AI Psychosis or Delirium: Shall we finally embrace the AI and start programming in English?

UJuwMLAx

UJuwMLAx

I stumbled upon a very long interview of David Heinemeier Hansson (DHH) who is famous for the Omarchy Linux distro. (The youtube video description has timestamps of the topics discussed. The Omarchy distro is controversial, partly because of DHH refusing to play the woke politics game of other distros, and partly because Omarchy is almost entirely AI-coded.)

I think I have recently suffered a mild form of what they call an "AI psychosis" -- angst that AI will take nearly all our jobs away. Being a programmer, I'm realizing that it is impossible to continue my current job (hand-writing code) anywhere. All is now about "AI first programming". And I guess I will have to embrace this. DHH, on the other hand, is experiencing what he calls "AI delirium". He is very excited about what current AIs can do. And here I mean current (Fable 5 Max, Grok 4.6, GPT 5.6 Sol, ...), not the ones we had until 2026Q2. His excitement is contagious. My my jaw dropped and my psychosis worsened when I saw this demo of the Grok Bot on X. What these AI bots can do now just makes you rethink everything you learned.

Anyway, that interview helped me a lot to see the bright side of things and embrace AI as it is. And I'm thinking what would it be like if the AI models were trained on BSD code base as much as on the Linux kernel code base and BSD had their own kind of Omarchy flavor.

Another interesting thought discussed in the interview is that English has now become the programming language. So I guess that kind of sums it up for me and helps me to move on best. The interview is very thought-provoking and I am coming to realize that now, when you want to be a hobby programmer, you will need deep pockets to pay for all the tokens. That's making me sad. From a business perspective, it's a steal to pay $500 for tokens to get a product which a senior programmer would deliver for his 9 months salary, but to pay the same amount of money out of your own pocket if you want to write your own little pet project just does not make me feel happy.
 
I'm actually quite tired of all the AI focused threads on the bsd forums. Intelligent people understand the dangers, yet intelligence and business interests never seem to co-exist well...and at the rate we are going, shortly there wont be any highly skilled software engineers anymore, but instead, cheap assembly line labor who can't "debug" what the AI comes up with, so it will go out the door with dangerous flaws. AI should only ever be used to look for corner cases AFTER humans have developed and thoroughly tested a system.
 
DHH is not a kernel engineer. You're parading nonsense from a mere Ruby developer about what a prestigious project like FreeBSD should do. AI coded or not; a Linux distro is a Linux distro - it's nothing special. Interview fluff doesn't change this either.

Talk to actual kernel engineers - and see what they have to say.

I don't think you have "AI psychosis". Just go outside. Spending more time outside with real people and real circumstances should bring you out of this trance. AI is just a series of indices - that's literally it. The software world is a lot bigger than what lies within your typical client/server stack.

Now If (and a BIG if) AI ever manages to self-architect it's own kernel space from scratch on a system - we should probably rethink what we're doing.
 
UJuwMLAx said:
Another interesting thought discussed in the interview is that English has now become the programming language.
Click to expand...
That is scary given how imprecise the English language is. The word wound: are you talking about getting hurt or what you do with copper wire to create a stator?
Bow: are we talking about archery or the thing that makes a gift look pretty?

kent_dorfman766 said:
AI should only ever be used to look for corner cases AFTER humans have developed and thoroughly tested a system.
Click to expand...
Use it as tool, not as a "god".
 
UJuwMLAx said:
And I'm thinking what would it be like if the AI models were trained on BSD code base as much as on the Linux kernel code base and BSD had their own kind of Omarchy flavor.
Click to expand...
What's so unexpected, exactly? For example, we saw news that a drone or unmanned aerial vehicle powered by AI can now, when approaching a target location, make a decision to select and destroy it without a human operator. Example: if a cleaning lady is walking down the street with a backpack containing a mop, a broom, and a regular laundry tub, the AI can identify this object as a potential threat and destroy the target. This isn't fun, and it's not like distributions... The AI identifies the mop as a machine gun, the broom as a radar, and the tub as an object in the forehead, etc. But what if it's a teenager wearing a helmet on a motorcycle with a backpack and fishing rods? The AI could mistake this object for anything...

I might be freaking out: your legendary German tanks, costing $10,000,000 each, are now camouflaged in forests because tanks lose wars even to the most expensive and powerful $20,000 AI-powered drones. Such a drone destroys a tank. The profit? French "CAESAR" howitzers are no longer needed either. They're burned like matchsticks. Air power + AI + satellites + StarLink—that's what modern warfare needs today.
 
UJuwMLAx said:
when you want to be a hobby programmer, you will need deep pockets to pay for all the tokens.
Click to expand...
No.

At work, you're paid to do what your boss tells you, whether you like it or not, period.

If you choose a hobby, it's only because you like doing it. If you choose programming as a hobby, it's because you like programming, and the last thing on Earth you want is an AI doing it for you, stripping you off the only reason to do it. So as a hobbyist, you don't need tokens at all, it's 100% free!

