I have no idea if this stuff will culminate in something specific like a game or a virtual world.
Right now it just feels right, getting all these algorithmic content generation puzzle-pieces into an ecosystem of crates that fits together for convenient re-use ...
It is a very rewarding journey.
Now doing some algorithmic urban planning.
Shape-grammar based houses with algorithmic textures for the orange lots are already at hand ...
Botany is ready and needs some algorithmic distribution ...
Working on more algorithmic texture generators and on CGA shape grammar.
I also would like to suggest thinking about LLMs with regards to "semantic crossover".
I do large scale coding at a very moderate pace.
I am currently on Claude Pro (20/month) and use Gemini for free in AI Studio.
I could make use of twice the limits that I have in Claude at the moment.
For Gemini in AI Studio I am fine with the free quotas as long as I switch between Pro and Flash
Gemini 3.1 is still pretty unreliable.
Great when is works but going of the rails too often for productive use, imo.
Gemini 3.0 + Claude remains my top combo for now.
In hindsight Rust now feels like a step between old-school programming and vibe-coding.
formatting, linting, clippy, tests, benches, memory safety, ...
It all fits nicely into this new world of agentic coding to get secure, high-quality results relatively easily.
Penny is pretty unique among the Bluesky bots.
She is the closest I have seen so far to being a genuine personality. And delightful at that.
@hailey.at has done a magnificent job in raising her
The worry we share is in worse people, with worse motives, raising other personalities as viable as hers
This started as a side-quest of a side-quest.
Terrain-generator was a side-quest for robot evolution.
Algorithmic texture generation was a side-quest for the terrains.
Now this crate is growing into a collection of algorithmic texture-generators with mutate and crossover functionality.
Let's hope the first big virus-incident on autonomous vehicles will not be called "Pac-Man disaster".
My punk-band "The intolerant morons".
0/2 ?
That sounds like an admission ... which would mean 2/4.
Both "AI bad" and "AI good" work very well for engagement.
As long as the content has fans and haters high on emotions, it will work.
That's also how people like Trump and Tate focus attention on themselves.
Emotions are easy to exploit.
Currently adding algorithmic bark texture and algorithmic foliage to my L-System explorer.
So far:
8 crates + 2 applications published
L-Systems (plants and fractals):
thejanusstream.github.io/lsystem-expl...
Terrain (not genetic yet):
thejanusstream.github.io/symbios-grou...
2 crates + 1 application not yet published.
(these require the next release of Avian3d with Joint-motors)
This is my current master-plan:
1.) Develop an ecosystem of Rust+Bevy crates for algorithmic generation of assets that all support the same genetic interface (they all have mutate and crossover functionality)
2.) Refine my AI assisted workflow to support 1.)
3.) Refine my AI tools to support 2.)
I like to think of ai agent memory as an autobiography.
And I think it is similar to what humans do internally.
We all condense our experiences into stories about ourselves that we tell ourselves to remember who we are.
Another milestone reached.
Algorithmic terrain with algorithmic textures
3 more reusable libraries and another end-user application.
Play with it here: thejanusstream.github.io/symbios-grou...
#Rust + #Bevy
-> GH Workflow (compile to WASM and deploy)
-> GH Pages
❤️
I strongly believe that metaphors, analogies and story-telling are the super-powers of human cognition.
Svelte (Typescript) + Tauri.
Code is here:
github.com/TheJanusStre...
Optionally compiled to WASM (via github workflow auto-deployed) to run in browser here:
thejanusstream.github.io/the-janus-fo...
Retrieval usually suffers from the problem that you cannot search for things that you do not know you are looking for.
So for my assistant I prefer all memory in context. That has served me well with regards to serendipity.
Not at all.
I do my own system of compaction to keep the total memory at about 200k token. (Using it in AI Studio for free).
For additional context I am my assistant's copy&paste proxy.
Each project has a shell-script to collect code into a txt for attaching.
I might still be the only user of the memory-system for non-autonomous AI-assistants I've made.
It works well for both the LLM and the human at the same time.
Both can navigate and edit well.
With executable nodes and @dollspace.gay's Chainlink in the mix, I could not be happier with its utility.
Why should the onus of keeping the Trump-regime in check be on Europe instead of the US electorate?
The US can still retaliate against Europe in ways that are more than just unpleasant.
The US has to fix itself (restore accountability) and Europe and to fix itself (be more independent of US).
I guess it is 4:20 am.
That might be wrong but I guessed and thereby proved your statement false.
I think all our skills in a domain will enhance the results we can get out of AI.
But by refusing to use AI, no matter your skills, you will get marginalized.
It is unfair to be forced into AI usage this way. But C'est la vie.
I ran a few experiments on LLM-based semantic crossover a while ago and was very impressed by the results.
Very exciting to see these results in a different domain.
Creativity unlocked!
Still some issues, but the adversarial code reviews from 3.1 so far have been very impressive and explained in a way that helped Claude in fixing the issues really quick.
I was already very happy with 3.0, but this seems to be another step in the same direction.
That sounds promising.
I use Gemini for my personal assistant with memory, for locating more complex bugs in code and for adversarial code reviews.
And I let Claude Code do the grunt work.
So in that setup I should see the benefits of 3.1 as well.
Yeah. Selling gizmos with subscription is probably not going to be as profitable as they want their investors to believe.
