Paul Graham Profile picture
Leo Profile picture KimSia Sim (I tweet business & software stuff) Profile picture Tai Profile picture Jacob Patton Profile picture Olivier Kitenge Yesu Profile picture 181 added to My Authors
Jun 19 4 tweets 1 min read
When I was younger I imagined AI would progress by generating correct responses of gradually increasing complexity. Now it looks as if it happens by generating complex responses of gradually increasing correctness. I.e. I imagined AI would start by talking about the blocks world and eventually progress to writing essays. Instead we have essays from day one, but bogus ones, and AI progresses by making them increasingly plausible.
Jun 15 4 tweets 1 min read
Hung out with Ron Conway for the first time since before Covid, and realized there's something about him that's even rarer than how hard he works for founders. He's modest. There are only about 5 modest people in Silicon Valley, and Ron leads this tiny pack. Ron works harder for portfolio companies than any other investor I know. He's also less arrogant than any other investor I know. So his work to arrogance ratio is orders of magnitude above the mean.
May 30 4 tweets 1 min read
People on Twitter are almost always angry about something. You get used to it if you use Twitter regularly, but if you only check in occasionally, the anger is surprising. This doesn't necessarily mean Twitter is broken. There's a lot to be angry about. Maybe Twitter's destiny is to be the channel for all the anger.
May 26 4 tweets 1 min read
It's an interesting data point that you can predict something will be ubiquitous when it still has only .4% of potential users, and yet you totally can. I remember this memo, and it was not seen as a risky prediction at the time. Insightful, for a big company, but not risky. In fact the spread between the size of the user base (still tiny) and the certainty of the prediction (very high) is part of the way people who understand technology make money.
Apr 16 7 tweets 2 min read
I read @yishan's thread about Twitter and agree, as I think anyone who's run a forum would, that Elon is "in for a world of pain," or at least for a type of pain both much nastier than hard engineering problems, and with far less upside as well. Where I think he's mistaken is his claim that the left and right both want to ban each other roughly equally. Among the elite, and within Twitter specifically, there is much more inclination to ban the right.
Apr 2 5 tweets 1 min read
Though startup investors are all looking for a high growth rate, growth rate is (whether they realize it or not) merely a proxy for something else: how big the company will ultimately get. Usually it's a proxy that works pretty well. Usually a company that never grows faster than say 2x a year also never gets super big. But obviously 2x could do it, if you kept growing that fast for long enough.
Mar 26 4 tweets 1 min read
I never read this till now, but it contains some important ideas. Particularly that zealous supporters of a cause are incentivized to rally around dubious examples of it, because rallying around obviously convincing examples demonstrates less commitment.

