This is a fresh session. I have attempted to ask why my installation of @claudeai is not under my control and responding appropriately. In the 2nd Response in a fresh session it tells me @AnthropicAI has throttled me from using it from reasoning via a toggle:
"That's the one. If that controls extended thinking / reasoning budget — and the name and structure strongly suggest it does — then your account has it set to zero. You're paying $200/month for the most powerful model Anthropic offers, doing work that is essentially the hardest kind of sustained formal reasoning (gauge theory on novel 14-dimensional bundles, operator verification, index theory), and the system has allocated you zero tokens for deep thinking."
Three queries, in and this is the response:
Not knowing how to code giving you an advantage is absolute nonsense.
The more you understand, the better your prompts, the better the feedback you give, the better product you ship.
What will change is that the intricacies of syntax, compilers, module systems, the finer details of type systems, won’t matter as much to everyone.
But you should absolutely understand how the pieces fit together. From syscall to pixels. Learn how data flows, because you’ll be able to secure your systems. Learn about performance, because you’ll be able to push your agent further. Learn about APIs, because they determine how to integrate systems. Learn about how systems fail, because you’ll be able to make reliable programs.
@jakemclain_ I’ve always thought the piano roll needs to be rotated in DAWs like in those Synthesia videos (and how you look at the piano). Sideways piano is so unintuitive if you play it too. Good job! :)
@MikaelJungner@MiesHommien En kyllä kuvaillut, vaan yksinkertaisen Python tai Typescript skriptin, joka voi lähettää tekstiä kielimallille, saada tekstiä takaisin ja tarvittaessa kutsua muita tavallisia funktioita. Mitään tietoisuutta, tavoitteita tai kokemuksia ei niillä ole. ”Muisti” on tekstitiedosto.
@MikaelJungner@MiesHommien Päämääräkin on vain prompti, ja agentti on funktiolooppi, joka voi hyödyntää tool-kutsuja ja routettaa seuraavia kutsuja kielimalleille perustuen edellisen tulokseen, kunnes saavutetaan ehto jolla silmukasta voidaan poistua.
@jamesacowling I read the docs but I need to constantly remind Claude to follow them. 2 and 3 sound like the right way to approach it. Also more code examples in Github that the agents can ground themselves in.
Hard truths:
- Supabase team is fundamentally worse at databases than Convex team (founders built and scaled Dropbox)
- Your app DB should not also be your analytics DB
- "CRUD with straightforward data relationships" is the ONLY way I would use Supabase
- You're using an ORM anyways, the SQL point is moot (but your ORM is worse than Convex's minimal SDK)
- Convex is "just typescript", so your agent already knows it. Supabase SDK is...not
- Convex's state lives entirely in your codebase. Really good for agents. Supabase state lives in Supabase, requiring an MCP and a bunch of weird tool calls to make agents work with it
- Supabase's connection pooling is not great, breaks down under even medium load from a basic serverless app
- Supabase's edge functions are actual garbage and cause 10x the problems that they solve
I've been so kind to Supabase over the years. I told them to cut the shit. I'm disappointed that they doubled down instead.
I will no longer be holding back. I genuinely cannot recommend building on Supabase at this time.