Next week I'll be giving a talk at EPC on how we create the world of @PlayPaxDei .Hope to see you there, tickets still available!.

Download from

It does not always result in working code, but at least you can quickly get something to improve on. This uses ChatGPT 4o, and there is a hidden parameter where you should put your OpenAI API key. (I put in a default key with 20€, so with some luck you can test it without one:)

Vibe Wrangles. Thinking of how to combine LLMs with @sidefx #Houdini . Here's LLM-enabled variants for Attribute Wrangle SOP and Python SOP. The difference is a string input where you can describe what you want the Wrangle to do, and it will initialize the code field

FWIW;

@sidahuj Claude is clearly worse at writing python for Houdini as for Blender, most times the custom code fails.

This was quite cool so I had to make a Houdini-MCP as well. Mostly a conversion of @sidahuj's one, but with the addition of Claude being able to get an OpenGL render to enable it to see what it's doing. Not really useful, but a fun gimmick and maybe a sign of things to come.

It feels like the new millennium is starting now -- it just didn't hit the nice round number.

I first tried with a local Deepseek-r1:32b, but kept getting errors with 85% of even the simplest wishes. GPT-4o makes valid code maybe a bit over 50% of the time. With a niche API like HOM these models just keep hallucinating. A finetune with Houdini documentation and examples?.

Don't worry, our jobs are safe, this can't really do anything of actual use.

A bit of LLM fun in @sidefx Houdini, less than one cent a wish ;)

I started a Houdini Starter Pack in bluesky. It's essentially a list of Houdini-related accounts you can start following - if you want to be included, ping me. I'll be both here and there for now.

Fascinating how easily replicators emerge from random code. A very inspirational listen!.

First chance to openly test the peaceful bits of our upcoming MMO. Still heavily in progress, but fun can be had already =).

If anyone asks me for career advice, I would recommend going for the infinite work.

FINITE WORK is e.g. accounting. There is only a finite set of work that needs to be done each year,.and 10x accountants will make others redundant. INFINITE WORK is e.g. research. You can always understand systems better, a 10x researcher just means better meds or faster chips.

How to think on AI and unemployment. There are conflicting views on AI taking our jobs vs AI making everyone better. I think categorizing makes this clearer. There is:.FINITE WORK, where AI will mean loss of jobs, and.INFINITE WORK, where AI will make everything better. 1/.

This announcement video has a bit more info than just the pretty pictures in the trailer. Still quite vague though ;).

This is how I've spent my time lately, doing the world building pipeline.

Want to come work with me, creating the world for a visionary MMORPG?.