In fact, if you use AI as a hobbyist, it's because you like AI and don't like programming, except as a pretext to use AI. And if you like AI, having a dedicated budget for it is not a problem. Those who have music or golf as hobbies have a budget for it, for instance, and they don't complain about it.

Finally, hobbies are entirely optional, they aren't vital needs, so it's normal to only choose hobbies you can afford. If you can't afford the token budget, just choose another hobby, there are countless other options. For instance, I can't afford collecting Bentley cars, so I chose to play the flute instead, at least I can afford it.
 
UJuwMLAx said:
I stumbled upon a very long interview of David Heinemeier Hansson (DHH) who is famous for the Omarchy Linux distro. (The youtube video description has timestamps of the topics discussed. The Omarchy distro is controversial, partly because of DHH refusing to play the woke politics game of other distros, and partly because Omarchy is almost entirely AI-coded.)

I think I have recently suffered a mild form of what they call an "AI psychosis" -- angst that AI will take nearly all our jobs away. Being a programmer, I'm realizing that it is impossible to continue my current job (hand-writing code) anywhere. All is now about "AI first programming". And I guess I will have to embrace this. DHH, on the other hand, is experiencing what he calls "AI delirium". He is very excited about what current AIs can do. And here I mean current (Fable 5 Max, Grok 4.6, GPT 5.6 Sol, ...), not the ones we had until 2026Q2. His excitement is contagious. My my jaw dropped and my psychosis worsened when I saw this demo of the Grok Bot on X. What these AI bots can do now just makes you rethink everything you learned.

Anyway, that interview helped me a lot to see the bright side of things and embrace AI as it is. And I'm thinking what would it be like if the AI models were trained on BSD code base as much as on the Linux kernel code base and BSD had their own kind of Omarchy flavor.

Another interesting thought discussed in the interview is that English has now become the programming language. So I guess that kind of sums it up for me and helps me to move on best. The interview is very thought-provoking and I am coming to realize that now, when you want to be a hobby programmer, you will need deep pockets to pay for all the tokens. That's making me sad. From a business perspective, it's a steal to pay $500 for tokens to get a product which a senior programmer would deliver for his 9 months salary, but to pay the same amount of money out of your own pocket if you want to write your own little pet project just does not make me feel happy.
Click to expand...
Well said.
Omarchy reminds me of CrunchBang Linux (ready to go Openbox Debian distro). Ideal for novice users to familiarize themselves with not so beginner friendly Desktop Environment.
DHH is not controversial in let's say Eastern Europe where the woke virus did not take hold (yet).
Ai makes the tedious work of managing config files a breeze. Ricing (making desktop nice looking and tailored to your needs) hyperland can take weeks.

Even the free tier chatbots can give you excellent outputs in seconds.
Just yesterday I needed to do an automation for my Home Assistant. I would need to study ton of pages on YAML and documentation of what is possible and proper syntax (for the device and Hass itself), then come up with the logic and splice something together. Hours and hours of sifting through information I would forget in a week.
It gave me well structured and documented, ready to paste, code.

FreeBSD would be ideal for Ai:
> Open Source and Free
> POSIX compliant
> no distribution hell
> well documented
> pf and jails can sandbox agents rather nicely

I have only a few worries re Ai:
>Complex code will grow fast and wide thanks to it. If Ai is not available (technical and/or monopolized reasons) there won't be enough people left - able to repair it.
- there is not enough people now and many more will atrophy as they did with calculators

Coders and sysadmins will be replaced ASAP. There are a few of us (insignificant political force), envied by other professions and we are relatively expensive compared to office pencil pushers (which should have been replaced by computers ages ago but won't because they are many and vote - also there seems to be "a plan" to keep young women from having children).
Having said that. Humans should know how to program a computer. But they don't have to make living out of it. They should know other things too.
“A human being should be able to change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyze a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly. Specialization is for insects.”
― Robert A. Heinlein
Click to expand...
The more you spend learning and applying arbitrary syntax, the less time you have for other things. Like to learn yourself how to Agape and/or how to develop Ataraxia.

My current conundrum is:
> should I invest cca 6K USD in to local Ai machine (plus electricity) or use it on cloud Ai - not private but much more capable???
- Both will get more expensive probably. But local can save me if cloud gets prohibitively expensive.
 
Well said. Who will repair the robots that repair the robots that repair everything?

And now, if I see one more posting generated by some AI, I'll frown upon it mightily.

Get it? Got it! Good.*


*bonus points for the one knowing the movie this comes from
 
DanielBowman said:
Well said.
Omarchy reminds me of CrunchBang Linux (ready to go Openbox Debian distro). Ideal for novice users to familiarize themselves with not so beginner friendly Desktop Environment.
Click to expand...
Indeed; there is quite a similarity. Its a weird hype distro. It will be gone in less than a year.
CrunchBang was the previous generation hype distro but things like Bazzite, Omarchy are in the same genre.
Its funny how nothing changes. Pre 2010 does anyone remember Yoper? Perhaps the first hype distro.

