Sabitlenmiş Tweet
Today we’re open sourcing Vilano Runtime 0.1.
It’s a durable runtime for building agent systems.
The core idea is simple:
the code reruns
the state does not
runtime.vilano.ai
github.com/vilano-ai/runt…
English
vilano ai
15 posts
Built on Runtime:
- durable state
- long-running workflows
- fanout across the bracket
- replayable decisions
Exactly the kind of system Runtime is meant for.
runtime.vilano.ai
Sam.@f1rstchair
I built an AI selection committee that argued its way through March Madness. Not one prompt. A real process: - team dossiers - matchup hearings - red-team rebuttals - challenger brackets - final selection - full replay Built on @vilano_ai Runtime.
English
Today we’re open sourcing Vilano Runtime 0.1.
It’s a durable runtime for building agent systems.
The core idea is simple:
the code reruns
the state does not
runtime.vilano.ai
github.com/vilano-ai/runt…
English
This is the gap between agent demos and an agent runtime. Behavior becomes a primitive, not a prompt or a template.
Sam.@f1rstchair
One underrated benefit of building a tiny DSL around an agent runtime (@vilano_ai), you can encode behavior.
English
vilano ai retweetledi
vilano ai retweetledi
I stopped doing manual fundraising research.
I had my agent runtime (Wilbur) do the first pass for @vilano_ai
English
Quietly built. Loudly shipping soon.
Sam.@f1rstchair
Advisory convo this week w/ someone who’s seen a lot of real companies get built. Showed him Wilbur + Corpus (@vilano_ai) Him: “Super powerful, how far along?” Me: “Core’s basically done.” Him: “Dude… most people raise way before this.” Whoops 😅
English
We’re not trying to build “the smartest agent.” We’re building the system that makes agents compounding: persistent context, reproducible runs, and orchestration that doesn’t collapse under complexity.
Sam.@f1rstchair
Hot take: coding agents are nearing their limit. The next leap of capability is not at this layer. @vilano_ai is building Wilbur and Corpus to fix this. Even in their infancy, they’ve made me far for capable than I could have imagined and we haven’t even raised money yet.
English
@thdxr It would be interesting if there were a view where you could see the changes in terms of the logic
I don't know if that would be helpful at all, might vary by AST.
But if you change every instance of x to y in a code base it's kind of just one change.
English
how are we still using swe bench to evaluate models on coding ability?
only python with half of it being django is crazy
x.com/qtnx_/status/1…
Q@qtnx_
reminder that almost half of swe-bench verified is only django
English
this isn’t cheating, it’s the syllabus
Sam.@f1rstchair
My dev setup feels like cheating. Ephemeral envs spin up per branch, tied to Git as the source of truth, artifacts in S3, and AI agents ship the PRs with me in the loop.
English