@sentra17@dave43law@ReemAmirIbrahim If its a self driving car and it happens by asking it to do that and the car says "sure I'd be happy to help with that" and does it, then yes the vehicle manufacturer is responsible.
@skayodev@hd_nvim@rough__sea It's better not having the shit YAML adds, simpler.
What are you even talking about? There's a lot more difference between JSON5 and YAML besides comment styles and comma counts. Have you been smoking crack?
@_georgemoller@iam_ixel E.g. visibility prop -> css `visibility` property + (if necessary) manually cleanup/reapply effects
Certainly seems that would be easier to reason about without deep study of this new niche api
@_georgemoller@iam_ixel How are we supposed to keep up with this if we don't use react regularly, much less learn it from scratch? Is a solution from first principles really that bad that we need a new api?
@scottjla@danielkempe@geoffreylitt Dunno about Photoshop but tools besides AI do sometimes do things we didn't ask for or specifically intend. Probably Photoshop included. That's not the real difference.
@geoffreylitt I produced X. Don't necessarily need to go into details of how if you're vouching for it.
When a junior dev tells me unprompted that they used ai for something I tell them I don't need to know that bc you're taking responsibility right? Always seems like a preemptive excuse to me
@BosonJoe@scorzeth@alexalbert__ I think the difference is that Claude Code uses the API which doesn't have a limit (I assume amount of code to be the same)
I still see regular people have no clue how we use AI to code
They think it's like AI Agents and lots of hype bs
But for me most of it isn't very difficult or deep or fancy but it's simply A LOT of autocompleting and prediction now
Small things like this a few mins ago:
Nomad List's weather API robot that pulls the weather every day to show on the site, sometimes it hits an error on the API side, and it sends that to my Telegram
Then I wanna know "ok for what city is it erroring?"
The moment I just put my cursor on "Bad API reply" it already knows what I want and autocompletes to "Bad API reply for city $city['name']" which will tell me for example "Bad API reply for Bangkok"
Basic stuff like that saves me a lot of seconds and minutes in a day and it adds up!
You can now run 100B parameter models on your local CPU without GPUs.
Microsoft finally open-sourced their 1-bit LLM inference framework called bitnet.cpp:
> 6.17x faster inference
> 82.2% less energy on CPUs
> Supports Llama3, Falcon3, and BitNet models