27B is already really good at coding-specific tasks. Fundamentally, there is little innovation on the core architecture: LLMs are all designed essentially the same, with minor differences in how they are trained. They are all feed-forward multi-headed attention models; it doesn't matter if it's a 4B model or a 1T model, that's just scale.

Further, the frontier models cannot afford to innovate: they have to scale as quickly as possible to "beat out" their competition. The frontier models fundamentally will not create the next "attention is all you need" monumental jump in AI.

Frontier companies are stuck on scale with zero capacity to innovate. You cannot point capitalism at "basic science research" and expect any ROI. This is a known reality. Innovation is much more indirect and a "random walk" style of knowledge acquisition.

Finally, these LLMs are quite literally designed with a human-in-the-loop, and we do not give ourselves enough credit for how well we ourselves tool-call. We are doing a lot of heavy lifting to make these models useful and you cannot simply remove us from the equation without also removing ourselves from the training pipieline.

These types of features are not worth it and need to be removed from the marketplace.

These models are a race to the bottom just like compute.

With unlimited tokens make it a lint rule or auto formatter.

A lack of structured output has never been a blocker for agents to work, that’s a traditional coding problem.

“Write good help text and error messages” is just good design which is self evident.

What originally got me excited to build TUIs was the concept of delivering apps over the wire via SSH. SSH apps resemble a browser in that way: no local installs required.

It's a major reason why I enjoy hacking on https://pico.sh -- deploying the TUI requires zero user involvement.

It’s always gonna be a multi shot process. And it can already write code good enough. That’s no longer the bottleneck.

Further, Qwen 27b is such an incredible masterpiece for coding and it can run on consumer hardware today. Anthropic/OpenAI are gonna give up on coding models very soon. There’s not gonna be any money in it when you can run your own local model for significantly cheaper.

Qwen27b is not SOTA but the value is insane. You can basically use it for small tasks and then route harder problems to opus or sonnet and boom you’ve said a lot of money.

The use case is different but this article strikes some vague similarities around an agent API to remotely execute commands.

They are still a WIP but it’s on our roadmap to continue to improve.