iiinova — where your digital self keeps evolving

I'm working towards publishing something peer-reviewed, so I needed solid grounding and now get to cite Husserl in German rather than in translation.

010

to the point that I ended up using glm ocr and lots of custom script to make markdown files of Husserl Ideas 1 in the original German and using those artifacts to better ground the conversation.

I would also say that I had similar feelings about learning phenomenology; I watched several YT videos and gave up on that, started working things through with Claude (informing some of the feedback I'm giving now), and eventually realized I needed more grounding for what I was after, to the point…

100

If you’re trying to dig into theoretical physics or quantum computing I can’t offer any factual feedback, but from what I do understand about quantum computing, probability mathematics would be more useful than qm if you’re not the physicist designing the machine. Quantum™ is probability and energy

That said, what I see in this thread seems hand-wavy and removed from q. mechanics, and I don’t know enough about your conversation, goal, or the scope to make any corrections. I hadn’t learned linear algebra before quantum mechanics, and haven’t applied la to qm since learning la.

Depending on your goals, it could end up being useless or embarrassing, not dangerous. I don’t disagree with your approach, just suggesting tweaks that might improve your outcome.

100

My use of the word psychosis wasn’t meant to convey a danger (to overall mental health or otherwise) but that you could end up firmly believing something is true and well reasoned, even generally accepted domain knowledge, when in fact it isn’t.

100

Shared via Claude, an AI assistant from Anthropic

I credit a really good teacher, and only had one hang up that I needed office hours for, and it was related to derivatives because I had dropped my calculus class that semester. This is a decent start to scoping QM vs. particle physics. claude.ai/share/0c5c69...

claude.ai

Claude

100

Also, a lot of big domains are lumped into “quantum”: e.g. q computing, q mechanics, particle physics, theoretical physics, etc. Mechanics is foundational and what I understand the best, but also defies most intuition though I became more intuitively comfortable with it than most in college.

So new sessions are the best way to make sure you flush out nonsense, and try not to bring any with you into the new session (which can be hard). The math should be the ground truth you return to; interrogate the equations, ask about the history leading up to the equations, ask about constants.

The training set is not clean, and long running conversations can reflect back misunderstandings that are easy to latch onto. Quantum physics is probably a better candidate for AI psychosis than consciousness (Claude, are you conscious), and probably has less instruction in the harness to steer away

240

Russ Ted Fugal @russ.fugl.dev · 21/05/2026

I think there’s been so much pseudoscience written about quantum mechanics on the internet, add to that the fact that most people whose aren’t physicists who think they understand quantum physics don’t (I might even include myself in that group), I’d recommend fresh sessions often

130

Now you have me thinking about site.standard.document records size limit. Is this specific to leaflet or broader. I’m abstracting my react blog helpers to tangled.org/russ.fugl.de..., leveraging @sequoia.pub and @markpub.at I know leaflet has a different value prompt than markdown by itself…

tangled.org

russ.fugl.dev/atproto-presskit

React blog kit backed by AT Protocol. Provides filtering UI, pagination, Bluesky comments/embeds, subscribe widget, and markdown rendering.

@xanlopez.net is right about the endgame: when we win it will just be “login with your account.” But to get to just “my account,” the meaning of “my” has to change first. @ you is the on-ramp.

@byjp.me drew a clean line: “Atmosphere” is the apps & tools everyone uses; “atproto” is for the devs. @ you is one layer above both. It’s what you tell the person who doesn’t want to learn either of those words yet.

@dholms.at set the bar: you lose if you say protocol, decentralized, repo, or pds. The cloud is “where I keep my photos when I drop my phone in a lake.” The moment you say “server” you’ve lost the thread. @ you clears that rule cleanly.

Coming back to this to make the case carefully. @ you or AT you isn’t competing with “Atmosphere Account.” “Atmosphere Account” is what the database row is called. @ you is the value proposition. 🧵

Russ Ted Fugal @russ.fugl.dev · 12/05/2026

I don’t know if I’ve ever done a full component or library or hook, but definitely sql and regex statements.

000

Russ Ted Fugal @russ.fugl.dev · 07/05/2026

when he made an indirect reference to tangled, i thought of you. i hope he takes you up on this. i'm also wondering if forgejo could become a knot server... 🤔

Russ Ted Fugal @russ.fugl.dev · 07/05/2026

after doing some research, i had drafted a plan to use forego.org as my private repo server, syncing my public repos to @tangled.org's knot1 (my original thought was to add extensions to my knot.smart-knowledge-systems.com). then i watched @t3.gg get forgejo pilled on his stream.

130

Russ Ted Fugal @russ.fugl.dev · 02/05/2026

Will be when I push, all local right now. My tokenizer already is open source. smart-knowledge-systems.com/kiri

100

Russ Ted Fugal @russ.fugl.dev · 02/05/2026

Good idea, fits well into what I'm building. Let me know if you want to work together on it.

Russ Ted Fugal @russ.fugl.dev · 28/04/2026

Which language? bsky.app/profile/russ...

000

There’s also no reason for you to read the code of those prototypes. But you can engineer an agent assisted rewrite, the agent reading the code. It may be useful to read the code to point out what definitely needs not to be replicated. But maybe that’s something the agent can handle itself.

000

That implementation is full of slop and tech debt, but still useful. Your job is to now engineer a rewrite. Prototypes and slop are not starting points any more than dry-marker on a whiteboard is a shippable product. While you're able to increase agility, no one is shipping what those agents wrote.

100

Imagine a workflow. You're whiteboarding an idea with a team, with an agent listening and updating the meeting artifacts in real time. Meanwhile, a second agent is vibe coding a sloppy implementation. Now you can show the product manager and designers real-time artifacts and prototypes.

So what I want to do is not engineer agentic workflows that seamlessly transform intent into product — I’m looking to leverage how inexpensive code generation is to be more agile without fear of technical debt. Slop only persists if you let it, and is highly adjacent to technical debt.

I will follow that up with engineering a rewrite that gets much closer scrutiny. I feel like this is highly agile. I’ve seen some discussion about how developing prompts for vibe coding is like a return to waterfall development. That level of planning is only necessary if you attempt to wish cast.

You can participate in a conversation without being 100% engaged and reading/hearing/parsing/scrutinizing everything that has been said & written. I will often generate a prototype through conversation and not necessarily read all the code.

130

Conversational generation requires a conversation and not just delegation with “accept all.” Conversational generation requires following the conversation, including the code being generated.

Though I’m not sure how much I like, the term “agentic engineering” it has has got me thinking about workflow and other ways to describe flows that I use. I don’t like wish-casting, which I guess is the Karpathian sense of vibe coding. But let’s call what I described as conversational generation.