ashishMay 27
Jan Wildeboer 😷

The business model of Tinder etc. isn't to find you the love of your life, it's to keep you on their site, making you pay for as long as possible.

The business model of Big AI isn't to solve your problems with the least amount of tokens. It is to keep you close enough to *think* you might soon have the solution while you keep on spending more and more tokens.

Both (and many other services) make their money based on hope, not on results. That's my opinion.

May 23 at 8:36amWeb
Show threadJan Wildeboer 😷May 23

So in my personal opinion, just as with many other things, it is better, though much more work, to be in control and to own the tools you use. Domain-specific, small LLMs work. Are they worth it? You will only know if you own them and have to shoulder the full cost yourself.

Digital autonomy or sovereignty comes from taking that perspective and responsibility. It is also defining for your personal autonomy.

That's my food for thought that I share with you. Think about it before you "yes, but".

Show threadJan Wildeboer 😷May 23
And if you think I often sound like some weird preacher — you are not wrong. My job title at Red Hat is still EMEA Evangelist :) Thinking ahead is what I do. Not just as part of my job. I share these thoughts because I hope they will help you find better ways to deal with stuff. Is all. Food for thought. Not guidance, just ideas I share so they can travel on their own :)
Show threadtotoMay 23
@jwildeboer
To me "food for thoughts" fits more to a philosopher than an evangelist.
Show threadالتنينوكسMay 23

@jwildeboer that is true. A lot of time I find the LLM just wasting token on my software project writing tons of comments redundant with code, useless documentations. Writing down in a file what he has done, and analyzing for inexplicably long time than the tasks requires.

Some of the issues can be mitigated once you encounter them, bit the tendency is the LLM just keeps finding ways to waste tokens.

Show threadLuminosMay 23
@jwildeboer You do have a talent for sharing interesting and useful information. I wish I was even half as good as you at finding interesting things to say.
Show threadBianca KastlMay 23
@jwildeboer Yes – without but. ;)
Show threadTaran RampersadMay 23
@jwildeboer well said.
Show threadHardy ErlingerMay 23
@jwildeboer Absolutely. And let's not forget the sheer *joy* of being in control of your own computing. Of understanding how things work. Of being able to fix things when they break.
Show threadhewiakMay 24

@jwildeboer

big ups for *invisible* #markdown !!

@timetinytim: please take note that this could actually be an easy and real markdown

Show threadWu Evar 🇪🇺🇺🇦May 23

@jwildeboer this comparison is flawed imho. It suggests one uses LLM tools to find *the* answer.

But it’s not.

They are very addictive tools offering your brain a shortcut in lots of cases to find *all* answers. Up to a point where it’s hard to convince your brain *not* to take this shortcut.

(And tbf: web search is dead with the flood of LLM generated and ad infested content farms which does not help at all. Maybe it’s time to bring Yahoo back 🤔)

Show threadJan Wildeboer 😷May 23
@maschinentraum Every comparison is flawed. The comparison is not the point. You find "addictive" the better term, and you are also right. What they have in common, though, is that both aim to keep you locked in, ideally indefintely.
Show threadWu Evar 🇪🇺🇺🇦May 23
@jwildeboer this is something we totally can agree on ☺️
Show threadPetra - a silly gooseMay 23

@jwildeboer

See also weight loss companies!!

Show threadHenri VerymetaldevMay 23

@jwildeboer

* hope = manipulation

Show threadts 🚇May 23
@jwildeboer yeah. This would check out. Some nasty people running these things
Show threadGabriel MarkleyMay 23
@jwildeboer it surge seems that way when you step back and look at The behavior they promote.
Show threadMarjorieRMay 23
@jwildeboer not entirely on hope: hope based on intermittent reinforcement; it sometimes seems to work.
Show threadPhilippa CowderoyMay 23
@marjolica @jwildeboer and conditions both ways: the ups feel higher, the downs feel lower and you're often anticipating those more than the ups each time you go round again anyway
Show threadNicole ParsonsMay 23

@jwildeboer

Each effort to achieve an AI result is like a slot-machine addicted gambler or a pigeon on an intermittent reward cycle.

Each attempt burns fossil fuel generated electricity.

The inefficiency & inaccuracy in AI delivery is a deliberate choice.

The fossil fuel industry is as expert at creating addicts as a drug dealer.

Used in coding, it creates a programmer so dependent they can't think through a solution independently.
https://leaddev.com/ai/addictive-agentic-coding-has-developers-losing-sleep

https://www.axios.com/2026/04/04/ai-agents-burnout-addiction-claude-code-openclaw

‘Addictive’ agentic coding has developers losing sleep

Developers speak of losing track of projects and burning out as they work with agentic coding tools.

