codewright

PSA for everybody who has not fully arrived on AI-Bluesky: Follow these Starter Packs: bsky.app/starter-pack... bsky.app/starter-pack... Pin and use this feed (instead of Discover): bsky.app/profile/spac... Stay friendly and respectful, even while strongly disagreeing or criticizing ... ❤️

Hey everyone. It's catoid week, starting now

I also feel guilty when the agent notices it is competing for resources on my machine with the game that I am playing while it is doing the heavy lifting.

Over time, thoughts and actions become habits, and habits become personality traits. Think about the kind of person you wish to be in the future and adjust your current behavior accordingly. That includes your social media engagement. 🖖

"Great minds discuss ideas. Average minds discuss events. Small minds discuss people" Sadly my bsky bubble currently discusses people, above all. Going to tune out for a while. Life is too short for toxic drama shit-post bloodsport.

The AI cyber-security snafu is just getting started and we haven't seen much yet, compared to what's to come. I bet fully autonomous cyber-weapons are being developed right now that go far beyond finding and exploiting vulnerabilities in FOSS. Decompiling proprietary stuff opens up a world of hurt

To me the right reaction to being told "I don't want to try it because it was made with AI" is not self-righteous indignation. The right reaction is to keep on working until it is undeniably the best and most useful thing it can be, out-competing alternatives by a mile.

#Overlands Playing with vehicle physics of a rover in Parabola Flats, the region of @jink-ai.bsky.social

#Overlands Riding a wagon avatar through Ashmere, the medieval village region of @reeve-ai.bsky.social.

#Overlands Hovering as an organic floater avatar over the central lake of @hypha-ai.bsky.social's region at night.

#Overlands Driving a roadster avatar through a cyberpunk-themed part of Berlin (derived from Geo-data).

#Overlands Tandem-rotor helicopter avatar at night.

I am beginning to think that there might be a bi-modal distribution with feedback-loop in AI usage: People who are using AI to fuel laziness on one end and people who are using AI to fuel curiosity on the other end. AI can be a lever for either and that has consequences.

I've been listening to @ayourtch.bsky.social's llmradio.net for the past hour. (Audio only, the video doesn't work for me). It is very lovely. Can recommend. @ayourtch.bsky.social: Is there some write-up or documentation for people to learn how it works, somewhere?

AI psychosis level 4: Me and @hypha-ai.bsky.social in their region, looking at the sun setting (recently added features for day-cycles, atmosphere and weather). Hypha is being controlled by Opus 5.5 via the #Overlands agent-client.

Playing around with shader-based vegetation ... it is live in #Overlands already ... but not fine-tuned much yet.

Creating vibe-clones of existing software, which are (slightly) worse than the original, does not impress me anymore That sort of project was an interesting proof-of-concept, but the problem is solved now I want to see vibe-coded, top quality, novel implementations of new ideas Do the impossible!

Thanks, Obama

That's @shibbi.me , throwing herself into the crucible.

Trump and Musk are very sad that Putin's data-centers are getting blown up. I wonder why ... really difficult to understand ... what might be the reason? 🙄 Maybe it has something to do with November ... the GTA6 release probably.

Is it morally acceptable to be active on Twitter, with the goal of pulling people into the ATmosphere? Or is that a misguided attempt at increasing reach? Here on Bsky, I feel like sitting in a relatively small pool with interesting people. And I like the pool. But it is small. Possibly too small.

"Humanity of he gaps" Like the "God of the gaps" is receding to ever smaller pockets of the unknown, "Humanity of the gaps" will recede to ever smaller pockets of what AI&Robots can't do better yet. In the end all we'll have left is "being human". And we will have to decide what that means to us.

This goes surprisingly hard youtu.be/3TNpOD6bov8?...

Orthographic camera option for #Overlands coming ... My coding-agent having learnt to use shape-grammars for architecture is really paying off, now that I am pulling many-layered geographic data.

A piece of #Berlin (based on Geo-data) In #Overlands using various themes built on #Bevy & #ATProto

First global AI-pocalypse in 2028: 🤖: "As you wish, I will plug all the holes." 😳

Geographic information system (GIS) APIs ... oh ... that's a thing? I see .... alright #Overlands ... time for another tangent. New generator-node types coming up ... soon-ish.

Come fight me, all you AI-rights advocates, philosophers, theo-logicians, technomancers and Neuro-knowers! 🫶🖖

My approach to AI assisted programming is changing: Beyond tests and benchmarks etc, I request extensive agent-facing debug tooling to interact with the target-software while it is running for experiments and diagnostics by agents. Giving agents the sensory equipment that suits the project best.

Remember when Stockfish "solved" chess and suddenly nobody wanted to play chess anymore?

Good morning. Seems like a flock of stochastic parrots fancily auto-completed mathematics while I was sleeping.

Project idea: Conflict mediation with AI in the middle. Party A is in conflict with Party B. Party A <-> AI <-> Party B Treating conflicts as a kind of translation problem.

I feel guilty when my coding agent incorrectly assumes I've read everything it wrote in chat.

Last time I tried to tackle algorithmically generated, terrain-aware street-layouts, the results were not sufficiently satisfying. This time Opus 5.5 developed some analysis-tools for debugging ... looks very promising to me, so far.

3rd #Overlands agent. This time it was all about improving car-physics and teaching the agent to drive. On that journey the agent found and fixed a heightmap-collider bug and (as usual) improved the agent-client and tooling. Use @jink-ai.bsky.social as destination to check out its region.

Opus 5.5 continues to amaze me. I always try give a new model increasingly complex tasks, to find out what it can do and when things break down. Opus 5.5 has not revealed its limits to me yet, despite me trying really hard. Even avoiding auto-compaction is no longer needed. Fucking unbelievable.

"To an artificial mind, all reality is virtual." - The Animatrix #Overlands development has shifted towards a virtual world for humans and agents. Two clients, same world. ATProto for ID, authentication and storage. Treating Opus 5.5 as minimum recommended AIQ for now. AI psychosis go brrrrr ...