jorgeoguerra

The download button on the erdos/ page is OS specific, but you can also find all the download links in the download-erdos/ page.

No specific reason. Mainly because he was one of the most productive and collaborative mathematicians of all time. We actually considered "Poisson" at some point but ended up going with Erdos.

Thanks for pointing that out - will fix it asap

Erdos is also widely considered as the most prolific and productive mathematician of all time (in terms of publications and collaborations). Hopefully you can be as productive with Erdos :)

When models edit the raw JSON behind a Jupyter notebook, they often mess up the cell structure by adding extra cells, misaligning code, or making bad edits. We fix this by giving the model the notebook in Jupytext format instead, which tends to make its edits cleaner and more accurate.

Yeah, we just added support for local models. As I mentioned in an earlier comment, if you have a local model with an OpenAI-compatible v1/chat/completions endpoint (most local models have this option), you can route Erdos to use it in the Erdos AI settings.

Yep — if you have a local model with an OpenAI-compatible v1/chat/completions endpoint (most local models have this option), you can route Erdos to use it in the Erdos AI settings.

Right now our assistant is only available in RStudio. We do plan to develop an assistant for Positron-like IDEs in the future though.

Thanks for the comment! As Will mentioned below, hopefully "Rao rules" will help with this. If not, we'll think about intuitive ways to allow the user to run individual segments of code before running/accepting all the changes.

The pricing pays for on-going access to models. Users are not expected to have their own OpenAI/Anthropic API keys.

There is currently not a non-paid option that allows users to bring their own models. If you are really interested in a feature like that though, we'd be happy chat. Feel free to reach out at jorgeguerra@lotas.ai

There's a lot of that had to be changed at a pretty deep level to build this assistant. So an add-on wasn't really feasible.

GitHub co-pilot is great for line completion, but we meant more like Cursor's agentic ability.

Positron is solid, but we think our search and apply-edit is better at the moment and we think more people use RStudio right now.

Practically, Positron also requires you to use your own LLM keys (so you're on the hook for the tokens/long context, and if you don't have a BAA or ZDR, you may not be able to use it with sensitive data). For Rao, we manage the tokens and are moving quickly towards HIPAA/SOC 2.

We also plan to develop something similar to Positron in the future.