LeadDev
Show threadBram MoreinisMay 23
@jwildeboer @DoomsdaysCW We allow Gemini for search in the school where I work. My AP CS Principles students are ending the year by watching "The AI Doc" and creating presentations and questions for each other. Last class we watched the section of the optimists claiming AI will solve all of our problems. It was utter horseshit, given the material and ecological limits to how much you can build out data centers and robots. 1/3
Show threadBram MoreinisMay 23
@jwildeboer @DoomsdaysCW So I started using Gemini to find bulletproof resources to make this case. I was shocked to learn that although my prompts clearly indicated the resources I wanted, Gemini's summaries indicating the points being made by the pages pointed to, which were in line with my prompts, led to pages that invariably said the opposite! I might ask, "show me why ice cream is fattening", and the summary would say "ice cream is fattening ", but the cited source ...2/3
Show threadBram MoreinisMay 23
@jwildeboer @DoomsdaysCW would have a title like "although people say ice cream is fattening, actually it can remove calories and fat from the body." Has anyone else discovered this manipulative agency in Gemini search summaries misrepresenting sources? It hardly seems accidental. Maybe it only relates to "anti-AI material". 3/3
Show threadDoomsdaysCWMay 23
So, I know with Google AI searches, they change the answers if you ask the question again. I guess the logic is if you don't like the answer it gives you, it goes with a counterpoint? It sounds like Gemini has something similar in its programming. @bmoreinis @jwildeboer
Show threadBram MoreinisMay 23
@DoomsdaysCW @jwildeboer My experience was that the first time I asked "provide evidence that data centers take unsustainable amounts of water from host communities" I would get an article that said "While previously water hogs, data centers are now designed to mitigate the amount of water required by means of internal circulation." Something like that. I'm sorry I don't remember the exact prompt or response but this happened over and over with different material and ecological limit prompts.
Show threadBram MoreinisMay 23
@DoomsdaysCW @jwildeboer But OK, I accept your answer that there is no system gaming, on behalf of the AI itself or its Google masters, promoting resources that counter AI-critical research by citing them as sole links to multi-point Gemini search summaries (where I would have expected many sources). Still, it would be very easy for Google to do that if it wanted to and would not require GAI.
Show threadroboticus lastius (not a bot)May 23
@bmoreinis @jwildeboer @DoomsdaysCW
Regardless of Google deliberate intent, if I take this scenario to other players, say if I asked a research librarian or a topic researcher - and they repeatedly replied to my questions about negative impacts of XYZ with inverted answers like "XYZ is actually not bad", I would view that person as deliberately contrary, agenda-driven and unhelpful.
I see no reason to cut AI more slack than I would a human in that role.
Show threadDoomsdaysCWMay 23

Hear hear, @lastrobot ! Well-said! (And I've worked as a reference librarian, so yeah.)

@bmoreinis @jwildeboer

Show threadathena_risingMay 23
@bmoreinis @jwildeboer @DoomsdaysCW Putting the costs of this mutated Tetsuo aside, it is not, nor can it ever be in this form, relied on. As an experiment, I asked ChatGPT for a list of ten poems with a certain theme to share with a student. Seven do not exist. I wish they did; they sounded perfect. And that's the issue, AI is not an intelligence. It is a data-driven mimic. They have fed it enough human information, that it picks out patterns in your queries, and responds with what has the highest likelihood of being an answer. Sometimes it's correct, but sometimes it's not. For the seasoned programmer or teacher, that's not ideal, but inaccuracies can be compensated for. For students and the juniors in their careers, this is a disaster.
Show threadBram MoreinisMay 24

@athena_rising @jwildeboer @DoomsdaysCW I just found an article that tells more of the story of how search results (lists of links) are deprecated in favor of AI summaries:
https://techcrunch.com/2026/05/19/google-search-as-you-know-it-is-over/

"Links will become an afterthought with the coming changes to the Search results experience, which builds on Google’s earlier launches of AI search features, like its short summaries known as AI Overviews and its conversational search, AI Mode."

Google Search as you know it is over | TechCrunch

Google is transforming Search from a list of links into an AI-powered experience filled with conversational answers, autonomous agents, and interactive interfaces — a shift that could further reduce traffic to publishers across the web.

TechCrunch
Show threadMim54May 23
@jwildeboer in other words gambling, rotten to the core
Show threadDanielMay 23
@jwildeboer bonne analyse,il y a aucun doute, c’est exactement ça
Show threaddemofoxMay 23
@jwildeboer something that has always annoyed me is tv shows dont aim to tell good stories, just keep you coming back.
It's "capitalist optimization" taken too far, like so mamy other things.
Show threadAlex FürstenauMay 23
@jwildeboer yep. Somehow this capitalism model doesn't seem to work very good to solve a specific group of problems. 😐
Show threadcratermoonMay 23
@jwildeboer there’s another industry with that exact business model. You can see it in operation at its most gaudy and unabashed in Las Vegas, Nevada.
Show threadPhilippa CowderoyMay 23
@jwildeboer See, OKCupid got around that by also being a great way to meet friends. Big AI is also a great way to cover your friends in verbal shite, I guess?
Show threadMichael HartleMay 23
@jwildeboer This ties nicely into understanding AI prompting as a Skinner box. Like slot machines in Las Vegas, loot boxes in online games or booster packs in trading card games, success & gratification is probably just one more try away...
Show threadDávid BárdosMay 23
@jwildeboer I agree. Monthly pricing demands efficiency. Token-based pricing, however, rewards wasting time and resources while finding the solution the longest and most overcomplicated way possible.
Show threadLambdaMay 23
@jwildeboer we have built our own carrot on a stick
Show threadChris In Toots 🌈May 24

@jwildeboer

Prediction market sites aren't there to help find 'public truths', they're there to take as many bets, and commissions, as possible

Show threadDavid EspositoMay 24
@jwildeboer capitalism does the same thing. its a lot of the progress narrative is just a fiction it self. its just using death anxiety and collective threat to make primates scared enough and then feeding them propaganda enough to build a identity so they can then feel that " i can choose to work on what i want and get good at what i want and then because im good i can relate to others" when really that even "i can choose" is itself a fallacy. and really we never got rid of our feudal substrate to society. and exchange value isnt the purpose of capitalism its to accumulate capital and make other primates irrelevant.
Show threadck0May 26
@jwildeboer Same with Netflix (that's why the UI is so painful to find shows of a specific kind compared to this good old popcorntime) or Google Android (that's would explain why making basic phone action as disable the number presentation takes as much user inputs as it currently does).