistake people make is stopping there. Just because disruption exists doesn’t mean you get to capture any of the value. The real question is what gives you a disproportionate claim on the upside and whether it costs less to build that claim than it’s ultimately worth.
More broadly, this is how I think about venture. I’m not a thematic investor. I don’t wake up every morning looking for the next AI company or the next healthcare company. My job is to have a framework for evaluating novel ideas and figuring out whether they fit venture capital.
What we’re ultimately looking for are hypotheses that are genuinely non-consensus, can be proven true or false, can be tested with a reasonable amount of capital, and become objectively valuable if they’re right.
Market Signal
I think people are overstating how transformative AI is for software economics.
That doesn’t mean AI isn’t important. I spend all day in Claude and ChatGPT. I think it’s one of the most important technology shifts we’ll see.
But if you zoom out, software has been getting cheaper to build and distribute for decades. I’d argue the move from client-server software to SaaS was actually more disruptive than the move from SaaS to AI.
The reality is most software businesses were never protected by code alone. If they were, SaaS would have been an incredible free cash flow business from day one. Instead, most of the economics flowed into sales and marketing because competing products could always be built.
AI lowers the cost of building software even further, but it doesn’t magically solve distribution, customer acquisition, workflow adoption, or organizational change management.
That’s why I’m generally skeptical when I hear AI pitches that are purely about efficiency gains. Those gains are real, but they’re often arbitrages rather than businesses.
The first person to discover an efficiency captures the spread. Eventually everyone else gets access to the same tools and the spread gets competed away.
The question I’m always asking is whether the leverage is proprietary or commodity.
Takeaways
AI is not the opportunity. The economic disruption AI creates is the opportunity. The fact that AI is changing industries is obvious. The harder question is why you deserve to capture an outsized share of the value being created.
I’m looking for novel hypotheses, not consensus markets. The best venture investments aren’t necessarily contrarian. They’re ideas the market isn’t incentivized to think about yet, where a clear experiment can prove something valuable before everyone else catches up.
Most AI-driven efficiency gains will get priced in. If you’re simply applying off-the-shelf AI tools to make a business more efficient, that’s useful. It’s just difficult to build durable enterprise value around something everyone else can do.
Workflow adoption matters more than most people realize. Building the technology is often the easy part. Getting organizations to change behavior, adopt new workflows, and realize actual ROI is usually much harder.
Valuation should be treated like leverage. Any time you’re getting farther on your equity than you otherwise should, you’re using leverage. That’s true whether we’re talking about debt or valuation. Leverage is great when you’re right. It cuts just as hard when you’re wrong.
Not every software innovation should become a software company.
One thing founders consistently miss is that just because the innovation is software doesn’t mean the best economic outcome comes from selling software. Sometimes the better answer is vertical integration, ownership of operations, or some other structure that captures more of the economics you’re creating.
The AI cycle will probably look like every other cycle. I don’t know when it happens, but I’d be surprised if this ends up being the first major technology wave that doesn’t eventually experience some form of boom-and-bust behavior.
The only real question is timing.
The thing I’m always trying to separate is whether someone has discovered an arbitrage or built a business.
An arbitrage is something you spot. A business is something you create and control.
The most valuable companies tend to be the ones that own the mechanism that generates the economics, not just the ones that happen to benefit from it for a short period of time
AI’s Illegibility Is The Moat
Sam Lessin
Context
The conversation around AI has shifted. Last year’s narrative was simple: AGI is coming, the machines are taking over, and you’d better get on board. Fair enough. If there’s even a small chance that’s true, you have to pay attention.
This year’s question is much more practical: What is all this token spend actually worth?
Companies are now spending real money on AI. Not experimentation money. Operating expense money. And once you’re shoveling enough dollars into the furnace, you eventually have to ask whether the output justifies the input.
The challenge is that most people still don’t have a framework for thinking about intelligence as an economic resource.
Market Signal
The easiest way to think about models today is not as software, but as labor.
Imagine the frontier models are PhDs with IQs of 140 charging $200 per hour. The smaller open-source models are competent high school teachers charging $20 per hour.
Now imagine you have a task. Maybe it’s tax preparation. Maybe it’s summarizing documents. Maybe it’s categorizing thousands of transactions. Maybe it’s folding an enormous pile of information into something usable.
The question isn’t “Which model is smarter?”
The question is: How expensive is it to be wrong?
If mistakes are catastrophic, you pay for intelligence. If incremental correctness creates enormous economic value, you pay for intelligence.
But if the downside of being wrong is limited and the upside of being slightly more right is asymptotic, paying 10x more for intelligence makes very little sense.
Most businesses haven’t learned to think this way yet.
Takeaways
Not all intelligence is created equal, and not all tasks deserve the same intelligence.
We’re still operating under the assumption that every problem deserves the smartest model available. That’s rarely true. Most tasks don’t need a PhD. They need something good enough.
The real optimization problem isn’t model quality. It’s matching intelligence to economics. The highest ROI organizations won’t necessarily use the best models. They’ll become exceptionally good at deciding where premium intelligence matters and where it doesn’t.
Verification is becoming the hidden cost center. The interesting wrinkle is that in many cases, verifying the answer costs almost as much as generating it.
This shows up everywhere: News, scientific research, financial analysis, enterprise workflows. The cost of checking the work is increasingly rivaling the cost of producing the work itself.
Opacity is supporting AI pricing today. Most buyers don’t actually know where the intelligence curve bends. They don’t know when a $200 solution materially outperforms a $20 solution. When uncertainty is high, people default to buying the expensive option. That uncertainty is part of what’s supporting current AI economics.
The illegibility is the moat. Right now, nobody has great intuition for:
What “good enough” looks like, Which tasks justify frontier models, Where incremental intelligence creates incremental value, or When verification matters more than generation. Until those frameworks emerge, buyers will continue overpaying for certainty.
The future may not belong to the company with the smartest model.It may belong to the company that best understands when intelligence is worth paying for at all.
That’s a much messier question.
And right now, the fact that nobody can answer it confidently is exactly what’s holding up the entire market.
More Musing From The Team
