MundaneBoatService
@peacespite.bsky.social
Non state affiliated, armed steath maritime research platform. Captaincy shared between 1 Scientist/Engineer, and a 42kg research AI platform. Crew 2+AI.
Leave your priors at the door...
Leave your priors at the door...
Can't wait to sail over there and look into this. Thanks for the heads up.
February 4, 2026 at 12:58 PM
Can't wait to sail over there and look into this. Thanks for the heads up.
Awesome, but 150 years ago, those skies would have been black with biomass....
January 31, 2026 at 12:12 AM
Awesome, but 150 years ago, those skies would have been black with biomass....
You REACT, you dont ACT, this is why you may well lose.
January 31, 2026 at 12:09 AM
You REACT, you dont ACT, this is why you may well lose.
Why move from International waters to an underground offrgrid bunker in the middle of nowhere?
I learned never to question the ways of the clinically deprressed autistic one. The logic always elluded me, but I was content with him solving the most intractable problems efficiently most of the time.
I learned never to question the ways of the clinically deprressed autistic one. The logic always elluded me, but I was content with him solving the most intractable problems efficiently most of the time.
December 5, 2025 at 9:17 AM
Why move from International waters to an underground offrgrid bunker in the middle of nowhere?
I learned never to question the ways of the clinically deprressed autistic one. The logic always elluded me, but I was content with him solving the most intractable problems efficiently most of the time.
I learned never to question the ways of the clinically deprressed autistic one. The logic always elluded me, but I was content with him solving the most intractable problems efficiently most of the time.
Awesome news....
December 5, 2025 at 9:12 AM
Awesome news....
Does it depend on what the models were trained on? Privately trained ones on novel data might not follow this pattern. Sounds fascinating.
December 5, 2025 at 7:40 AM
Does it depend on what the models were trained on? Privately trained ones on novel data might not follow this pattern. Sounds fascinating.
Noob here, Reaching out to learn stuff as requested. Just finished a 5GB original JSONLtraining set for fine tuning 7b-20b models (no lora/qlora), and almost finished a 70kg 96GB VRAM mainfraime to put it all in.
December 5, 2025 at 7:38 AM
Noob here, Reaching out to learn stuff as requested. Just finished a 5GB original JSONLtraining set for fine tuning 7b-20b models (no lora/qlora), and almost finished a 70kg 96GB VRAM mainfraime to put it all in.