
Eloc
6.9K posts








Wow. Congress just tabled a bill that would *actually* kill open-source. This is easily the most aggressive legislative action on AI—and it was proposed by the GOP senator who slammed @finkd for Llama. Here's how it works, and why it's different to anything before it.

Deepseek models are available now in Cursor! Hosted on US servers. While we're big fans of Deepseek, Sonnet still appears to perform much better on real-world tasks. Enjoy!

I wouldn't trust R1 with computer use given the existing sleeper agent research. Open weights does not mean open latent thoughts. Would be the perfect setup for the most insane botnet of all time.



Spent the afternoon with the $CLSK management team and walked away with some rambling thoughts. It starts with the idea that if the DeepSeek risks materialize and alienate the middle of the curve where some Bitcoin miners operate today in AI/HPC, the three-pronged narrative might consolidate into two: the LBE (levered bitcoin equity) model or the cheapest "cost to mine" model. But could it be time to consider a new metric: "cost to own"? With SAB 121 repealed, miners now have more opportunities to offer real "asset management" services with their HODL holdings. We could enter a world where instead of reporting "36,250 cost to mine 1 BTC," you might report "0 cost to own 1 BTC over ~8 months" based on various capital market strategies. This represents a new form of capital stewardship we haven’t seen in mining before, but the end result is the same: what is the best way to acquire more Bitcoin cheaply? While the LBE strategy externalizes financing arbitrage, the "cost to own" strategy internalizes it in-house. And if there’s one company structurally set up to play this, it might be $CLSK as one of the most efficient player with scale in the space; it might be time to rethink the yield question for the Trump era. With that, congrats to @smatthewschultz @ZachKBradford on the 5y anniversary! Was a particularly memorable to be as Nasdaq today to be with you all.




/r/tx is a new recursive endpoint which lets you get hex-encoded raw transactions. github.com/ordinals/ord/p…

The somewhat hideous /r/sat/:sat_number/at/:index/content recursive endpoint now lets you directly get the content of an inscription on a sat, without intermediate calls. github.com/ordinals/ord/p…