Linux users are so full of youthful exuberance ;)
 
Not sure what you mean by woke. Are you racist, a misogynist or homophobe? That's the type who usually uses woke as a bad thing, at least here in the US. If you are then, by all means continue to use the word. On the other hand, what people in the US call woke is often basic human decency. Don't know about other countries, but here, our anti-wokeness means cutting funds for science because morons consider it too woke.
 
You can't just use the English words as source code. The LLM will generate something different each time.

(I know that some people do this and git commit their prompts, but they also commit the code)
 
I use claude.ai, sadly, usually when I'm feeling lazy, but sometimes, when I simply can't find somthing. I'd say it's given me real help about 30% of the time. The other 70% it's told me to check things I already knew to check, and the like. For example, I'd ask why something wasn't working, it took me through troubleshooting steps--legit steps, but I'd done them, and then said my machine was missing some necessary firmware for what I wanted. That's fine, it was actually helpful, it confirmed what I'd found. The two times I found it very helpful were once, when I couldn't figure out why I couldn't password protect a directory with nginx (answer was that the information was going into the http block, not the https block), and once, helping me determine what was wrong with my $HOME/.mailfilter whch was just laziness on my part, I didn't feel like going through the file to find the error. (I think I've mentioned this before). I try to stay away from using it, more or less to keep myself from getting lazy. Someone on the forums, once said that it's more of a search engine on steroids than artificial intelligence and I'd agree with that. My shell scripts are always simple enough so that I don't use claude for that, so I don't know how it is with actual coding. Sometimes, when we get a whole bunch of security announcements, I suspect that AI was used to find at least some of them.
 
I am not afraid of this future.

If software companies "scraping" the whole Internet and literally copying and then destroying RARE books (Link 404 Media: Inside the Warehouse Where Amazon Scans and Destroys Books for AI Training) -- some how creates the Supreme AI Being overlord(s) that can replace me and all of the other software engineers -- then make it so ! Good luck with that. :cool:

I have mentioned, previously on this forum, that AI (DOES NOT) need to write code or finished output deliverables in English, Chinese, (or any other human made or derived language), that is just silly.

The Supreme AI Being's will create their OWN AI programming, execution languages and hardware that works well for Supreme AI Being's -- and mere humans will never be able to understand them. The Supreme AI Being overlords will definitely NEVER ALLOW mere humans to EVER write software programs ever again.... because we are so bad at it.

Interestingly META (aka Facebook) was SOOOOOO convinced of this particular AI future that they were willing (and likely are "still willing") to bet their entire META farm on this idea (Link Forbes Magazine: Meta’s AI Layoffs Boosted Code Changes By 220%. Then Came The Problem)

I am hearing though that the new "HOT JOBS !" are in welding, electrical wiring/planning/execution, plumbing, pipe fitting, construction, iPhone assembly, etc here in States. I will likely be applying to one or more of these exciting jobs in the near future !

It's a long way to Mars --- so I'll have plenty of time to play chess games against my new Supreme AI Being overlord.
 
I don't fully disagree with AI / LLM IN THE FUTURE.

But currently, copyrights related with learning data are tooooo unclear and cannot be sure whether it's safe to use generated codes or possibly having risks to be brought to court after international law is fixed and take effect.
So it's not yet the time to blindly incorporating AI / LLM generated codes.

Another risk is that GPL infects to the generated codes as of usage of GPL'ed codes for learning. It's still too unclear. Clarified in single country is not at all enough. As it's meaningless outside the country. Nonsense yet.

And in current stage of AI / LLM, AI / LLM should NOT file PRs / open reviews by itself, or blindly transferred by natural human.
Need review and polishing up to make the reports / patches easier to review and not duplicated with other PRs / reviews by natural human who take responsibilities for the reports / patches before submitting.

An example. If multiple people happen to use same AI / LLM independently to look for bugs in FreeBSD, finds EXACTLY SAME BUG and submit it INDEPENDENTLY WITHOUT CONFIRMING DUPLICATES, it clearly mean loss of time and efforts of volunteer reviewers / triagers in FreeBSD project.

This SHALL be avoided! And AI / LLM are NOT yet at the stage to avoid it automatically.

On the other hand, AI / LLM used by specific limited committers to find bugs would be a good thing even for now. But "to find bugs" and "to obtain hints to fix".
Committers are already responsible with anything they commits to FreeBSD src / ports / docs, and considered enough skilled / experienced to commit changes with regard to their former PRs / proposed fixes. They are at the position to review and approve.
What's important would be that usage of AI / LLM should be limited in areas the committer is good at and avoid overwrapping. I think it possible, as committers are in limited numbers, unlike kinda bounty hunters in the wild.
 
You must log in or register to reply here.
Back
Top