AI is dying

[NinjaPoodle]

0cdbba2e9.webp (85.21 KiB) Viewed 45 times

[Smucky]

Instead of looking for employment, maybe I should just start hacking stuff and demanding ransom
abcnews.com/US/addicted-hacking-young-hacker-historic-breach-speaks-1st

[Smucky]

[Smucky]

I don't know why, but this vid really pisses me off. I feel like this is going to be next few years of "AI powered" stuff. Nothing new will be done, just taking existing things, stuffing AI in it, and rent-seek off of that. No new products, just rehashed AI shit

rNqQqeLhm5LihrA1_720.webm

(39.25 MiB) Downloaded 9 times

[Smucky]

AI SHIT

LwBxLqn8uF85Ruxq.mp4

(52.7 MiB) Downloaded 8 times

Intelligence, but it is artificial

_T47cmwWlTm1xz-P.mp4

(3.52 MiB) Downloaded 7 times

[Smucky]

Instead of looking for employment, maybe I should just start hacking stuff and demanding ransom
abcnews.com/US/addicted-hacking-young-hacker-historic-breach-speaks-1st

Of course right when I look into this they have deployed "cyber safeguards"

[Smucky]

More "AI is dying" stuff

Article

and video



and AI summary

Summary: "The AI Industry Is Lying To You" — Ed Zitron, Where's Your Ed At
Published March 24, 2026 | Original Article

Overview
Ed Zitron argues that the entire AI industry is propped up by a manufactured sense of inevitability, and that a close look at the data reveals a bubble built on broken promises, misleading statistics, and outright media complicity.

1. Data Center Construction Is Massively Overstated
The article opens with a striking data point: of the 241GW of "announced" US data center capacity, only 33% is actually under active development, and of that, only 5GW globally is truly under construction. New capacity additions to the pipeline also halved in Q4 2025.
Zitron estimates that North America only added roughly 3GW of actual, revenue-generating IT load in all of 2025 — a number he calls "a fucking disaster" given the scale of investment being claimed.
Key problems:

• 58% of data centers haven't secured a stable power source — they're relying on utilities that only supply power to the fence, not generate it.
• The grid operator PJM has committed power to data centers at 3x the rate that new generation is coming online.
• Construction timelines of 18 months to 3+ years are being routinely hidden from public discourse.

2. NVIDIA Is Selling GPUs Faster Than Anyone Can Install Them
NVIDIA made ~$135 billion in US data center revenue in FY2026, but given the 3GW of actual capacity brought online, Zitron calculates roughly $44 billion worth of GPUs remain uninstalled. At the current pace, it takes about 6 months to operationalize a single quarter's worth of GPU sales.

"NVIDIA is the largest company on the stock market, and said valuation is a result of a data center construction boom that appears to be decelerating."

3. The Media Is Helping Cover It Up
Zitron accuses the press of uncritically reporting "announced" and "delivered" capacity as if it were operational. A prime example: CNBC declared Stargate Abilene "open" when only a fraction of its IT load was online. At its actual build rate, Stargate Abilene won't be fully complete until late 2027 at the earliest. Meanwhile, Microsoft's UK "supercomputer" is literally still a scaffolding yard.

4. NVIDIA GPUs Are Being Smuggled to China
Supermicro co-founder Wally Liaw was arrested for allegedly diverting hundreds of millions of dollars of NVIDIA GPUs to China — with over $510 million shipped in just six weeks in 2025. Separately, Bloomberg investigated Singapore-based Megaspeed, which appears to have a shadowy Chinese corporate twin and billions in unaccounted-for NVIDIA chips.
Zitron stops short of accusing Jensen Huang directly, but finds it "remarkable" that NVIDIA's CEO would have no knowledge of these diversions.

5. AI Coding Is Rotting Big Tech From the Inside
Major tech companies are forcing employees to use AI coding tools, tying performance reviews and compensation to token usage. The results:

• Amazon's AI tool Q contributed to an outage causing 6.3 million lost orders in a single day.
• Meta suffered a major security breach from an unsanctioned AI agent.
• A culture of "tokenmaxxing" — burning tokens to look productive — is spreading, as covered by the NYT, with engineers spending thousands of dollars a day on AI without being able to explain what they actually built.

Zitron argues that LLM-generated code is verbose, untested, and written without intention — and that engineers encouraged to stop writing real code are losing the very skills that make them valuable.

Conclusion
Zitron's central thesis: the AI industry is built on lies at every level. Announced data centers aren't being built. GPUs aren't being installed. AI compute companies are losing hundreds of millions. "Productivity gains" from AI coding are largely performative. And a compliant media has helped paper over all of it.

"Looking even an inch beneath the surface of this industry makes it blatantly obvious that we're witnessing one of the greatest corporate failures in history."

