This is a great idea. Can you share more about what "24 specialized agents" mean in this context? I assume each agent is not simply an LLM model with a specific prompt (e.g. "You're the world's best biologist. Review this biology research paper.") but is a lot more sophisticated. I am trying to learn how sophisticated it is

Here is a description of how it works: https://github.com/robertjakob/rigorous/tree/main/Agent1_Pee...

> To my surprise, Zoo’s API didn’t perform particularly well in comparison to LLMs generating STLs by creating OpenSCAD

This is interesting. As foundational models get better and better, does having proprietary data lose its defensibility more?

Zoo co-founder here. Our product is still pre-v1. But getting to v1 very soon. We actually built a whole new CAD kernel from the ground up. I say this because we can't actually train on models the CAD engine does not yet support. Just 2 weeks ago we shipped csg boolean operations to the CAD engine. This unlocked new data to train our model on that use those operations. So its actually fair to say at the time he used our model we using about 2% of the data we actually have. Once we can use more and more the ability of the model will only get better.

Training an audio model this good from 0 prior experience is really amazing. I would love to read a blog post about how you guys approached ramping up knowledge and getting practical quickly. Any plans?

Congrats! What made you not want to take any investments and what was the most counter intuitive thing that made you succeed?

We were always profitable and growing at a sustainable pace, so never needed funding.

Most counter intuitive thing was to not think I was missing out on anything as we didn't have VCs backing us.

I'm a fan of your work! I'm curious about how you decided to work on building a terminal for your next project among your other ideas. If you have time later, could you share your main motivation with us or link to an existing post if you already mentioned it elsewhere?

To people who tried using it, what are the reasons to use it over iTerm2?

It’s fast. In the same league as Alacritty and Wezterm, which is to say on the doom fire benchmark those terminals can do 500fps plus, where iTerm2 does 90fps on my M4 MacBook Pro. In practical terms if you use tmux or vim in the terminal, it makes a big difference for typing latency and scrolling.

Thank you! Interesting, because I haven't perceived "terminal performance" as a problem before. I will try it out

The parent is probably referring to this benchmark: https://github.com/const-void/DOOM-fire-zig/

Where should I look first if I only get around 160 fps, while WezTerm gets 400+ fps?

I am not a very adept terminal user. I just hook in to tmux. So all emulators are basically the same for me. Switching windows in tmux is much faster than clicking on tabs. ghostty at least looks much nicer. It has a lot of built in fonts and themes. Command line editing is similar. Man pages for some reason are very slow to open in iterm2, but that isn't the case here. The only deal breaker for me is that it takes up a lot of cpu (in macbook air). Unless there is a simple, non-consequential config to change that will fix this, I will stay with iterm for now.

> Switching windows in tmux is much faster than clicking on tabs.

Sure, if you're clicking tabs... A fairer comparison would be between the keyboard shortcut to switch windows in tmux and the native keyboard shortcut to switch tabs.

which is prefix+number in tmux. I don't know if you can easily jump to a tab with a visible index in ghostty.

On Mac at least it seems to be cmd+number

I tried it but I just can't jump since it doesn't have support the global hotkey to autofocus iterm that I'm so used to.

I tried using their toggle_visibility keybind but it's a bit wonky since it doesn't return focus back to another window when you toggle away.

I just use rcmd for that. But that iterm feature is nice.

I love the landing page. Great job!

Thank you!

What problems did you have with Twilio?

Several years ago Twilio wasn't so bad. It always had pretty antiquated API practices and felt clunky but it worked.

Nowadays, it's expensive and contains a ton of hoops to jump through for business and a2p verification. They charge you for many things (monthly for a number, for verification, every submission for a2p and re-submission if they reject it for whatever reason). I tried to contact support and they wanted me to upgrade to a support plan to even ask a question.

If I don't find a simpler solution I'll remove transactional SMS messages entirely from my application. I'd rather remove it than deal with Twilio.

Where does the data come from?

Scraping

Wow this is really cool