Welcome to Slow Ventures’ Snailmail, where we slow down and share what’s been on our partners’ and founders’ minds this week.
TL;DR:
Others:
PortCo Activity: Coco Robotics launched Coco 2 Delivery Robot (Slow III, Opp I), Blackbird partnered with MetaMask for exclusive crypto card for dining program (Slow V), Parcl partnered with Polymarket for real estate prediction markets (Slow IV)
AI Won’t Seal a Permanent Underclass, But It’s Raising the Bar on the American Dream
Sam Lessin
Context:
This riff started in response to Jack Altman’s question about whether AI is going to create a “permanent underclass.” My take: AI isn’t going to create a 100% sealed-off, permanent underclass. But it is going to fundamentally change the calculus for upward mobility in America and not in the way most people are ready for.
Market Signal:
Historically, if you were in the 50th to 75th percentile of willpower, intelligence, and luck, you could reasonably expect to find real upward mobility. The system needed a lot of solid, average-to-above-average folks to scale opportunity. “Have a shovel, will work” was enough in the 19th century. “Smart enough, will go to school” was enough for most of the 20th. There was a built-in safety net for the able and committed.
AI blows that up. Now, thanks to power-law returns and frictionless scaling, we need fewer “medium-good” people to create, build, and scale things. The winners are bigger, but fewer people ride along with them. The bar for upward mobility shifts from p50/p70 to p90/p95. Upward mobility isn’t gone, but it’s way less accessible for the merely competent and committed.
Takeaways:
AI doesn’t create a fully locked underclass, but it does make upward mobility rare—no longer a mass-market, retail product.
The old “guarantees” that powered American optimism (work hard, be a bit above average, get ahead) are breaking.
This breeds a new kind of fear: not that you can’t win, but that you can’t even try unless you’re already exceptional.
The core American promise, that average people with grit can move up, is at risk.
Shipping Fast Isn’t a Moat Anymore (And Maybe It Never Was)
Yoni Rechtman
Context:
After writing about the SaaSpocalypse and the headwinds facing software defensibility, I got some pushback (h/t Comron) that’s worth unpacking: If you could outsource software to India for cheap for 20+ years, and code was already a commodity, why did SaaS ever have a moat? If all vertical SaaS is “just Salesforce with extra fields,” what exactly did LLMs kill?
Market Signal:
The market is finally internalizing something we’ve all quietly known: code is not a moat. It never was. The bottleneck was always execution: finding people who could do it well, quickly, and at scale. Outsourcing was a blunt attempt to commoditize code, but it never worked at the level people feared. AI code generation, on the other hand, is a genuine shift—both in kind (it multiplies the best, doesn’t burden them) and in degree (it’s orders of magnitude faster and cheaper than the old offshoring playbook).
Takeaways:
Code = Execution, Not Moat: Writing software was never a moat, just a high-friction execution bottleneck. AI relieves that bottleneck, but doesn’t change the nature of true moats (network effects, data, brand, economies of scale).
Moats vs. Execution: Moats set your ceiling; execution determines how close you get. They’re independent. Confusing the two is why so many SaaS businesses are in trouble right now.
Everything Shifts Down a Tier: Application software gets easier, so the hard problem moves up the stack. Pure app software companies start to look like services businesses: easy to start, hard to differentiate, tons of competition. For everyone else, this is a gift, resources can go to the real differentiator.
AI = The Great Leveler: More people can now execute at a “good enough” level. The value of code as a differentiator compresses at both ends: it’s free/cheap at the low end, and a force multiplier at the high end.
The Only Four Jobs Left: Product/vibe/PM/slop cannon, security/SRE/infra, “hot people” (in every sense), and grown-ups (legal/finance/ops). The latent traits always cut across job titles, but the stack is compressing.
Dear VCs: Deep Tech Isn’t Your Get-Out-of-Software-Free Card
Sam Lessin
Context:
There’s a new refrain echoing through VC land—“Software is uninvestable, so now I’m focusing on deep tech.” Give me a break, Mr. Software VC. Suddenly everyone’s a quantum physicist, a biohacker, a rocket scientist. It’s the meme du jour.
Market Signal:
What’s happening? The market’s gotten tough for software, so the same folks who once waxed poetic about SaaS multiples are now chasing “deep tech” like it’s the last lifeboat off the Titanic. But here’s the rub: Deep tech is hard. The learning curve is vertical, the business models are even trickier, and the talent pools are fragmented. It’s not like you can just read a few Substacks and suddenly have taste in quantum, fusion, or CRISPR.
Takeaways:
Deep Tech is Not for Dabblers: You probably need to be a lot smarter than you think to have a real edge here. Taste doesn’t come from ChatGPT-ing your way into a new sector. If you’re just jumping into deep tech because it’s the new narrative, you’re probably about to get smoked by people who’ve been living this for decades.
Where’s Your Edge? If everyone’s making the same pivot, what makes you special? You’re not going to out-GPT the next guy. You’re not going to outwork the next gal. And unless you have a true, embedded advantage—networks, expertise, proprietary insight—you’re just another tourist.
Option A: Prepare to run on a treadmill that only speeds up. Your only edge is working harder. Is that really why you got into VC? To run races forever?
Option B: Specialize and say goodbye to the breadth that makes VC fun and valuable. You can’t cover all of deep tech. You’ll have to pick a lane (space, bio, defense, whatever), and suddenly you’re not a mile wide and an inch deep—you’re half a mile wide and three inches deep. No one likes that.
Asks:
So here’s my advice: Don’t be a lazy pants. Don’t default to the obvious narrative (“no more software!”) and couple it with the obvious pivot (“deep tech!”). That was clever five years ago, before it was the meme. Now? It’s just lazy sauce.
There are plenty of other solutions. We’ve never been more active at Slow. There’s tons to do around cultural shifts, new return-on-scale monopolies, and smart, non-obvious bets. Don’t just chase the meme. Do the work. Find your edge.
More Musing From The Team