July 19 2025th. I'm back in Beijing. My coauthor flew me back, and home is great.
We talked a bit and most of the theory problems they are into were somewhat irrelevant to my interests. One of my coauthor's students was working on a idea that I brought up about like a year ago for using semidefinite programming to find bounds on finite size qldpc codes, and he pulled me into the overleaf for it. He had another visitor who was interested and asking about using rl to optimize circuits, but from my understanding it wasn't looking that optimistic. A while back Jan told me about how their efforts were stuck on low distance codes and it seemed the only way around it was more compute. The compute costs didn't seem to scale linearly at all though. It took them a ton of effort to go from distance 3 to 4 and we are going to need at least like distance greater than 20 for it to be practically relevant. I was also hopeful about quantum chemistry before. My coauthor mentioned that one of the top researchers in the field had lost hope in it for all kinds of reasons, and was steering people away at QIP. He got some backlash on twitter for that.
I asked my coauthor about his first paper. He told a story about how he was first into string theory, then realized the field was dying about half a year in, then switched to quantum information. He chose his advisor because he was doing all kinds of interesting stuff in quantum information and the conversation were always great, but his advisor didn't actually advise him much. So he worked on some communication problem inspired by something he saw in the PBR theorem paper. I didn't really understand the details but I think he said something about the idea of using an infintely long message to communicate 0 information, not that it was going to be useful for something, but he found something about it interesting. Scott Aaronson was also a collaborator on this paper. In hindsight, it was amazing to him that he could produce good research himself, and without much guidance His second paper was a major hit and its his most cited paper to this day. He said it's kind of amazing the state people are in when they are new and striving and unburdened. I think my coauthor's phd was partially funded by epstein's 2017 donation. His advisor took money from epstein and seemed to have knew him well and visited him in prison. Sometime I will get around to asking about it. We also talked a bit about one of his theories for proving how time travel was impossible because of how it would force quantum mechanics to incoporate some nonlinearities that would break things down. His advisor did some work on post selected quantum mechanics and what not back in the 2000s. It's comforting that the grandfather paradox is impossible under this formulation, but has a host of other problems like breaking computational complexity. I asked why he didn't go about writing up a paper for it and he said it was largely because he couldn't really give it to one of his students because there's a possibillity it would fail and delay their phd graduation date, and he was still in survival mode to churn out papers since he doesn't have tenure yet. He told me about how his advisor had all kinds of wild theories for things and how that's what many theorists live for. There's some very exciting stuff to think and talk about, but it seems few early and mid career academics get the chance to dedicate themself entirely to these things because how brutal survival is in the academic world.
I stopped working on the startup, I thought I had better ideas for what to do with my time. One of the cofounders I first met around a year and a half ago. He was easy to get along with and had a decent amount of personality. He was cofounder at a series a startup that was training their own diffusion models to do interior design and they had pretty fast growth and raised 36m which seemed big to me. Also, his family owns some factories in China and they're wealthy. He talked all about how they would use it as a wedge into doing more interesting things with video and image and worlds and what not.
We kept in touch, and like a year and a half later when we were talking he told me he figured his current startup was going to be stuck doing interior design forever and was going sell it off for like 150m and do something new, I moved in with them later after I reported my landlord and came on as cofounder. Much time was spent just talking and thinking and we tried out some interesting things here and there. Eventually him and the other cofounder somhow decided to pivot to some consumer social app, so I pulled the plug since that wasn't what I was there for. I was under the impression we would do something a bit more innovative, since his last startup was doing interesting stuff with diffusion models and he was knowledgable about them.
I met some old guard people in semiconductors who made fortunes and invest now. They go around the world golfing and what not. I heard one of them telling another about some itinerary to iceland that would be 270k usd. They were all pretty interested in talking about quantum computing since I had some experience in the field and most of them believed quantum computing could change the game. None knew about the algorithms crisis in the field. I told them about it but I don't know if they really believed me. People outside the field really don't know much about it, and people in the field are heavily motivated to talk positively about it. My prediction is that they are probably going to end up like particle accelerators or gravitatonal wave detectors or huge telescopes and what not, essentially another device to probe physics and what not with.
I also talked to someone who recomended me just go straight for bci and not worry about anything else. She did some major work at circle and they invested in her startup, she raised like 8mil. She said that if she were in my position she would just go straight for it and a few decades out everyone who got into it early would be in an amazing position.
It was a good point. I spent some time looking into BCI before and my conclusion was that reading and decoding the brain was far along enough already, and mostly inevitable in the coming years whether i did anything or not. The real challenge would be writing/encoding to the brain. Once writing is solved I think we might be able to become immortal in some sense. After I took a MEMS class with lin liwei in fall 2024, I realized we would need some serious progress to be able to do any meaningful writing.
The idea is roughly that if we can figure out how to write internal monologue to somebody's mind, this somebody will be under the impression that it was their own thought. Then their consciousness can kind of be slowly extended with the read write loop until it can persist without their biological brain. Theres a ton of ways this can go wrong and we get screwed over or screw eachother over though, and what happens after could go a lot of ways.
A few new acquaintances of mine tried to convince me to join their startups as an engineer. Some were well funded or had been through top labs or were twitter microinfluencers with a few thousand to tens of thousands of followers. I also met a decent number of people who had worked directly with some of these mythical billionaire founders and what not before. I even met some people who tried to figure out what stuff I was interested in and talk me out of them. I was surprised that was a recruitment strategy people used. I'm glad I didn't meet anyone early on who was capable of brainwashing me.
One of the sources I would diagnose for the huge swathe of generic software startups is the idea of "searching" for product market fit. This idea has totally infested the silicon valley startup circuit, and for many early stage founders it is something they wrestle with daily and rarely if ever question. This idea often frames figuring out what people want and are willing to pay for as the truely important part, while building the product becomes just a matter of execution and nothing innovative technically.
I had always wanted to attack problems that were obviously important and just lacking solutions, it was strange how this preference seemed to place me in the minority there. This made me think a bit about what the role of an entrepreneur should be, and eventually I skimmed through some scholarly theories on the whole matter. In summary, for Frank Knight the entreprenuer's role is to bear uncertainty about demand, for Israel Kirzner it is to notice gaps in the market and exploit them, and for Joseph Schumpeter it is to innovate and bring about "creative destruction".
The idea of searching for product market fit echoes of Knight and Kirzner. Schumpeter's idea of the entrepreneur has not caught on as well, but I much prefer it.