can i talk to an openclaw bot using internet relay chat?

[ariadne@social.treehouse.systems]

i just wanted to put an openclaw on irc as a fucking shitpost man

[ariadne@social.treehouse.systems]

and you tell me people legitimately are using this software.

how?

is it really magically better when you hook up claude?

[ariadne@social.treehouse.systems]

(don't worry, i am running this in a MicroVM under kubernetes, I wouldn't dare give it access to anything I care about.)

[jfkimmes@social.tinycyber.space]

@ariadne what model did you finetune on? For a 1B model you need something really specialized on tool calling.

[ariadne@social.treehouse.systems]

@jfkimmes i built an LLM from scratch with transformers kinda loosely following the scripts the qwen people released

the LLM is basically trained on ~30ish GB of mostly furry smut and public Linux IRC logs.

*nods sagely*

[srazkvt@tech.lgbt]

@ariadne @jfkimmes can we uhm, inspect, the furry smut ?

[ariadne@social.treehouse.systems]

@jfkimmes i am, however, using the 35b parameter qwen3.5 reasoning model for the "thinking" portion of this exercise

[ariadne@social.treehouse.systems]

@SRAZKVT @jfkimmes i am keeping my typefucking logs to myself, thanks

[ariadne@social.treehouse.systems]

i wonder if the problem is that the model i trained is too shit to do anything other than really bad ERP

[hayley@social.applied-langua.ge]

@raulinbonn @ska @ariadne I made a talking clock in gmod which would cuss out the user

[albertcardona@mathstodon.xyz]

@ariadne

The key is to realise that the average is so low – we can't all be experts at everything, so we are bad at most things – that a model performing slightly above average at one of the tasks we aren't good at means a majority of users will perceive its outcomes as positively better than what they could do themselves.

To any expert, the model falls very short, as it performs well below its own ability.

[jfkimmes@social.tinycyber.space]

@ariadne Oh, is that a OpenClaw specific feature where you can specify that reasoning traces are generated by a separate model than the actual response? I'm not really familiar with OpenClaw's internals.

[jfkimmes@social.tinycyber.space]

@ariadne In any case: as long as the final response is generated by your trained model it will never make a valid tool call since there are probably about zero training examples of the necessary JSON structure required by the tool handling in your furry smut (this is an estimate that could be quite the way off knowing the furry community but still)

[ariadne@social.treehouse.systems]

@jfkimmes yes, you can have it use a different model for planning.

[ariadne@social.treehouse.systems]

@jfkimmes this does explain something: it seems to be able to invoke tools when it is planning, but then those tools do not get invoked in the final step.

so it uses tools to read files when planning, then fails to use tools when executing.

what a fascinating conundrum.