This, but for education, too!

At this point they have to create so much more than AGI. They have to turn whatever LLMs are into the next iPhone and the next Azure and the next Office Suite in two years. They’ve pissed off everyone and already spent so much money and it is in no way even remotely worth it

AI is best understood as a cyberattack against a functioning society and should be treated as such

Yes this is essentially how LLM companies operate. Selling compute with a heavy subsidy. No plan to turn profitable, because the users they have today ONLY use their product because of the subsidy. The moment they pay the real price, bye bye most users. It'll be a disaster when it pops. 🫧📌

Imagine that you spent a year writing a book, only to find that AI is distributing your text without your consent 🤯 In March, the European Parliament called to update copyright rules to protect creators. Today, on World Book and Copyright Day, we highlight authors' rights 📚

Its always so interesting how "Anti-AI extremists" are always "hysterical" and "loud" but AI "enthusiasts" are always just hopeful and optimistic (maybe hyping at best) and nothing else

The kids these days say “AI” in the same way that people our age say “bullshit” (“you didn’t do that, that’s AI!”) and I think that’s beautiful

Dad of two grade-schoolers here: can confirm that the kids are using “AI” as a synonym for “fake”

I'm announcing a new business: I'll sell $1/month Linux VMs that has 64 cores and 32 GB RAM. 🙃 I expect to get millions of users and billions of revenue shortly! All I need is a bunch of clueless investors who'll give me a trillion dollars to burn on this without any plan for profitability.

It's starting... 🍿🫧📌

The Child, aged 17, witheringly asks "where did you hear that? ChatGPT?" when I've said something she thinks is stupid.

Despite LLM booster attempts, you can't separate a product's capability from its price. LLM companies have been massively subsidizing a mediocre product, and got a bunch of users. But the product isn't good enough for people to pay the full price. This is how the bubble pops. 🫧📌

Deadlock over. The EU just cleared way for the €90-billion-loan for Ukraine and the 20th sanctions package. Russia’s war economy is under growing strain, while Ukraine is getting a major boost. We will provide Ukraine what it needs to hold its ground, until Putin understands his war leads nowhere.

The Matrix is literally about AI making people dumber, exploiting them, and forcing them to doubt their sanity, all to centralize power. Literally the effect that LLMs are having on humanity right now. Yet most of the nerds boosting LLMs pretend to be Neo. But they're Cypher.

Morning friends. There are still those amongst us that need to wake up.

“Why don’t you want to have a conversation about AI?” “I broke into your house and stole your shit and now I just want to talk.”

How are you going to replicate the unimaginable amount of compute it takes to run this monstrous pile of stolen data through the brutest of forces to produce marginally useful results 30% of the time, none of which anyone is paying for with their own money?

AI boosters are, like, really smart

Oh he's somehow both

I just keep thinking, writing the code never was the slowest part of my process anyway.

AI has not changed the way software is built. Code is written using the same syntax, version controlled using the same version control systems, compiled using the same compilers, deployed to the same servers, and neglected by the same developers.

Once again, this level of incompetence in software by Anthropic is the strongest proof that LLMs can't replace actual software engineers 🫠

+1

“AI is here to stay” says the makers of NFTs, the metaverse, Google Glass, Windows Phone

The subsidies are slowly going away, and suddenly asking your LLM to rename a file will start costing you real money. This is why I think it's naive to build a workflow *now* around this subsidized technology. Better to wait until they stabilize on real prices and THEN let's evaluate the tech.

One thing I don't think I see talked about enough is how AI services like Claude Code are psychologically similar to gambling. When you spend tokens you have a chance of a good result, with lots of smaller good results along the way to keep you hooked. No wonder some people say they're "addicted" 😬

I'm not joking when I say mRNA technology is more important than "AI" and it's a tragedy we're throwing billions into one while our government is aggressively defunding the other.

Just imagine if we invested at least one tenth of resources into vaccines that we're burning on fucking LLMs. But for that, we'll need to wrestle power away from the same ~5 billionaire dudes who control everything. Bc they're doing the inverse - trying to use LLMs to steal our jobs. Fight back.

A few days later and leaks from Anthropic confirm that the actual reason why Mythos isn't out isn't because "it's too good, it's dangerous", but because they don't have the capacity to serve it at their current massively subsidized prices (no shit, Sherlock). What a bunch of clowns 🙃

Congrats to everyone who fell for this just because Anthropic said it lol

This is tech I can get excited about and that actually deserves trillions of funding: mRNA vaccines www.nbcnews.com/health/cance...

Nothing to see here, everything is fine

Every new minor update to an OpenAPI/Anthropic model is greeted with a bunch of vibe-coders declaring "this changes everything". I've been observing this for two years now. Yet, nothing has really changed? Some folks have a fast buggy autocomplete, but software engineering looks identical to 2022.

"Token usage [of 4.7] on the same problems is up 35% on 4.6" This is why the claims of inference cost going *down* is a lie by the LLM industry. While per-token cost might be down, if the same query requires a lot more tokens, the per-query cost is up. Plus is still subsidized 🫢 🫧📌

I just donated to @eurosky.social. We need independently run European atproto infra (pds, relays, appviews), and a funding and legal framework for the community to flourish. The folks at Modal Foundation, Eurosky's parent, are doing great work to help us get there. Let's fund.eurosky.tech ✌️

If AI was so great, then the tech companies wouldn’t be in its 3rd or 4th year trying to force it onto the public. Good shit sells itself. Bullshit needs all the hype it can get.

Reese Witherspoon? The NFT evangelist?

Same people who tried to sell you NFTs and crypto are now trying to sell you LLMs - totally legit

Not everything that breaks is technically the fault of vibe coding but the kind of culture that would foster vibe coding is the kind that produces a lot of shit that breaks.

I don’t know or really care if they’re lying or telling the gods honest truth: I want it to percolate into the money ghouls petrified brains that AI is a stinker, a loser, a bad investment

This aged well 🫢

i’m not saying that vibe coding caused this outage but i am saying that if you vibe code everything then it leaves you woefully unprepared to fix outages

Bsky staff is scrambling to convince ppl that today's outage has nothing to do with their embrace of ai. That's great. We should make them scramble more. The cost of replacing competent humans with AI needs to be greater on every single level for every single company.

It, again, truly doesn't matter whether vibecoding actually caused the outages or not - when you start bragging about how claude is making every engineer on staff more productive and the site goes through months of persistent downtime, that's a Statement.

Sifa is a new professional networking platform on the Atmosphere - fully compatible with your Eurosky account. ➡️ go to sifa.id ➡️ login with your Eurosky account ➡️ import your LinkedIn profile and upload to Sifa You’re set. Sifa have a very exciting roadmap of features coming. Be ready! @gui.do

If AI was genuinely useful your employees would choose to use it, you wouldn't have to mandate its use.

The biggest reason why I call bullshit on Mythos being a super intelligence is because the Claude desktop app Anthropic just released is just so bad. So many bugs, confusing interactions, clearly no human cares about this product. Still, only passionate humans can produce high-quality software.

Every company's marketing strategy in 2026:

If this doesn't definitely prove we're in a massive AI bubble, I don't know what will