I'm using luna (the smallest), at low thinking in my 9-to-5 job and I'm quite happy. No groundbreaking tasks so far, but typical small jira issues and fixes are done in a matter of low minutes. Very fast loops have their pros.
The concept of them being more than compiler bookkeeping, but as propositions about program behavior and invariant encoding is more than 50 years old at this point.
> I've been around many conversations over the years where people would say goofy things like they couldn't use Python because it's untyped.
I'm one of those, but what I mean is that there are no static types you (and the compiler) can reason about.
Moreover there's a difference between types as compiler bookkeeping, as in Fortran or C, and types as propositions about program behavior like in ML, which highly influenced modern type systems, including TypeScript, albeit Ocaml, F# and the Rust type system belong to the ML family.
I learned that outside of tech, Gemini is widely used in enterprise.
E.g. in the insurance company where my SO works, the major tasks are writing Gemini "gems" (some kind of prompts I think) and NotebookLM is a killer product for e.g. collecting and summarizing new laws, cross checking documents and what internal regulations are.
I then learned it's used in a chemistry consultancy company of a friend of mine to process reports. Flash and Pro models are also wildly popular in another European bank I know people in to assist in customer care (pre processing tickets before handing them to humans), translations, reporting, etc.
Google suite is already at the core of many businesses and Google easily adds these offerings without new contracting being needed.
Don't confuse our bubble with the real world. You can have a disaster product like teams and still dominate enterprise because you were already there with excel, outlook and SharePoint.
> You use the previous gen model to prepare datasets for the next model iteration
I've read multiple times that this approach is harmful in training.
You're essentially describing what many call distillation, but it's only useful in post training to guide behavior, it teaches how to behave, not how to think.
I might be wrong though and would be glad if someone more knowledgeable provided more insights.
It's generally a major downgrade in acting like an assistant.
I don't know what's wrong but it is just bad at multi turn discourse even on a limited amount of content with no MCP or bash calls of any sort.
The thing that makes me mad is how stubbornly confident it is even whets wrong.
I have to tell it many times to actually re read the conversation as it even insists I said something else.
It's like it had a scratchpad where it has some summarized bullet points which it fills of made up content.
I'm so confused. On one side I like to connect it to honeycomb/otel logs and I can see it figures out difficult bugs in the code better than other models.
On some others I feel I'm assisting at a continuos disaster and consistent degradation since Opus 4.6, it's a tragedy.
I'm more and more the assistant to a capable, yet confidently stubborn and wrong LLM.
I've attended courses from some of the best researchers on the planet (like Graetzel at EPFL) and you did yourself a favor if you skipped the confused ramblings and just studied on the books.
Plenty of courses taught by brilliant individuals that were just bad at teaching or borderline not prepared.
Some courses (like biochemistry) were effectively useless as de facto you had to memorize 600 pages of Lehninger's book anyway. There's nothing to understand in the Krebs cycle.
I also vividly remember exams like advanced algebra were the professor genuinely did nothing but rewrite canned content on a board and could not really shed light on anything, you were on your own.
Mostly interested in functional programming and effect systems.
contact me at
enricopolanski at gmail dot com