davkap92

Tbf its in the description on the page "Competitive with models 4x larger on tool use, instruction following, and multi-step agentic tasks."

Well Karpathy did. this is basically his work (with the big caveat if there are no LLM mistakes/hallucinations). Its just exporting his transcript to a different format. I think if its presented as 'look what i built' its misleading. but if its presented as, this format of information was helpful to digest, and properly attributed to the author (which it was here) theres no real issue

Interesting but by telling it to check X for mentions of itself, that is an action.. wouldn't this essentially direct it and hence be steered/controlled by random individuals on the internet?

Nice interface, a few tweak suggestions, its not clear when things are auto updating maybe some more indicators some tooltips, such as retirement spending, is that annual? nice job though!

Looks cool. Just a note might be worth mentioning you're benchmarking the desktop scores as the mobile ones can be significantly different on pagespeed insights. Maybe show/track both? Seeing as mobile usage is significant

This one looks good too https://material.io/resources/color/#!/?view.left=0&view.rig...

I was just thinking the same thing! May also build something like this. This from cloudflare seems quite close to what i had in mind https://cloudflare.design/color/

I'd test this myself but don't have a guitar with me at the moment

Looks very cool will try this out and report back but just curious first thing that comes to mind, how are you differentiating between the several frets that could produce the same note, or for example 6th string 5th fret(A) vs open 5th string(A) ?

I had ssl error... and then realised it was because I have dnsmasq on *.dev at the time there was no dev tld :P

I've been in the same boat... guess in any profit focused business eventually the cs stuff has to be wrapped up into some form of value adding service, and 'CRUDified' and delivered. I suppose which team you're on or maybe at a startup depending how early on you get into it, could impact how much time/opportunity you'd have to work on the interesting problems

Nice idea thanks! One suggestion maybe add some sort of priority marking/ time spent for each one, as I would say some are definitely higher priority than others

I like the changes nice job!

Ok, so as an example you're a developer new to Node and associated technologies, I think it may be helpful if one could visually see how all the tools interact with each other. Grouped by say CI, Task Automation, Libraries etc. A visual story of the complex of web of technologies that exist and how they interact with one another.

Regarding your first point, not sure if author covered it, but in Google Content Experiments with multi bandit approach cookies are stored, so user who sees variation b will keep seeing b while the experiment is running

Interesting but since the writing of this article (2012) Google do offer Multi Armed Banned Content Experiments... https://support.google.com/analytics/answer/2844870?hl=en

Microsoft Office 3D here we come, walking past life size graphs and real life spreadsheets

this is the little bit I have so far... https://github.com/davkap92/codecontext

I made http://www.valueforest.co.za a website that aggregates all the major South African classifieds into one place. Growing well 50K unique users a month, will probably have to get partners at some point, as they say you need a hacker/hustler/hipster and I'm the hacker.