Mastodawn

Happy to have a paper, a performance and an 'alt.nime' contribution accepted to NIME (new interfaces for musical express) conference London !

The paper is a collaborative one, about Uzu languages, the draft is here https://codeberg.org/uzu/nime2026/src/branch/main/uzu.pdf , we'll work on it some more so feedback very welcome (although already have a lot of feedback about the typos etc)

The 'alt nime' thingie is this fun paper with @danstowell , about learning the pyramid tala:
https://codeberg.org/Algopaca/pyramid-paper/src/branch/main/pyramid-paper.pdf

The performance will be a collaboration with top drummer Matt Davies, also exploring the pyramid tala.

Anyone else going to nime?
https://nime2026.org/

#nime #nime2026

nime2026/uzu.pdf at main

nime2026

Codeberg.org

Dan Stowell 1d ago

Lots of amazing imagery at the "Woman, Life, Freedom" exhibition in Amsterdam's Resistance Museum. Recommended. #Amsterdam #iran #museum

Dan Stowell 1d ago

NomNomDan 1d ago

Visited Moychay tea house in #Amsterdam. Very nice place. My #tea was an "experimental" semi-fermented gaba red tea from Georgia... I know, lots of words. It was v tasty and calming. #nom

Dan Stowell 2d ago

David Nicholson 3d ago

RE: https://neuromatch.social/@jonny/116324676116121930

Incredible thread from @jonny doing a "code review" of the Claude leak

An Eldritch horror of Typescript consuming tokens forced into json parsed by regexes then battered around by a Pachinko game of ALLCAPS prompts orchestrating an infinite regress of Agent Tool Processes and Subprocceses.

I guess everything I know about coding is wrong and this is the completely inevitable future
🤷🤷🤷

Show thread

Dan Stowell 3d ago

@wim_v12e @jonny If there's anything likely to change my gradually-thawing attitude to LLM coding, it's this. Wow. The leading system is itself an irretrievable pile of spitball approximations to good code...?

Dan Stowell 3d ago

Show thread

jonny (good kind)5d ago

My dogs I am crying. They have a whole series of exception types that end with _I_VERIFIED_THIS_IS_NOT_CODE_OR_FILEPATHS and the docstring explains this is "to confirm you've verified the message contains no sensitive data." Like the LLM resorts to naming its variables with prompt text to remind it to not leak data while writing its code, which, of course, it ignores and prints the error directly.

Dan Stowell 3d ago

Show thread

jonny (good kind)5d ago

So the reason that Claude code is capable of outputting valid json is because if the prompt text suggests it should be JSON then it enters a special loop in the main query engine that just validates it against JSON schema (it looks like the schema just validates that something in fact and object and its keys are strings) and then feeds the data with the error message back into itself until it is valid JSON or a retry limit is reached.

This code is so eye wateringly spaghetti so I am still trying to see if this is true, but this seems to be how it not only returns json to the user, but how it handles all LLM-to-JSON, including internal output from its tools. There appears to be an unconditional hook where if the JSON output tool is present in the session config at all, then all tool calls must be followed by the "force into JSON" loop.

If that's true, that's just mind blowingly expensive

edit: please note that unless I say otherwise all evaluations here are just from my skimming through the code on my phone and have not been validated in any way that should cause you to be upset with me for impugning the good name of anthropic

edit2: this is both much worse and not as bad as i thought on first read - https://neuromatch.social/@jonny/116326861737478342

jonny (good kind) (@[email protected])

Attached: 3 images OK i can't focus on work and keep looking at this repo. So after every "subagent" runs, claude code creates *another* "agent" to check on whether the first "agent" did the thing it was supposed to. I don't know about you but i smell a bit of a problem, if you can't trust whether one "agent" with a very big fancy model did something, how in the fuck are you supposed to trust another "agent" running on the smallest crappiest model? That's not the funny part, that's obvious and fundamental to the entire show here. HOWEVER RECALL [the above JSON Schema Verification thing](https://neuromatch.social/@jonny/116325123136895805) that is unconditionally added onto the end of every round of LLM calls. the mechanism for adding that hook is... JUST FUCKING ASKING THE MODEL TO CALL THAT TOOL. second pic is registering a hook s.t. "after some stop state happens, if there isn't a message indicating that we have successfully called the JSON validation thing, prompt the model saying "you must call the json validation thing" this shit sucks so bad they can't even ***CALL THEIR OWN CODE FROM INSIDE THEIR OWN CODE.*** Look at the comment on pic 3 - "e.g. agent finished without calling structured output tool" - that's common enough that they have a whole goddamn error category for it, and the way it's handled is by just pretending the job was cancelled and nothing happened.

neurospace.live

Show thread

Dan Stowell 3d ago

@wim_v12e I'm fascinated to understand a bit more of this. If you don't have time to write it up please do boost anyone else who does a meaningful dive!

Dan Stowell 3d ago

David Lawrence Miller 3d ago

"""
We are happy to inform you that your research article, especially titled, "Bayesian views of generalized additive modelling", which is mostly aligns with our conference theme got the attention to brings our invitation. We are inviting global professionals in the field of Recycling and Waste Management, we sincerely hope you will be the part of this event.
"""

Dan Stowell 4d ago

Currently submitting a journal article with 22 co-authors. The best decision I made was to BEGIN the writing process by demanding they each fill in a form with their ORCID, email, and institution details! A simple afternoon of copy-paste ahead... (Instead of a week of pestering co-authors) #academicchatter #academicpublishing