Venture investing is unusually time-dependent. Early-stage judgment depends on live context: who is building now, what has recently become possible, and what excellence looks like at inception. It is less like accumulated wisdom and more like an athlete’s motion, learned through repetition, strengthened by use, quietly lost through disuse, and constantly evolving as the global standard of play changes. That judgment only stays sharp through sustained exposure. Being in market—meeting founders, seeing incomplete ideas, and engaging with new technical primitives—is not a sourcing tactic; it is the work.
The “time = 0” problem is especially acu…
Just published an essay, just edited a friend's essay this weekend. Realized I am hypersensitive to rhetorical crutches now. It all reads to me like AI slop, even when I see it hand-written by a human
1/ new essay: AI Adoption is a coordination problem.
People adapt to AI faster than organizations do.
That gap explains most of the confusion about “slow” AI adoption.
2/ Individually, adoption is fast. People use AI to write, learn, plan, and build because it makes their work better. Once you’ve worked with better tools, you don’t go back.
3/ At work, the calculus changes. The unspoken worry isn’t that AI won’t work—it’s that it will, and the people who make it work first (or even provide "human data" and "traces") won’t be the ones who benefit.
4/ Enterprises slow things down because real change is costly to coordinate, collides with en…
1/ One productivity question stuck in my head: when AI can draft, explain, and propose solutions on demand (across a huge terrain!), what still has to live in your head for that to turn into real work, rather than a pile of plausible slop?
2/ Now you can wander into a field you barely know and still coax out ideas that feel specific, original, and coherent. But the moment the idea appears, you run into the tough bit: is it true, is it actually new, what’s it assuming, is it meaningful?
3/ Most obvious in software. You can vibe code a LOT (until state, edge cases, security, performance, integration, or debugging show up). And then there's a …
it's end of year planning! open roles at our small team at
@conviction:
- engineering
- general creatives (experience, storytelling, design)
- investing
- always happy to have amazing operators camp with us
referrers may get to name my next child
short of that, can enjoy a vacation home :)
Field reports > feed reports. A lot of AI “analysis” today is just the same circulating threads from people with zero primary access... but a lot of confidence. The result is a market-wide group chat that ends up driving the narrative. Careful!
that narrative would have led you to believe in '23-'25
- OSS foundation models will never matter
- no value in the AI app layer
- scaling pre-training is all there is
- scaling pre-training is at an end
There's a particular kind of cleverness that gets celebrated in startup circles: the ability to see through everything. To identify the "real" motivation behind every action. To reduce human behavior to a simple calculus of self-interest. This usually gets called being "clear-eyed about incentives." Often it's just cynicism wearing a Patagonia vest.
Understanding incentive structures genuinely matters. If you're building a company or investing in one, you need to think carefully about why people do what they do. You need to design compensation that rewards the outcomes you want. You need to anticipate how partnerships might evolve when inter…
my kid learned to read with this thing. also, loves that purple dancing cow. seemingly not concerned about the cow’s attitudes toward tech or misogyny.
complete failure that education sector is lagging this badly in AI. the result of this lack of adoption is that we get many of the AI abuses folks are afraid of (eg cheating) but miss the benefits (eg personalized tutoring) https://t.co/4MqKJTwn2O
The fact that there are at least three Dan Rob’s very good at math in the tech ecosystem seems like a collision risk: @danrobinson @danlovesproofs @danintheory