Previously: Why “guess culture”?
Guess culture in management
I don’t like asking for things. I want my employees to figure out what I want, and make it happen without me having to tell them.
Even better is if they can figure out what I should want, if I were wiser. Or, what our org should want — understand its coherent extrapolated volition and implement it.
Guess culture at work goes both ways. As a leader, a lot of what I want do is try to build good model for each person I work with. Where they’re strong and where they need to improve; what they’re excited for, where they want their career to go.
(The best manager I’ve ever had was Jing Shan. My first day at Google, she asked me what I wanted to do and I said “work on frontend”. Unfortunately, she ran the backend team of what would become Hangouts Chat, and we were in launch go-go-go mode. But she remembered. 12mo and one promotion later, she found a chance to put me on a full-stack initiative.)
Sometimes, people I work with tell me that they’re surprised at how quickly I extend trust, responsibility and power; eg handing off my babies (Manifest, Ada) to people I barely know, at first. Partly, I accept this as a compliment; partly, I just don’t know of another way to manage.
“Founder Mode” is a recent-ish meme among SF startups, extolling the virtues of managing hands-on, Steve Jobs-style. But I notice that there’s a countervailing meme in Anthropic-land, where Dario gives biweekly all-hands to help everyone build “shoulder advisor” model of him — and by most metrics, Anthropic has done better than AirBnB.
Guess culture in community
To be fair, my “management problems” are kind of tiny. Across all of Manifund + Mox there is ~6 FTE right now. It’s not clear that advice for AirBnB or Anthropic-scale orgs applies to me.
But: many of my projects are community-shaped, like Manifund and Mox and Manifest and Surplus. Managing a 200 person coworking space or a 900 person conference have a bit more in common with running a medium-to-large org, than most eg software startups.
Besides me personally having a preference for guess culture, I think guess culture might be important for scaling up communities. I often think about how the biggest bottleneck in EA & AI safety community will be trust, as waves of money are unleashed, salaries spike, all kinds of people try to join, and the space enters hypergrowth.
Asks like “please give me money” or “please hire me” become quite costly, requiring trust that doesn’t yet exist. Instead, I encourage newcomers to learn to guess what funders or employers want to see from them, and do those things.
Returning to Anthropic as a case study in community. Why is Anthropic winning? Obvious answers might be “hire really smart people” or “take AI safety seriously” or “raise a lot of money” or “be really scaling pilled”. But: OpenAI has smart people, SSI is all about safety, everyone has boatloads of cash and the cat’s out of the bag on scaling. I suspect it’s something about internal culture, which bears fruit in insanely good retention and hiring choices and sustained product quality and shipping velocity. Something about those Dario shoulder advisors (and no doubt, many other personas) work to facilitate trust & alignment at scale.
Guess culture in product
Speaking of product quality, one thesis I have is that being a really good “guess”-er is key to designing good consumer software.
YC advises that you talk to users. But talking to users can be expensive — you have to find them, convince them you’re worth talking to, line up time to chat. If instead, you can just model them inside your own head, you can converse with them any time for free.
In fact, I think that the point of talking to users is to build a mental model of them. So that beyond getting answers to your immediate product questions, you can also just design things correctly oneshot, just by querying that mental model. The point is to know them better than they can articulate, ideally better than they know themselves; otherwise you end up making a “faster horse”.
Guess culture in religion
At this point in my life I’ve listened to hundreds of talks, but one concept that’s stuck with me is from a talk by Emmett Shear (in a series hosted by Misha Glouberman)
Guess culture in writing
Reading lots of longform can be an exercise in developing models of people. I have “shoulder advisor” models of eg Paul Graham and Scott Alexander and Patrick McKenzie, built across years of following their writing, long before I ever got to meet any of them.
I’ve now had the privilege of meeting some of the writers I look up to in person, and collaborate on things together. But I think this was partly possible
When James, Stephen and I where applying for what would become Manifold, we scoped the proposal to exactly fit Scott Alexander’s tastes. This was only possible because all 3 of us had spent years reading Scott’s writing and thereby building a very good model of him. And (I think) Scott endorses this, as ACX has gotten some value from the existence of Manifold and then Manifest & Manifund afterwards.
FWIW: our original application:
Motivating example: A Substack writer could create a prediction market for “Will the FDA approve X drug by Jun 10 2022?”
Also: from the writer’s side, the advice of “writing for a single person” seems to be an exercise in practicing guess culture?
Misc, other interesting reading on guessing and asking
“Instead of asking for things, guess what people want”