Thinking about the filters debate, it dawned on me that it's all a moot point: 80 bytes may be enough to use as a latent space in an autoencoder [1]. For example, you could encode an image (ie. embed it in the latent space), convert the latent vector to a consensus valid address, and publish to the chain. Then you release the decoder half of the autoencoder publicly (say on nostr) and anyone could interpret that image (or any image encoded with the encoder half) if they ran the decoder. That's effectively what happens when inscribing a jpeg. Granted we use traditional software for rendering/decoding the image but I don't see much distinction. 1.

Autoencoder - Wikipedia

#FreeSamourai

Baby steps towards a Dyson Sphere [1]. Initially this is a satellite constellation around the earth but over a long enough timespan it’s a race to capture as much energy as possible from the sun. And if this works you can imagine ASICs making their way up to space too since electricity makes up a large percentage of the operating costs. If starship makes good on its promise to drive the cost of space access down 100x and we already have the research and designs for running servers efficiently in space (from this Google project eventually) then ASICs quickly follow.

Exploring a space-based, scalable AI infrastructure system design

I love you all, but some of you need to get back with your ex or something because you clearly don’t have enough drama in your life 😂

People were hating on Gensler for pushing back on ETFs but have you considered that maybe an ETF for money doesn’t make sense and those products encourage the “bitcoin is property; lever up against it instead” narrative? It was inevitable that Wall St would show up but it’s not something to celebrate. They blew a hole in the housing market years ago and we still haven’t fully recovered. They’re not good bedfellows.

Support the retailers willing to accept sats for goods. Experience the difference between an ETF and free range bitcoin for yourself! (I have no affiliation with any vendors nor have I ever had one. Just an npub trying to get off fiat) View quoted note →

What a crazy coincidence all those predatory bitcoin loans came on to the market with high interest and terrible LTV ratio requirements right before the price took a nosedive so that debtors have to pledge even more collateral or just straight up lose what they’ve loaned against. The banks got lucky!

In one of my first courses on linear algebra we were taught how to do things by hand initially but from there we did everything in Mathematica. That may seem like "cheating" or that the students wouldn't retain the lesson since a computer was doing all the work. I strongly disagree with that view in hindsight. We were able to explore visually what matrices did to the inputs. For example, eigenvectors are the axes along which some action would happen and the eigenvalues determine the quality and degree (ie. reflection, expansion, contraction, etc.). If you're doing calculations by hand it's hard to get enough examples to build intuition and without spending a lot of time you can't "tinker with" the numbers to the same extent. The same is going to be true of LLMs and similar tech. Sure, they can do a lot of "thinking" for you and some will argue it leads to brain rot (and maybe it will!) but it also enables you to explore many more possibilities in your finite time than you otherwise could.

Rightly or wrongly when all the influencers have been saying we’ll have a melt up or the Fed has to print trillions I’m starting to think the odds of an overt default are increasing.

Happy bitcoin whitepaper day to those who hodl