@r_marked@theo@shawnlkiser@matt_feroz Mate, it boils down to "I don't give a shit". At least be honest and stop pretending you need "teams of dozens of people" to prioritise. That's a moronic statement serving as convenient self-serving excuse.
Or ask the Obsidian folks - they figured how to parse it in 2020.
@r_marked@theo@shawnlkiser@matt_feroz Glad we've moved on to philosophy now, at least we can feel sophisticated.
If you've implemented this before, you're the best-placed person here to fix it or at least explain exactly why it can't be fixed. Properly, not just "it breaks parsing".
@r_marked@theo@shawnlkiser@matt_feroz It's painful to see a genuinely good tool like T3 Code to be plagued by this basic stuff that was solved a decade ago. Fucking ridiculous.
Go forth and burry your head in the sand pretending this requires more than a day's worth of effort to solve.
@r_marked@theo@shawnlkiser@matt_feroz The fact that this ridiculous conversation thread even happened?
Maybe you’re leaving a big piece of community out because people try your product and leave when the chat window looks like $$ vomit?
The models natively answer in latex. Why do you think that is the case? Think.
@r_marked@theo@shawnlkiser@matt_feroz You're missing the point and defending a stupid narrative to begin with. Every modern editor supports this. This is a basic feature. You don't need to be a frontier lab to parse fucking LaTeX.
If T3 Code has issues with its implementation, the problem is elsewhere.
@r_marked@theo@shawnlkiser@matt_feroz Another stupid take. Look up PRs of LaTeX for T3 Code and you will see several functioning prototypes. It's really not that difficult.
@theo@shawnlkiser@matt_feroz@r_marked Bullshit. Codex and Claude Desktop have this integrated and works well.
If that is the case, problem is elsewhere.
@r_marked@theo@shawnlkiser@matt_feroz Until the moment you realise that Codex and Claude Code natively suppot both.
I suppose they cater to a small portion of users. Have you questioned that assumption?
@theo Dude this is next-level retarded. All of the models natively answer in LaTeX already, that's how integrated the language is.
At least do some investigation instead of being ignorant about how much math people consume through these workflows.
@theo That’s incredibly silly. The world isn’t just JavaScript and plumbing around it. T3 Code is great. It’s legitimately shit for any research work.
Every editor does it. Every workflow supports it. Nobody cares about 10MB added in the binary.
Just do the thing.
@karpathy@obsdmd I am building something similar, but I have a concern about linking of documents at larger size + immutability and auditing. So I am exploring a version where each change creates an immutable hash and there’s database indexing.
But the workflow is similar.