Stop hardcoding one model name in your code. Now you can give each request a policy: a small rule for what the call needs It picks the right model for the job, on your own keys This is unhardcoded, our new open source routing for AI models, live today

Sakana's Fugu routes models behind one API They're right: AI is many models now, not one winner. But it's a closed conductor - you trust whatever it picks. The question isn't which model is best It's who decides which one runs

With unhardcoded, every decision is out in the open You see which models were considered, which one ran, which were rejected, and what each one cost Every routing is replayable from its fingerprint. Nothing happens in the dark

One policy handles one call. But real features are never just one call A workflow wires several policies into a full job: one step reasons, another writes code, a third merges them into a single answer Each step picks its own model, and the whole run leaves one trace

We built unhardcoded for ourselves first, running real traffic on it Now it's open source and yours to run An open project by GenLayer Labs, the team behind GenLayer Try it now 👉

curious @levie what do you think on this?

Congrats bros

Finally, routing that’s actually transparent instead of another black box. Seeing every model considered + rejected + cost per call is the kind of visibility we needed.

Every decision is done open with unhardcoded

@konigthegreit This is how you actually ship reliable agentic stuff without brittle single-model lock-in.

It looks quite interesting and promising

Well done, looks interesting 🔥

@kstellana Yes that's why you should try @zeitgeist__AI Not only one model 👀powered by genlayer

The interesting part isn't another AI model. It's moving model selection out of code and into policy. Models, pricing, and providers change constantly. Hardcoding those decisions never felt scalable. @GenLayer @RallyOnChain

Definitely gonna give this a try 💯

LFGen🦾😎😍

This looks quite amazing!

this is good

Sounds cool! Can’t wait to try it out 🚀

This make sense For genlayer usecase too I mean for some contract you get timed out when the wrong model validates your work

I see synergy here guys, you have the routing, we have the inference at a huge discount.

Have good time by genlayer testnet in live

Smart routing saves so much time.

Ditching hardcoding model names is a game-changer for flexibility! Giving each request its own policy helps ensure the right model picks up where you left off, saving future you time and sanity. Nice move!

The question "why does this still live in code" is the one that stings because the honest answer is usually "because nobody scheduled time to revisit it." Unhardcoded makes revisiting unnecessary by making the decision continuous.

Definitely a cook

Looking great

Per-request routing with the policy on your own keys is the right primitive — that control layer is what matters once you're running inference at scale, not just swapping model endpoints. Good ship.

Amazing tech indeed

Amazing genm

Dynamic routing saves so much time.

终于不用改代码改到哭了

A new way forward. Another integration shipped successfully

That sounds useful for swapping models in the future.

Unhardcoded code once, route smart forever your AI finally picks the right tool for the job.

Lfg thanks for this Genm

This is pretty impressive

Let's go Genlayer