At least they must give you unlimited model usage π
Posts by Francesco
From AI-and-I to I-and-I the gap is small
What can go wrong π?
/superpowers:brainstorm
ralph-loop "find issues in the plan"
/create-beads
ralph-loop "keep context window under control and implement all beads"
Productivity at its peak, burning tokens like there is no tomorrow,
The quickest way to do something is taking a shortcut, it will compound the speed at which you will win or lose.
What makes the difference now is that the builder's vision can be so high level that you don't interact with the code anymore.
But the devil remains in the details
Opus 4.5 in ClaudeCode is quite smart, really happy of the work he has delivered today. Very good boy!
Saturday morning, and here I am building an Astro web app with a Supabase backend
The new #pixel smartphone is awesome, everything is adaptive π
the @supabase.com approach definitely gives mileage.
Really enjoying how low level you can get on Postgres but at the same time all the convenience methods to quickly access the data from the client, plus the rest of the awesome Supabase extras.
Good artists copy, great artist steal... ironically this is pretty relevant to how LLM models are trained, using any available data
I built a MCP server that can wrap any electron app, published it on GitHub, was expecting it would get some hype, but so far zero starsπ
The daily life of a clearly mentally unwell person; trapped in a permanent spiral of needing to get closer to the hardware
Advent of code is so cool!
But still a lot of ppl like them, crazy world π
I see you
Man does it suck to set up a regex in C
It's only good at solving problems which have been discussed throughly on the internet (therefore encountered during the ML training phase).
It would be a much better product if it could admit not knowing something rather than hallucinating.
What do you mean by generated in a loop?
Webcodecs are not the friendliest web api
π today I received my Bsky invitation, this is literally the first post I read