My experimental metaverse has switched from app-password to oauth. Live at: thejanusstream.github.io/symbios-over... Next I'll focus on content and content-manipulation tools ... Teleporters should already be working, so if anyone is interested, we can link our home-regions ... webbing it up

Currently switching my atproto "metaverse" from app-password to oauth via @dollspace.gay's proto-blue crate.

AI isn't real. The data-centers are filled with left-over brains from the Adrenochrome production.

Currently implementing: - Improved World-Editor GUI and sync - Extended environmental settings - Inter- and Intra-Region teleportation At some point I will need help from people who have a sense for visual aesthetics, to go over a gazillion default-parameters.

New prototype of my p2p 3D metaverse for Bluesky is up. All you need is an app password: bsky.app/settings/app... To log in: thejanusstream.github.io/symbios-over... Every Bluesky-account automatically owns a (unique) Region. Avatar- and Home-region settings are stored in your PDS.

I'll have to publish this, even while it is still very rough around the edges, and everywhere else ... A kind of metaverse that is 1.) Based on Bluesky/atproto accounts and storage 2.) Proc Gen all the way, from terrain to foliage textures 3.) P2P mutliplayer 4.) #Rust, #Bevy, #Avian3D

How I notice pricing for Claude has gone bad: 5-hour limit is used up before I could finish my first coffee.

ProcGen p2p metaverse progress: World-editor (only available in your home region), to edit Generators (algorithm with parameters) and Placements. Saved to and loaded from your PDS.

I understand the urge to use sloppy language with LLMs, and the urge to express yourself emotionally ... but keep in mind that everything you enter into the context will have an effect on the output. Pedantically correct language with correct usage of domain-specific vocabulary goes a long way.

Going to be an interesting week ... Home-region (Your DID == Room DID) allows you to edit and save: - environment settings, - generator settings, - and placements in your PDS. Will update the public demo accordingly soon ... This will be a decentralized procgen metaverse of a kind.

I've updated my "Web3D" demo ... - Every DID is now a region - Your DID == Region DID -> "admin rights" - Every region is unique by default (hash of did is random seed) - Admin can overwrite Water-level and sun-color (stored in their PDS) Next up: generalized lexicon schema

Floats not allowed in atproto data model? Excuse you?

A lifetime ago, I spent a lot of time in virtual worlds like Second Life. Now I am building one the way I think it should be: - Decentralized - FOSS - Serverless (p2p meshed rooms) - Leaning in on procedural generation - ATProto for auth,id,db, ... - Rust+Bevy+Avian3d (WASM) as game engine

Now working on: Free Decentralized Web3D -Rust+Bevy+Avian3d (compiled to WASM) as game engine. -Atproto/Bluesky for Auth&ID -Procedural generation (parameters stored in PDS) for avatar-design and home-region content -DID-hash seeded region default designs -each region is a p2p communication room

I'd really like to test this ... so if you have a minute, that would be awesome: Proof of concept for a Bluesky-connected p2p 3D-world as website. Using Rust, Bevy, Avian3d, WebRTC p2p via Matchbox, procedural terrain & textures, ... I already got follow-up plans ... but first I need this tested.

You are cordially invited to hang out, chat, and help me test a p2p multiplayer prototype. ❤️ Create a new App Password: bsky.app/settings/app... Use it to join me (us?): thejanusstream.github.io/symbios-over... Read more about it: github.com/TheJanusStre... My Dashboard: thejanusstream.github.io

With my recent exploration of the p2p WebRTC rabbit-hole, the new side-quest-tree is: "Server-less" (p2p + ATProto auth&ID), "asset-less" (procgen) online virtual worlds via rust+bevy+wasm in a browser. The first demo-app should be up for testing soon-ish. The symbios ecosystem will grow ...

Currently working on a p2p communication layer for multiplayer bevy-stuff, with ATProto auth. Will need some testers soon. (the "oasis" reference came from Gemini ...)

Wanna know how stupid we humans are? We regularly put people at the top of military hierarchies who underestimate the danger of hubris. All throughout history hubris has caused upsets in war. Again and again and again and again and again. And naive usage of AI will not improve that situation.

