$ whoami

sam lessin

founding partner at slow ventures. intern-writing at the information. host of more or less. trying to fix harvard. previously vp product at facebook; founded drop.io and fin.

// links

// social

// positions

capital-light, one-and-done seed

raise exactly the round you need to reach a real product, then scale on your own profit. later financing is fine, but by choice from a place of profitability, not forced onto the industrial vc-factory financing line that ran 2010–2023.

creator-led businesses

build community, brand, and trust — then layer products on top. creators can be serious entrepreneurs, and the best of them are massively under-financed in the early days. the future of meaningful brands looks more like this than like the old dtc / saas playbooks.

customers replacing vendors

saas is dead, not just pressured. anyone even a little technical can now build their own tools — i ripped out my zapier stack last month. the real ai disruption isn't new startups replacing old vendors; it's end users skipping the vendor tier entirely. five times cash flow, not 20x revenue.

crypto, still

over the years i've gotten crypto more right than most — 2000x on solana from seed, early to bitcoin personally (2013), seed checks into ripple, aleo, algorand, many more. the ecosystem's fundamental importance hasn't wavered. the right opportunities to seed change over time; the conviction doesn't.

sub-10-post deals

i don't pay up for seed, pretty much full stop. high-priced seed means too-consensus takes — lots of people willing to invest with little differentiation. i want capital going where it's most needed, most in demand, most expensive. first-check bets i'm most proud of in this bucket: venmo, birchbox, teamshares.

american entrepreneurship, local edition

periods of technological change come and go; what makes america great is the insanely strong culture of entrepreneurship everywhere in the country. last decade everyone wanted to be the next mark / elon / sergey. the next generation is building real, profitable, locally-meaningful businesses — not hail-marys for global 'monopoly' platforms.

ai as a startup category

ai is extraordinary technology and a terrible business. foundation models commoditize toward zero, and the incumbents — google, meta, apple, microsoft — capture the upside across a hundred surfaces overnight. what's funding 'ai startups' right now is narrative, not unit economics: a triangle trade between washington, silicon valley, and wall street where everyone needs the gdp number to go up.

agent unit economics

i ran fin. error rates are what kill agent margins, and they don't disappear at scale — they compound. 'ai will do X for you' businesses priced on full automation mostly get caught by the long tail of cases where the agent is almost right and therefore expensive. the math doesn't work the way the pitch deck implies.

seed for the 'industrial-vc' pipeline

that game is over. please don't pitch me on how your seed financing gets you to metrics some series-a investor will value for the next money-losing round. the factory line isn't coming back.

expensive rounds

money is a commodity; i like to sell it for high prices, because that means it's going where it's scarce and valuable. if rounds are expensive and money is cheap, there are better sources of capital. the capitalism i'm proud of is money into people and things where the capital is very very expensive because few else 'gets it' yet and the money actually matters.

double bottom lines

bad idea. if you can't reduce to a bottom line you don't know how to trade your goals against each other. capitalism works because it forces reduction to 'profit'. the second you debate how much A you'll trade for how much B, you're screwed.

even co-founders

even when they're best friends to start, this basically never works. great startups need someone ultimately fully responsible and in charge; splitting that across multiple people produces sub-optimal outcomes. great 'junior' co-founders you recruit, fine. 50/50 (or 33/33/33) founding teams, no.