⚡️🌙 @dystopiabreaker.xyz · 12/08/2026"the so called meteor so called scientists have detected headed towards earth is actually just fake to prop up their big science bubble. no i will not look up. i am very serious buy my substack" 1290
⚡️🌙 @dystopiabreaker.xyz · 07/06/2026miss this discourse tbh. it was very funny in this specific way 1401
⚡️🌙 @dystopiabreaker.xyz · 27/05/2026one of my favorite cheeky papers, from 1999, is this one where the authors argue that sometimes it is better to simply wait and do nothing, because your astrophysical simulations will complete faster if you simply wait for the next epoch of compute arxiv.org/pdf/astro-ph... 511613
⚡️🌙 @dystopiabreaker.xyz · 18/05/2026average MLeng: 'well, the adversary would have to know the details of how our defense works' the humble Kerckhoff axiom: 1251
⚡️🌙 @dystopiabreaker.xyz · 18/05/2026you may think your service on localhost cannot be accessed by remote services, because it is listening on localhost and the same origin policy protects you. however, have you considered: 1131
⚡️🌙 @dystopiabreaker.xyz · 18/05/2026neural networks try to design towards “all structure” in their layers, whereas block ciphers must be “no structure at all” (any structure is a break, usually) 2211
⚡️🌙 @dystopiabreaker.xyz · 09/05/2026friday night, perfect time to work on something completely normal 0291
⚡️🌙 @dystopiabreaker.xyz · 25/04/2026the pollution would from gas turbines would be substantially less concerning if they were actually 'burning pure methane' 2614
⚡️🌙 @dystopiabreaker.xyz · 24/04/2026one cool thing you can do with simple, lightweight systems level mitigations/instrumentation like this is stick them in a zkVM (like SP1) and produce succinct proofs over arbitrarily sized agent trajectories, and show compliance to some agent policy without revealing the transcript at all 3142
⚡️🌙 @dystopiabreaker.xyz · 24/04/2026sometimes i bring things to claudes and am just like 'do u like this' 41063
⚡️🌙 @dystopiabreaker.xyz · 24/04/2026side note the new image model is pretty good at understanding scale 0140
⚡️🌙 @dystopiabreaker.xyz · 14/04/2026what a beautiful scene. no nuclear radiation or datacenters in sight. 819523
⚡️🌙 @dystopiabreaker.xyz · 07/04/2026claude mythos is far more token-efficient than opus. continuing the trend 61428
⚡️🌙 @dystopiabreaker.xyz · 31/03/2026the breathless reporting on datacenters has escalated to “datacenters are dropping 18 hiroshima scale nuclear bombs on their communities every day” 1120431
⚡️🌙 @dystopiabreaker.xyz · 31/03/2026yes, you might imagine using a rater model as the 'correction loop' here, but unsupervised pretraining results in all the models converging on very similar representation spaces, so your rater will have the same 'taste'. a difficult bootstrapping problem 170
⚡️🌙 @dystopiabreaker.xyz · 31/03/2026claude made this in 10 seconds and it felt important to share 5473
⚡️🌙 @dystopiabreaker.xyz · 30/03/2026i dunno man. every single capabilities graph looks like this 140
⚡️🌙 @dystopiabreaker.xyz · 26/03/2026and for context on the actual size of the rate limits, this includes implementing three different complex games in rust (p2p rollback netcode, 3d rendering, server-authoritative state, ...), including procedurally drawn assets 230
⚡️🌙 @dystopiabreaker.xyz · 15/03/2026even under extremely generous assumptions, the “AI is drinking all the water” thing is empirically fake. (infographic from @andymasley.bsky.social ) 713117