This looks very promising. It would be cool if support for Bonsai-style models would land in mainline MLX soon, looking forward to trying it out.
It seems PrismML has implemented a better version of an idea I had a while back: what if we had a 1-bit model where the scale of the weight is determined by its position. The model would have to be trained from the ground up for this though, which is why I never tried it. The interleaved scale factor approach of Bonsai is a much more flexible approach at almost the same cost.
I know this is an outsider position, but I always felt that the AoC leaderboard was a mistake. Very few people had the time, the commitment, and the capability of making it on there in a meaningful fashion, and it put an emphasis on something that didn't match the vibe of the event at all. If speedrunning the problem solving was the point, then why package every episode into an enjoyable little story?
This also ties into the comments that AoC has become moot or was "ruined by LLMs". If you enjoy solving the problems, nothing should have changed for you. What's the difference if a given problem was already solved by an LLM, or a group of IQ 200 superhumans from MIT for that matter?
As time marches on, there will eventually be absolutely nothing left where an unaugmented human outperforms a machine. That doesn't mean you have to stop enjoying things. In a few years at most, all programming will be purely recreational.
> You're happy to not use it. I'm happy whether you use it or not.
I think you misunderstood the OP. The criticism was about the landing page, not the project.
And fwiw, it's not unwarranted. Putting the words "terminal emulator" somewhere on the page would make less people bounce off. You may be fine with them bouncing off, but still, I think it was meant as constructive criticism.
If I was GitHub, I'd make a distinction between two random people shuffling 10 bucks back and forth and two people shoving hundreds or thousands around. The first scenario is probably an accident, the second is not. Come to think of it, they probably should have capped individual contribution matching to something reasonable (maybe they'll still do that).
The biggest takeaway for me here should sound familiar: don't overengineer your data model. The more decisions you make about specific properties, the more special cases you will need to handle (and those are a continuing source of frustration for both programmers and users down the line).
In this case, most of these issues can be avoided by allowing a free text field instead of dedicated fields for street name, number, apartment, floor, state, county, district, and whatever.
In this specific case merely forging a POST request externally does nothing as the user has to be signed in for this to work. Thus, restricting actions such as voting to POST (and restricting to same-origin) does constitute adequate protection if there is no XSS vulnerability on the site.
It seems PrismML has implemented a better version of an idea I had a while back: what if we had a 1-bit model where the scale of the weight is determined by its position. The model would have to be trained from the ground up for this though, which is why I never tried it. The interleaved scale factor approach of Bonsai is a much more flexible approach at almost the same cost.