jderick

Bo Burnham

dev vs validation team

I think with guitar it is easy to enter flow state because it is easy to avoid playing the "wrong" notes. Probably similar to these other hobbies. Perhaps they just happen to have a tactile component, although it is nice to do something a little different from time to time. It seems computer games probably provide a similar form of satisfaction.

Can you put them all on one page and zoom in/out? Might be interesting to see some patterns in the sequence. Maybe allow filters for different factors or number ranges or different groupings.

Was this in SV?

Could this be used to train an LLM? It seems the hidden states could be used to learn how to store history.

Distributed systems are hard. I like the idea of "semantic locality." I think it can be achieved to some degree via abstraction. The code that runs across many machines does a lot of stuff but only a small fraction of that is actually involved in coordination. If you can abstract away those details you should end up with a much simpler protocol that can be modeled in a succinct way. Then you can verify your protocol much more easily. Formal methods have used tools such as spin (promela) or guarded commands (murphi) for modeling these kinds of systems. I'm sure you could do something similar with the lean theorem prover. The tricky part is mapping back and forth between your abstract system and the real one. Perhaps LLMs could help here.

I work on hardware and concurrency is a constant problem even at that low level. We use model checking tools which can help.

I wonder if this can generalize to mandelbulb?

Step one should be having an edtech ecosystem that doesn't allow students to use the web. There are just too many distractions online. I think noone has really invested the kind of capital required to do a good job with this. Most of the software my kids have to use for school is pretty bad.

It's a lot harder to make a single player game have depth and complexity than it is for a multi player game, since you don't have human opponents.

It doesn't make sense unless zooming out shows some higher level representation of the code.

It is interesting to see the type of analysis he does and the visualizations are impressive, but the conclusions don't really seem too surprising. To me, it seems the most efficient learning algorithm will not be simpler but rather much more complex, likely some kind of hybrid involving a multitude of approaches. An analogy here would be looking at modern microprocessors -- although they have evolved from some relatively simple machines, they involve many layers of optimizations for executing various types of programs.

I'd be curious to see an example of what you are talking about wrt his analysis here.

What is GPR?

Even great regular coffee isn't ubiquitous yet.