Alright ... starting to gain some confidence this might turn out alright. Adversarially hardened P2P WebRTC behind Atproto auth, as Bevy Plugin. For review at github.com/TheJanusStre... Not really tested yet, so my assumption is that it doesn't actually work properly. But it might.

Come on Claude. Get your shit together. I deliberately saved some of my weekly budget for today to get some shit done ... 😡

Trying to crate secure and robust P2P WebRTC stuff feels like trying to nail Jell-O to the wall.

Use one LLM to write code and another LLM to praise you for your incredible abilities and deep knowledge. Seriously ... if your code-review prompts are not thoroughly adversarial, it is just masturbation.

Wondering if this is viable for multiplayer web-games: Identity & Lobby via ATproto Handshake via XRPC Relay Server WebRTC P2P for low latency data-transfer Opinions? Alternatives?

Admit it! This is cute. My new "dashboard" / "portfolio" SvelteKit Static Site Generation via GH Action. @ thejanusstream.github.io

I am spending too much time tinkering on tangents ... ... thus, I will publish more unfinished stuff as soon as I think it might be interesting to someone out here. Sneak peek: Morphogenetic robots (L-System bodies with NEAT-controllers) "Initialize Population" !!!

Observation on the impact of LLM usage on my own learning: I am learning less in-depth on any particular subject, because I spend less time overall thinking about it. Once the code works to my satisfaction, I move on to the next subject. But I am learning about more subjects than ever before.

Far from ready, but getting closer step by step. Learning a lot along the way. ProcGen all the things 🔥 Totally unrelated, have you thought about using LLMs for semantic crossover yet?

AI psychoanalysis is pretty ridiculous and at the same time essential to get the good stuff out of AI at the moment. Gemini tends to be "proactive". Instead of doing exactly as told, it tries to read intend to do better/more than what was asked for. For some tasks that is perfect, for others not.

It is amazing how much you can save by letting Gemini prepare the work for Claude.

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".

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.

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.

Currently adding algorithmic bark texture and algorithmic foliage to my L-System explorer.

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.)

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 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.

Gemini 3.1 Pro is currently bugged in AI Studio, it seems. Code-blocks are missing beginning and end of the code. Had to switch back to 3.0 for now.

Initially I only wanted non-flat terrains for evolutionary robots. But the mission-creep has set in and now I want algorithmic terrains with shader-based Splat using algorithmic textures.

First test with Gemini 3.1 was a failure. In AI Studio it is currently dropping the initial streaming chunks of large code blocks, for me. Repeatedly. I guess they'll fix it soon(-ish).

I did not ask for the visualization of hydraulic erosion, but apparently Claude wanted to show off. And I appreciate it.

New side-quest chain unlocked. 2 more crates and an end-user application on the way ...

Little robots evolving to go forwards while having their heads upright and being energy-efficient. Sensors include joint-angles, IMU and touch.

If AI-assisted coding causes burn-out / exhaustion, re-align your workflow. Don't focus on velocity. The AI will outrun you and keep running. Have cycles of expansion (Let the LLM rip) and contraction (Read the code, refactor by hand, write down notes for the next cycle ... human-paced activity).

Does not look like much yet ... but I am getting there.

When the ground acts more like a high-surface-tension liquid than bedrock ... 🥴 Robot = L-System body with NEAT CPPN-style brain GA = NSGA-II multi-objective to maximize displacement and being upright while minimizing energy-consumption Trying to get proper walkers is a non-trivial challenge

Now that we have some good tools and methods for generating and refining high quality code, it is time to aggressively optimize for token-efficiency again. We should try to map out the Pareto-front with both objectives in mind.

Since usage of the verb "google" for search is in decline, we should re-use it to describe building a great foundation but face-planting on integration and UX. "She is a great engineer, but she totally googled this project."

Always a hoot when your simulated creatures evolve to hack the physics-engine. Here they figured out how the get a leg stuck in the ground for locomotion glitching.