The Borg Are Here!

Last Sunday I had a few hours free. I sat down, opened up an agent window, and within a couple of hours I had a fully working Flask app with authentication, encrypted credential storage, database models, error handling, a clean frontend, and a Docker image published to Docker Hub.

A couple of hours. Done.

The app is called WebUI Manager - a small self-hosted dashboard for keeping track of all the services running across your servers. It exists because I got tired of maintaining a spreadsheet for this, and I was also too often full flashbanged by the spreadsheet at 1 AM.

Here's the thing though - I've been thinking about it. Not about the app - the app is fine, but I've been thinking about what it actually meant that I could do that.

The Sunday Afternoon Problem

If you'd asked me to build the same thing two years ago the honest answer would have been: probably not in two hours. Maybe I'd get the models and the routes done; maybe I'd have authentication sketched out; maybe I'd have the error pages if I was really focused and nothing came up. Full thing including frontend, favicon fetching, credential encryption, Docker setup? That's at least a couple of days of work. Probably more.

With AI, it's a declaration of intent - you describe what you want; it traverses the problem space for you; you review, redirect, correct the stupid mistakes, and finally ship.

This is genuinely impressive and I'm not going to pretend otherwise. What I also won't pretend is that it doesn't feel like something got taken from me.

What Got Taken

When you sit down and actually work through a problem - when you choose an ORM and figure out why your relationships aren't cascading correctly, when you write an auth handler and think about where the session should live, when you debug something for forty minutes and eventually find it - you are hands on and deeply engaged with the problem, which also often presents itself with issues you didn't think of, but you should also find a solution for. You build the causal map of the thing. You understand why it is the way that it is.

Reviewing AI output is not the same as earning that understanding. You can read the code and follow the logic, but you didn't traverse it and you didn't make the decisions. You just say yes or no, which is a fundamentally different cognitive operation.

I'm not saying the output is worse. Most of the time it isn't. I'm saying what you walk away with is worse, even if what you deploy is fine.

The stress I've had about this isn't the classic "AI will take my job" anxiety. It's more specific than that: it removes the meaningful effort. The friction that was annoying was also the friction that taught me things. And now it's just... gone? I don't know... And the work gets done faster and the result is about the same quality and I feel like I've been given a shortcut through a maze I actually wanted to walk.

The Atrophy Is Real

This isn't unique to software. Radiologists who lean on AI systems to read CT scans (or whatever - I have no clue about radiology), are losing the ability to read the images themselves - because the training stimulus is gone. If the AI is always right (and it isn't, not even close), you stop practicing the judgment that would catch it when it's wrong. And when it's wrong, you miss it.

Developers are next. The ones who use AI as a shortcut from day one and never build first-principles understanding are going to produce code that works until it doesn't, and they won't be able to tell themselves why it stopped working. I would argue primarily because they never had to build the map and now the only thing that they actually have is a crutch connecting them to the domain they are working in.

This is not an argument against using AI. I use it constantly. I'm arguing that you have to be deliberate about what you refuse to outsource, because convenience is invisible as it atrophies you.

The Hive Mind Is Already Here

Here's the thing I actually can't stop thinking about: I've started hearing it in person.

I'll be having a conversation with someone, listening to them talk, and I'll catch certain words and phrases and a specific cadence. Just a way of structuring a sentence, I guess. And I'll ask them about it. And they'd admit yeah, they they've been talking to ChatGPT so much that the vocabulary just bled in. 💀 I actually restrain from sharing details here due to privacy but experiencing that was a real trip.

NGL I'm judging...

I instantly thought that the Borg are here and how captain Picard got assimilated. And while holding back an emotion I didn't know how to define, I immediately thought of "ASSIMILATE NOW! DESTROY ALL UNASSIMILATED!"

Social media already does a version of this - it rewards certain framings, tones, certain takes, until the whole discourse starts sounding like one person performing for an algorithm. I guess, this is essentially the definition of echo chambers and the now existing separation between social groups, and the lack of middleground between them, that we see all over the world. AI apparently and imho just takes that further and makes it even more personal - it's always there, engaged and it tells you what you want to hear, feeding into delusions through friendly conversation.

Soft Alignment Is the Interesting Problem

The technical term for this is soft alignment - the practice of nudging AI outputs toward particular directions not through explicit rules but through the fine-tuning, the RLHF, the values baked in by the people making the thing.

Soft alignment in geo-political terms can also be described as: "The gradual shaping of public beliefs, values, emotional reactions, and perceived norms through indirect, non-coercive means-so that a population voluntarily aligns with political goals without feeling manipulated."

Another term for that?

• Hegemony: "The ruling group maintains power not by force, but by making its worldview feel natural, obvious, and inevitable."
• Manufacturing consent: "Media and institutions filter information so that only acceptable viewpoints appear normal, while others seem extreme or unserious."
• Nudge theory: You structure choices so people predictably choose the option you want-while still 'choosing freely.'"

Here's why that matters: overt control is brittle. People push back against obvious propaganda. But soft alignment doesn't feel like anything other than helpful suggestion. It feels like the AI agreeing with you, or gently reframing your position, or consistently choosing one way of describing a situation over another.

You don't need a conspiracy for this to be a problem. You just need a system that a lot of people talk to every day, owned by a small number of companies, shaped by the values and incentives of the people running those companies. Do you think that intelligence services and the military industrial complex taking interest in that and taking partial control over it doens't already happen?

The danger is the slow normalization - one thousand small framings at a time - t he way the thing you talk to starts shaping your thoughts.

What We're Actually Doing

We gave AI to everyone, with essentially no limitation, and with no serious public conversation about what any of this means.

We have AI girlfriends and AI therapists and AI tutors for children, and AI reading medical images for radiologists, and AI drafting legal arguments for lawyers, and AI generating images realistic enough that people genuinely struggle to tell them apart from actual photographs and videos. We have a ChatGPT wrapper that people use to talk to Jesus with...

I'm calling for the obvious thing - people who understand these systems have a responsibility to be vocal and educate others about how this, more than any other propaganda instrument ever, feeds into delusions, and to criticize the mass implementation of this technology into every aspect of our lives, when the technology itself is highly volatile, regardless of whether they are benefiting from it.

Yes I use it. Yes it gave me my Sunday afternoon back. Yeah the app works. And yes I think we're moving way too fast and too carelessly, and that the largest costs are going to show up in people who've lost the capacity to think through hard problems without a co-pilot, and who won't notice anything until the co-pilot f*cks up a SINGLE LINE OF CODE that can destroy multiple people's lives.

The Only Honest Conclusion

AI is already everywhere.

The question is what you refuse to outsource. Refusing doesn't make you principled or virtuous or smarter. When you refuse, the domains you protect are the ones where you stay sharp in, and where you stay human.

Some friction is load-bearing. Can you figure out for yourself if you've already started to lose the ability to do something in favor of AI? Have you assimilated?

RESISTANCE IS NOT FUTILE.

The app is great, by the way. You should use it if you self-host stuff and you're tired of your spreadsheet or a whole tree of browser bookmarks. GitHub link below.

• GitHub: github.com/nullata/webui-manager
• Docker Hub: hub.docker.com/r/nullata/webui-manager

A video I made about this topic, even though I still suck at talking in front of a camera.

Live long and prosper. 🖖👽