The problem with being a pragmatic LLM user is that you have on one side corporate America shoe-horning the tech in mediocre products none wants, and on the other side a large portion of the internet who loathe it but don’t use it and don’t even know what it does. Those conversations never go anywhere man. You’re talking to someone who thinks accuracy of 57% on SpreadsheetBench means the model gives wrong answers 42% of the time.
Hate to agree with Microsoft but yeah, Excel is probably a great place to introduce an LLM. It’s in that sweet spot between natural language and light programming, in an environment with math baked in so you don’t really care about the model’s accuracy or exact recall. All the data is here, and the model only has to manipulates cell numbers and writes formulas in this dumbed down language.
I’m sure you can get away with pretty small models too. It doesn’t need super human knowledge to implement 90% of common Excel use cases, and i suspect in real world scenarios the accuracy must be pretty interesting.
Supabase unironically do exactly that. You’ve got your client, you login through OAuth, then use your JWT to connect directly to postgres.
Your JWT contains your user id, which is used with row level security rules to determine which rows you can and cannot access. It’s pretty amazing what you can do with PG alone. The tooling is not quite there yet but that’s probably where we’re headed.