if you think proprietary model developers like anthropic or openai will release their best models to the public, you’re dead wrong. they’ll always keep their strongest models for themselves and use them internally with zero and i mean zero care or obligation to the public good. this is why open models matter. no private or public company should have access to a model more capable than the best open weight model available to everyone.

until ai gets better (way better), you human devs are the testers of your own ais. not the devs, testers. like qa people. thats the current bottleneck.

bip110 supporters complain about spam, bitcoin getting captured by centralized mining, etc... but somehow they’re totally comfortable with a one man show.

llms have absolutely zero intrinsic model of spacetime. time enters only as an externally supplied symbol like a timestamp or token order. there is no duration concept or anything close to lorentzian geometry. reasoning about spacetime as an abstract concept is different from experiencing it. but here is the problem. we do not perceive spacetime directly either. what we call the passage of time may be an emergent construction built from memory, biological clocks or prediction at best. the brain continuously compares a remembered state with a present state and anticipates a future one. that creates the subjective sensation of duration and motion through time but not the intrinsic temporal frame itself. so in a way either there is no time to perceive or we’re just primitive biological llms isolated from most of reality, trying to infer its deeper structure from a tiny stream of signals available to us.

nostr analytics. View quoted note →

the core principle of decentralization is giving everyone an equal amount of freedom. i’m still debating myself whether spammers and ai bots deserve that same freedom. to be perfectly honest, im not sure yet and im not ashamed to admit it.

next up: a native Nostrich checkout. pay in sats, zap the seller and done. because a store front should eventually be more than a list of links. and we're gonna build it. View quoted note →

cought the lucky 7

manage your timeline with @Nostrich and carry your filtered preferences across compatible nostr clients with nip-51. View quoted note →

nyc 1900 vs 1913. only 13 yrs apart and cars had basically replaced horses in just over a decade. now apply that same rate of change to ai and whatever you’re doing today.

ai took away the mechanical part of building and left us with the ideas and creativity. take full advantage of this era.

1 million believers are more than enough to take $btc to $1 million a coin. right now maybe 700–800k people own 1 btc or more, including the large holders. 5 yrs from now that number could look very different and so could the price.

$btc moving upward even though the 4 yr cycle still isn’t quite complete tells me one thing: as supply keeps shrinking, the cycle’s dominance will keep fading too.

ai still doesn’t get unit tests right. it writes a bunch of noncritical stuff and misses what actually matters. still not sure if that’s a software limitation or just ai not fully understanding the business logic.

yesterday's $btc etf inflow +$200m. we're so back...

gm! someone pls create a nostr account called rotten potatoes and start posting actually good movie reviews with some nice visual galleries. 💡🥔

building and testing on nostr relays feels like running through a desert for days with no water. then you switch to another project with direct db access and suddenly it feels like you’re swimming in an infinite amount of water. devs know exactly what i mean.

apple just dropped the new m5 ultra and the fully specced out version costs 0.25 BTC. yeah we're going to a million per coin.

About two yrs ago on july 21, 2024, gold was about $2,400, silver $29, and copper $4.20. Today, august 24, 2026: Gold: $4,639 → +93% Silver: $68.54 → +136% Copper: $6.61 → +57% It's just the beginning. we've seen nothing yet. especially for bitcoin. #supercycle View quoted note →