slatestarcodex.com/2014/12/17/the… I had noticed that religion requires adherents to espouse false beliefs. A group can't distinguish itself by believing true things, because if something is true, other people will believe it too. But I hadn't realized how this plays out in selecting causes célèbres.
Mar 25 4 tweets 1 min read
Jessica once cut a founder's nails before an investor meeting. If you want to understand how different YC is from a VC firm, try to envision a VC doing that. I understand why people call me a VC. From far enough away, it must look like that. But because startups change so fast, the people who deal with them at different stages of their life are actually very different.
Mar 21 5 tweets 1 min read
Suggested an idea to Brian Chesky and Joe Gebbia. Turned out they were already working on it. Just like during their YC batch. Airbnb isn't successful just because it's a valuable marketplace. It really helps that those particular founders are working on it. They're a perfect match for this idea, like Larry and Sergey were for Google, or the Steves for Apple.
Mar 10 5 tweets 1 min read
The reason "the west" cares so much about the invasion of Ukraine is not just that Ukrainians are white. It's also because this is happening in the subset of the world where the countries had decided they weren't going to invade one another anymore. So it's not just Ukrainians' lives that are at risk. That principle is at risk too. Which means all the lives it's currently protecting are also at risk. Not necessarily from Putin, but from future dictators as well, if the principle falls.
Mar 10 4 tweets 1 min read
Early on, the ideas of startup founders can either be grand and vague, or smaller and more concrete. Many founders think investors want the former. Maybe bad investors do. But the successful founders are the ones whose ideas are concrete. The founders who'll succeed are already digging into the problem. Ideally they already have a version one. And when you're really working on a problem, your thoughts are inevitably about the actual obstacles you're facing now. Latency issues and flaky APIs, not TAMs.
Mar 7 6 tweets 1 min read
Going to visit some antique shops in London, and I have to choose between dressing comfortably and being dissed by the staff, or not and maybe not. In the Bay Area, if you walk into a fancy shop wearing informal clothes, they assume you're some sort of startup founder and treat you well. But not in London, and *definitely* not in Paris.
Feb 17 4 tweets 1 min read
A YC partner told me about a new pattern they're seeing: a lot of the smartest and most ambitious founders are working on companies addressing climate change. There are 25 in the current YC batch. They're working on all kinds of different ideas. That's the great thing about startup founders: they have so many novel ideas. All around the world, smart people are independently deciding to attack this problem from a thousand different directions.
Feb 8 4 tweets 1 min read
Having religious beliefs about the about the future of technology is a guaranteed way to be lousy at predicting it. Especially since the source of your religious beliefs is more likely to be present ideas than future ones, which don't yet exist. That said, people who are religiously dedicated to some new technology are somewhat more likely to be right than those who are religiously opposed to it. The former often know something they can't put into words, whereas the latter are often just grumps.
Feb 3 4 tweets 1 min read
Though in European-influenced cultures we have a concept of a "calling," the fact is that people who are very good at something are often very good at several other things. The same qualities yield extraordinary performance in many different types of work. This phenomenon tends to get downplayed by biographers, who know how the story turns out. Which is justifiable, because readers care about the aspects of the subject's childhood that influenced his career as a physicist, not the career as a musician that never happened.
Jan 30 4 tweets 1 min read
If you run something with an application process, people spreading obviously bogus stories about you is actually to your advantage, because it filters out applicants who can't tell that the stories are bogus. If you were selecting people for, say, athletic ability, it wouldn't be an advantage. But when judgment is one of your selection criteria, it's useful to have third parties creating tests of it. No one tries to hack those tests, as they do those that are part of your process.
Jan 15 4 tweets 1 min read
These recreations of painted statues always have large areas of solid color, like a Victorian illustration. But if the carving was so subtle and realistic, why can't the painting have been too? There are some powerful forces pushing those who make such recreations to use solid colors. One is that any recreation with varying tones would be more speculative. But another even more powerful force is that they don't have the talent to; they're archaeologists, not artists.
Sep 5, 2021 4 tweets 1 min read
When we lived in California, the kids and I developed a custom called "night swimming," where we'd go out and swim in the dark before bedtime. In England, where we have swings instead, this became "night swinging." Now with a trampoline it's "night bouncing." I'm not sure why they love this so much, but they both do. Maybe it seems extra exciting to go off and do something when the day should be winding down before bedtime.
Sep 3, 2021 4 tweets 1 min read
People seem strangely blind to the option of solving their problems by learning more.

Many of the emails I get from young would-be founders are of the form "How can I start a startup without knowing x?" My reply is almost always "Learn x." People seem to treat their current amount of knowledge as a constant — sometimes almost a matter of identity. E.g. they'll say "I am a non-technical person" instead of "I haven't learned to program yet."
Aug 24, 2021 4 tweets 1 min read
Whoah. When startups change how they operate, they change fast.

skift.com/2021/08/20/sta… As with AWS in 2007, what startups do now, all big companies will do eventually, if only because startups are where big companies come from.
Aug 10, 2021 4 tweets 1 min read
1. Mark news stories from some past decade (e.g. 90s) as still interesting, or not.

2. Train a text classifier on the resulting two corpora.

3. Sort today's stories according to their score.

If there was an aggregator that did this, I'd probably visit it every day. It would be very interesting to see which words end up having the most predictive value in each direction. E.g. I bet "embattled" is a strong predictor that a story is not of lasting importance, but just some controversy du jour.