bits per weight

> If anything, I want these models to be less persistent at their focus of completing their goal, and instead just call defeat and say “I’m not sure how to proceed next”.

that might end up like the older gemini models which frequently gave up and called itself a failure.

Gemini still gives up too easily

will be releasing weights per their tweet announcing the model https://xcancel.com/Alibaba_Qwen/status/2078759124914098291

repo appears to be private, getting a 404.

I'm just stupid. Sorry about that. Made it public now - https://github.com/Simbastack-hq/sentinel

i read around launch that anthropic will fallback to opus if fable is used for frontier LLM development. did you run into anything like that?

Strangely, I did not. I was expecting it and looking out for it the whole time. At least I never say a warning!

that's because they said they'll silently fallback in case of model training. Then they backtracked, and said they wont silently fallback, but how will anyone ever know?

large companies aren’t buying subscription plans. my org has a 2k per month token budget per person and starting to explore optimizations like automatic model routing.

Mind sharing a link to the source code? i couldn’t find it on the website.

thanks

Github icon in the header on the right.

i use this for my personal projects. some features are gated behind a license but the basics like provider proxy, logs, metrics are covered in the free version. https://github.com/maximhq/bifrost

can you explain what you mean by symbolic recursion tricks in this context?

The model can call a copy of itself as a tool (i.e., we maintain actual stack frames in the hosting layer). Explicit tools are made available: Call(prompt) & Return(result).

The user's conversation happens at level 0. Any actual tool use is only permitted at stack depths > 0. When the model calls the Return tool at stack depth 0 we end that logical turn of conversation and the argument to the tool is presented to the user. The user can then continue the conversation if desired with all prior top level conversation available in-scope.

It's effectively the exact same experience as ChatGPT, but each time the user types a message an entire depth-first search process kicks off that can take several minutes to complete each time.

How is this different from a standard tool-call agentic loop, or subagents?

Each stack frame has its own isolated context. This pushes the token pressure down the stack. The top level conversation can go on for days in this arrangement. There is no need for summarization or other tricks.

Is this related to the paper on Recursive Language Models? I remember it mentioned something similar about "symbolic recursion", but the way you describe it makes it sound too simple, why is there an entire paper about it?

The RLM paper did inspire me to try it. This is where the term comes from. "Symbolic" should be taken to mean "deterministic" or "out of band" in this context. A lot of other recursive LLM schemes rely on the recursion being in the token stream (i.e.. "make believe you have a call stack and work through this problem recursively"). Clearly this pales in comparison to actual recursion with a real stack.

This is just subagents.

i don’t think it’s “eat more, weigh less”. they are appetite suppressants, so you’ll eat less, weigh less.