Week 1161
The OpenAI situation is finally resolved with Sam coming back as CEO and a board changeup. I read some OpenAI lore to get more context on the events leading up to this weekend:
- How ChatGPT Fractured OpenAI (The Atlantic, Nov 2023)
- The Inside Story of Microsoft’s Partnership with OpenAI (New Yorker, Dec 2023)
- A Brief History of OpenAI (Wired, Sept 2023)
- More History on OpenAI (MIT Tech Review, Feb 2020)
- Sam Altman Profile (New Yorker, 2016)
My main takeaway from these articles is that I should think bigger and be more ambitious. “Always think about adding one more zero to whatever you’re doing” (Sam in his New Yorker profile). I also found this snippet about his personal finances quite interesting: “he decided to rid himself of all but a comfortable cushion: his four-bedroom house in San Francisco’s Mission district, his cars, his Big Sur property, and a reserve of ten million dollars, whose annual interest would cover his living expenses. The rest would go to improving humanity.”
--------
Justin and I flew to NYC! We are here to visit some friends and to try working out of a new city. We decided not to fly the redeye flight so we wasted a day just traveling but I think it was worth. I’ve had extremely bad experiences flying redeye flights and I don’t think the extra time of day is worth the poor sleep and feeling of lethargy that comes with it. It might work for other people but for me the next day is usually wasted anyways because I’m too tired to do anything.
On the flight I took a few naps and read some articles in between. I read about SIMD and how control flow and memory operations are the main culprits for slowing down code execution. Stability AI released it’s SDXL-Turbo model which claims to generate high quality images in significantly fewer time steps than existing models. They trained it using a method called “adversarial diffusion distillation” which incorporates ideas from GANs to achieve better quality images on the earlier time-steps. I reread Jessica Livingston’s recounting of her childhood and the early YC years. I also enjoyed Molly Mielke’s Moth Fund interviews. Here are two quotes that stuck with me:
- “I come to California and start a company and have that company be acquired by Apple and have a pretty senior leadership position at Apple when I was pretty young. I was 18 when I moved to the US, 23 when I was running a bunch of machine learning stuff at Apple, and I'm 31 now … but underlying all of that is I by no means made it. I definitely have sleepless nights, things I'm working on all that sort of stuff, but I'm in a much better position than I was when I first started and I do get joy out of paying that forward to other people.” - Daniel Gross
- “[Patrick Collison] spent 50 hours with me before they made an offer.” - Claire Hughes Johnson
--------
The first 4 days in NYC have been fun although the weather has been cold and rainy. We stayed at a friend’s place in Williamsburg who’s been extremely kind to let us crash at their apartment. We also visited another friend’s new startup office and caught up with some NYC Retool folks.
I have a few observations about NYC vs SF so far.
- SF streets smell like human feces while NYC streets smell like trash
- NYC streets are less pungent
- NYC has less homeless people
- The most noticeable difference has been that I don’t feel hyper-alert while walking around NYC like I do in SF. There’s so much activity that even walking around at night doesn’t seem that dangerous.
We got invited to a tech meetup at the General Catalyst NYC office and I learned that they started their own healthcare company?? They have 20+ partnerships with existing health systems but want to eventually own and operate their own health system. It’s seems odd to have a hospital be run by a venture fund but I’m curious to see how it plays out.
--------
I’ve been reading a lot of engineering blogs lately. Here are some that I’ve enjoyed:
- Database in the Browser by Stepan Parunashvili
- A Graph-Based Firebase by Stepan Parunashvili
- The thesis behind Stepan’s startup InstantDB: a DB with the devX and flexibility of Firebase but the query strength of Postgres
- RethinkDB: why we failed by Slava Akhmechet
- Spent too much time building product instead of doing distribution. By the time they felt the product was “complete,” they found it hard to sell the product because competitors were already entrenched in customer workflows despite being worse product. However, the competitors were able to survive for longer because they invested more in GTM. Their product quality eventually caught up to and surpassed RethinkDB over time.
- The data model behind Notion’s flexibility by Jake Teton-Landis
- Farewell EC2-Classic by Amazon CTO Werner Vogels