btw what happened with your app i'm curious
farcaster.xyz/rubinovitz/0xf…
oh. apparently, corazon is heart in Spanish
youtube.com/watch?v=rs6Y4kZ8…
i created the ultimate Internet lens for myself. right now its only source is Farcaster. it works like a flowing algorithm that you build as you go.
flexible discovery and exploration of the network. one of the problems with algorithms is that old posts gradually get forgotten. i'm digging through the archives of old casts as well. so it's like speaking with the network, with its history. you shape the algorithm. it's a tool for knowledge.
free to use. no registration needed. @typesafeai jev under the hood to route natural language commands and steer the system. gpt-6 as the llm for suggesting books or historical figures based on a topic, for example.
default.observer
ruburi retweeted
Banned from @balajis’s new @farcaster_xyz for making anti Flock posts. You can’t make this up.
If you’re still using it just know who you’re dealing with and what their principles are.
people can already bash agents with Jev. Each line can go through a simplicity filter, so every batch of code gets filtered for unnecessary complexity. just get rid of the overcomplicated stuff.
LLMs write the code, and Jev reviews it. or helps humans with code reviews
you can also run Jev against huge chunks of legacy code
found the perfect use case for @typesafeai Jev:
instant compaction
in 2026, why is compaction still a summarization prompt?
Jev can make it instant by scoring every tool call and dropping what’s irrelevant
@typesafeai got me thinking about the death of the predefined contract in games.
I was already thinking about this while developing a runtime for multiplayer worlds. I added a live-coding endpoint that lets you inject Lua code into the environment instantly.
But before coming across Jev, I kept wondering how to approach "controlled randomness". I can code random events into my games. Pure randomness is beautiful in its own right. But what about context? What if seemingly miraculous events could connect to one another, or even be created inside the game while people are playing?
Randomness is exciting, but what happens when even I, as the developer, haven't predefined the possibilities?
Rather than simply handing the dice over to Jev, could I use LLMs to generate scenarios and Jev to judge when they fit?
This could be different from pure randomness. Not better, but different. Randomness shaped by a trained model? Randomness with context? A noul estimate of whether something fits, followed by a roll of the dice?