Le Chat: Custom Mcp Connectors, Memories

Inspired by the Greek word for human: Anthropos / ἄνθρωπος, the same etymology as English words like anthropology, the study of humans.

(I'd hazard a guess that your first language is something like a Romance language such as French, where people would pronounce that "anthro..." as if there is no h? So a particularly reasonable letter to forget when typing!)

This said, I am really supportive of Mistral, like their work, and hope that they will get more recognition and more EU-centric institutional support.

https://news.ycombinator.com/newsguidelines.html

It’s pretty rare though. Really solid model, just a few quirks

Would you like to elaborate further on how the experience was with it? What was your approach for using it? How did you generate synthetic data? How did it perform?

In slack/teams I do it with anything someone might copy and paste to ensure that the chat client doesn't do something horrendous like replace my ascii double quotes with the fancy unicode ones that cause syntax errors.

In readme files any example path, code, yaml, or json is wrapped in code quotes.

In my personal (text file) notes I also use ``` {} ``` to denote a code block I'd like to remember, just out of habit from the other two above.

This is maybe more maybe less insidious. It will literally just insert a random character into the middle of a word.

I work with an app that supports 120+ languages though. I give the LLM translations, transliterations, grammar features etc and ask it to explain it in plain English. So it’s constantly switching between multiple real, and sometimes fake (transliterations) languages. I don’t think most users would experience this

I got this format from writing markdown files, it’s a nice way to share examples and also specify which format it is.

It also helps with having a field of the json be the confidence or a similar pattern to act as a cut for what response is accepted.

Mistral medium is ranked #8 on lmsys arena IIRC, so it’s probably just not your style?

I’m also comparing this to gpt-5-mini, not the big boy

I’m making an app to learn multiple languages. This portion of the pipeline is about explaining everything I can determine about a work in a sentence in specifically formatted prose.

Example: https://x.com/barrelltech/status/1963684443006066772?s=46&t=...

Yes I use(d) structured output. I gave it very specific instructions and data for every paragraph, and asked it to generate paragraphs for each one using this specific format. For the formatting, I have a large portion of the system prompt detailing it exactly, with dozens of examples.

gpt-5-mini would normally use this formatting maybe once, and then just kinda do whatever it wanted for the rest of the time. It also would freestyle and put all sorts of things in the various bold and italic sections (using the language name instead of the translation was one of its favorites) that I’ve never seen mistral do in the thousands of paragraphs I’ve read. It also would fail in some other truly spectacular ways, but to go into all of them would just be bashing on gpt-5-mini.

Switched it over to mistral, and with a bit of tweaking, it’s nearly perfect (as perfect as I would expect from an LLM, which is only really 90% sufficient XD)

> Lumo is powered by open-source large language models (LLMs) which have been optimized by Proton to give you the best answer based on the model most capable of dealing with your request. The models we’re using currently are Nemo, OpenHands 32B, OLMO 2 32B, and Mistral Small 3. These run exclusively on servers Proton controls so your data is never stored on a third-party platform.

That's a bit of a double edged sword. My support goes as far as giving local offerings a try when I might not have done otherwise. But at that point they need to be able to compete on merit.

I could never get anything useful out of Lovable and was frustrated with the long editing and waiting process.

I'd prefer a site builder template with dropdowns. Lovable feels like that type of product, just with an LLM facade.

I don't hate AI, I just wasn't getting into the groove with Lovable.

Our engineer used lovable for about a day, then just cloned the repo and used cursor since it it was much more productive.

1. A general purpose LLM chat interface with high reasoning capacity (GPT-5 thinking on web is my go to for now)

2. An agent that has unrestricted token consumption running on my machine (Claude Code with Opus and Amp are my go to for now).

3. A fine-tuned, single purpose LLM like v0 that is really good at one thing, in this case at generating a specific UI component with good design aesthetics from a wireframe in a sandbox.

Everything else seems like getting the worst of all worlds.

Western Europe and the US had essentially the same level of government safety nets (and government spending and economic growth) from the 1950s to the 1990s.

Who do you think started all of the European industrial giants that are still globally competitive today? Europe had no problem competing in the industrial revolution, if these were actually European values there would be no competitive European industry like there is no competitive European tech. It's only the digital revolution that Europe has struggled with.

Government spending makes up a smaller % of Chinese GDP than in the US, so definitionally their economy is more privatized than the US economy. China is at 33%, US at 36%, Europe is at 50%.

For every Jack Ma, there's a million other Chinese businesses flourishing in every niche imaginable with very little oversight from the CCP.

A bit of a shame if they're going to launch these connectors that the model isn't suitable for them.