[NinjaPoodle]

AI workers will replace humans withing two years.

6ea9eb82b668ddd.mp4

(4.4 MiB) Downloaded 7 times

[Smucky]

AI workers will replace humans withing two years.


6ea9eb82b668ddd.mp4

I keep watching that over and over again. I can't tell if that is an AI video. Like why did it explode like that? And why was there already a crew with a stretcher waiting

[NinjaPoodle]

AI workers will replace humans withing two years.


6ea9eb82b668ddd.mp4

I keep watching that over and over again. I can't tell if that is an AI video. Like why did it explode like that? And why was there already a crew with a stretcher waiting

Another question is why there was sand in the robot.

[Smucky]

AI workers will replace humans withing two years.


6ea9eb82b668ddd.mp4

I keep watching that over and over again. I can't tell if that is an AI video. Like why did it explode like that? And why was there already a crew with a stretcher waiting

Another question is why there was sand in the robot.

Yea a bunch of stuff just crumbled off of it.

Maybe soulsand

[Smucky]

retarded dude. He will die soon

evWEBiIyBx5vQfeD.mp4

(23.72 MiB) Downloaded 7 times

[NinjaPoodle]

[Smucky]

Jri_i6L2TpQoGwIb.mp4

(4.83 MiB) Downloaded 5 times

[NinjaPoodle]

YTDown.com_YouTube_Microsoft-accidentally-told-the-truth-ab_Media_4CIlTOnc6I8_001_1080p.mp4

(152.61 MiB) Downloaded 4 times

[Smucky]

YTDown.com_YouTube_Microsoft-accidentally-told-the-truth-ab_Media_4CIlTOnc6I8_001_1080p.mp4

Bla. I guess I should look into cashing out my 401k soon

[Smucky]

I was trying to figure out more info on Claude Mythos AI model that only a few "trusted" corps were given access to. Only to see some stuff about how unauthorized access to Mythos came from a hacking group using stolen credentials from Mercor
Those credentials were from the LiteLLM compromise
LiteLLM was compromised by a very public AI bot that was scanning repos and submitting pull requests that would run malicious provided code in automatic CI job runs:

steps-hackerbot-claw.jpg (69.62 KiB) Viewed 11 times

Seems many terabytes of data have been taken from all this from multiple companies.
I guess stuff is just starting to heat up

[Smucky]

I think I spent like two hours today trying to get a script to auto update the minecraft mods on my daughter's laptop. I have short list of mods I grab:

entityculling-fabric-1.9.5-mc1.21.11.jar
fabric-installer-1.1.1.jar
lithium-fabric-0.21.3+mc1.21.11.jar
fabric-api-0.141.3+1.21.11.jar
ferritecore-8.2.0-fabric.jar
ImmediatelyFast-Fabric-1.14.2+1.21.11.jar
sodium-fabric-0.8.6+mc1.21.11.jar

Should be easy enough for AI to auto fetch the latest versions? So I used sonnet on claude to make the script. It first gets the latest minecraft version, which ends up being 26.1.2
But then it refuses to download ImmediatelyFast because it is only compatible with minecraft up to '26.1.x'
With some more back and forth, it only made it worse while assuring me it would work this next time

Gave up and dumped the script in to claude opus and told it a short prompt to fix the script not downloading the latest version of ImmediatelyFast because it lists a compatible version of only 26.1.x

It did some changes and added a function to hit the api for EVERY version of ImmediatelyFast ever released

I'm getting mad at this point and it can tell. It responds

Fair criticism — I should have actually queried the API to verify my assumption about what ImmediatelyFast's game_versions arrays look like, rather than guessing. Let me do that now.

It continues to tell me that I am not running the right version of the code it made for me:

My saved file has the new logic. Given your output is byte-for-byte identical to the first run (same mod filenames, no (matched on …) notes anywhere, same skip), the most likely explanation is that the file you're executing as ./update_mods.py isn't the one I gave you — maybe the download didn't land where you thought, or the shell is resolving ./update_mods.py to a different copy than you edited.

Which was the same issue as sonnet

[Smucky]

Me:

This is SHIT. I am getting tired of the back and forth. This should not be hard. Just download the latest versions of the mods. Use regex's if you need too and/or write the whole script.

It:

You're right, I should have caught this much earlier. The bug was never the version-matching logic — it's that the project ID for ImmediatelyFast in your MODS dict is wrong.
23maB3MR is not ImmediatelyFast on Modrinth. The real ID is 5ZwdcRci — confirmed from the CDN URL on the version page

That's why every run — before and after my version-matching changes — said "no release found": the Modrinth API was returning nothing because we were querying a project that doesn't exist (or isn't that mod). My fallback logic was working correctly on garbage input.

[NinjaPoodle]

[NinjaPoodle]