Marko Njegomir

Њ | PhD student | TA | Machine learning, graph neural networks, software engineering, medical neuroscience

The limits of my compute mean the limits of my world

[Marko Njegomir @njmarko]

@elonmusk @TMTLongShort "datacenter" is a dumb name for "big computer".

"cloud"? just someone else's "big computer".

[Marko Njegomir @njmarko]

@tszzl I wonder who they believe the llm whisperer employee is that teaches models about the goblins.

[Marko Njegomir @njmarko]

@AnthropicAI @AltimeterCap @Greenoaks @sequoia Dario Amodei spotted celebrating a $1 T valuation.

[Marko Njegomir @njmarko]

@eurofounder AI laws are a scam by Big Model to sell more tokens.

[Marko Njegomir @njmarko]

@XFreeze The EU 🇪🇺 is competing in regulation, while the whole world is competing in innovation.

But in space, no one can hear the bureaucrats scream about their AI laws.

[Marko Njegomir @njmarko]

Daio Amodei, if Anthropic AI displaces human labor, including his own job, and he decides to move to Europe:

[Chubby♨️ @kimmonismus]

Anthropic co-founder Dario Amodei has been saying this for over a year now. And he keeps saying it. Louder each time.

In May 2025, he told Axios that AI could eliminate 50% of all entry-level white-collar jobs within five years and push unemployment to 10-20%. In January 2026, he published a 20,000-word essay calling AI “a general labor substitute for humans” that will cause “unusually painful” disruption.

At Davos, he warned of a “zeroth world country” forming in Silicon Valley, decoupled from the rest of society, running at 50% GDP growth while everyone else faces mass joblessness. In his own words: “We, as the producers of this technology, have a duty and an obligation to be honest about what is coming.”

And the data is starting to back him up. Tech entry-level hiring dropped 30-50% in 2025. Wall Street banks are cutting ~200,000 roles concentrated at the junior level. S&P 500 companies shed employees in net terms for the first time since 2016. Anthropic’s own labor market research confirmed that 77% of businesses use Claude to automate tasks, not to augment workers.

Now another Anthropic co-founder is echoing the same message:

“There is a real possibility that AI will displace human labor at a very large scale. Supporting those people will be a moral imperative of historic proportions.”

This is no longer a warning from the sidelines. This is the company building the technology telling you, repeatedly, that the disruption is real, it’s fast, and society is not ready for it.

[Marko Njegomir @njmarko]

@paulg The dog is plotting how to ruin the cat's plan. As is tradition.

[Marko Njegomir @njmarko]

@JulienPasteur1 The failure looks spectacular, though. 🔥

It's like Elon's engineers demonstrated that SpaceX can do the impossible by landing the rocket, and then gave us the fireworks 🎇to enjoy afterwards.

[Marko Njegomir @njmarko]

@RnaudBertrand Pro tip: Change your VPN to Wuhan, so your posts go viral. 🦠

Or just post “the algorithm is cooked”, the Grok model already predicts maximum engagement on that one. 🤷‍♂️

[Marko Njegomir @njmarko]

@Noahpinion Europeans are still reaping the benefits of the wet-tower-around-the-neck technological revolution.

[Marko Njegomir @njmarko]

The error-correction mechanism of elections could alleviate the crisis if the student-led list wins an absolute majority on its own. Whether that actually produces meaningful institutional change, or just another reshuffle, will be the real test.

It is fascinating how, in the USA, despite its imperfections, power still regularly changes hands from one party to the other, even 250 years after the Declaration of Independence.

In a two-party presidential system, coalitions are built inside the parties before elections. When elections come, if people are dissatisfied, they know exactly who to punish, because responsibility is concentrated in the ruling party.

In other words, the USA has a higher political learning rate.

In many European systems, coalitions are often built between parties after elections. This means that minor coalition partners can decide who gets power, giving them disproportionate influence relative to their vote share.

Because of compromises inside the ruling coalition, no party can fully own the program. Voters often cannot produce a clean “throw the government out” result when they want error correction. A change of government can look more like a reshuffle. So it starts to look like rule by bureaucracy, not democracy.

In other words, many European systems are more like a lower-learning-rate system.

The AI technological revolution will test not only companies and technologies, but political systems.

It will be interesting to see whether faster electoral error-correction helps countries adapt to this new reality, or whether it makes their politics more unstable precisely when long-term strategy is needed.

[Marko Njegomir @njmarko]

@elonmusk This feels like a phase of perfecting the technology, so one day it could dig tunnels under the surface of the Moon 🌔 and Mars.

Underground bases are safer from cosmic radiation.

[Marko Njegomir @njmarko]

@beffjezos Worth remembering that optimization is in third place in Elon's algorithm:
1. Question every requirement
2. Delete any part or process you can
3. Simplify and optimize
4. Accelerate cycle time
5. Automate

[Marko Njegomir @njmarko]

@AJamesMcCarthy The EU 🇪🇺 mind cannot comprehend this.

What you see vs what the EU 🇪🇺 mind sees.

[Marko Njegomir @njmarko]

@Polymarket Are workplace problems just a psyop by Big HR™ to sell you more mandatory team-building?

[Marko Njegomir @njmarko]

@birdabo He did the interview with them, and Anthropic passed Karpathy's interview.

[Marko Njegomir @njmarko]

A Titan in Geek mythology.

Karpathy's next video when he gets back to education:

[Andrej Karpathy @karpathy]

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

[Marko Njegomir @njmarko]

@karpathy Who had Karpathy doing AI research on Colossus hardware in the set of possible outcomes for Anthropic?

[Marko Njegomir @njmarko]

@djcows Ilya simply called it a big computer.

[Marko Njegomir @njmarko]

@GoodfireAI Shape rotators can only rotate shapes, but wordcels can rotate shape rotators.