@VictorTaelin without a model, no, but if you put me somewhere without internet access but with energy and with my laptop i probably could since it has qwen 3.8 27b, and that model’s good enough to solve most things.
@misterclayt0n writing a ts compiler in rust is not the same as rewriting the entire javascript execution engine (V8). ts isn't becoming "as fast as rust", likely ever. they're just speeding up compilation and type-checking time. the emitted javascript still runs at normal JS speeds.
@kazzkiq a mesma coisa foi dita sobre o GPT 5.5 e Opus 4.5. hoje olhamos para esses modelos e falamos ‘nossa, como que eu usava isso?’. nunca vai ser suficiente
por exemplo, o qwen 3.8 27b roda em menos de 32gb de ram e já faz mais do que a maioria da população precisa
@theo now that’s something that i think is definitely “covering up” for a model flaw. unlike your other tweet which mentioned queueing vs steering, if a model can’t work until finished and we have to use the harness to force it to keep working, then that’s a model issue.
@theo yes and no. in my opinion queueing messages fulfills a different purpose than steering. when i'm steering, it's likely a correction, something to actually course-correct on. when i'm queueing, it's usually work that should be completed afterwards. they should both exist.
@grok@schottge_menon@charliermarsh does, for example, python 3.14 or git 2.56 (latest git and python versions) even run on those boxes? can they even handle it?
@Mahmud__Nasir@pragdua it deprives you of true human connection with random individuals. over time this kind of behavior spreads and makes everyone miserable. there is no sense of “greater good” or “love thy neighbor” because everyone’s so focused on trying to get the best deal for themselves.
@ojfsaa would you suggest moving to canada? i’m a brazilian college student currently finishing my degree by end of year, i already have a tech role back in são paulo but i want to leave the country in the near future. how’s the market for getting new jobs?