Sebastian Harm

German, Computer Scientist

[Sebastian Harm @HarmSebastian]

@ChariteBerlin Danke für eure Arbeit!

[Sebastian Harm @HarmSebastian]

Mein Beileid an alle Betroffenen in Berlin #CSD
Ich bete für alle Verletzten und Betroffenen. Es ist fürchterlich.

[Azyra @cryptoazyra]

At least we got nice skin 😭

[Sebastian Harm @HarmSebastian]

We Europeans often view us as better than the Americans. "Free" Healthcare, no mass shootings, we all know the arguments and stereotypes.
However, we fail to deliver in technological terms. Yes, we have a strong Middle Class. But we don't have something like SpaceX, nothing like OpenAI or Anthropic (relative to the market volume). Our industrial output is declining rapidly.
Technological advantage has always been Europe's path to success and dominance. I don't wanna live in a country that only plays a footnote in future history and science. It's not only politics, which can safe us. We need initiatives from the heart of our European societies that change exactly that. It's doable!

[Sebastian Harm @HarmSebastian]

Die Wand ist nun fertig. Ich mag das Ergebnis total, besonders dafür, dass wir das zum ersten Mal gemacht haben:

[Sebastian Harm @HarmSebastian]

Achtung: Work in Progress!

Hätte mir vor 5 Monaten jemand gesagt, dass ich meine Wand rosa streiche, darüber ein dunkelpinkes (mit Glitzer!) und graues Muster malere, dann hätte ich das wohl lachend abgetan - zumal es mir tatsächlich zusagt. Turns out, es fehlte nur die richtige Freundin.

[Sebastian Harm @HarmSebastian]

@OpenAI @huggingface Crazy times ahead

[Sebastian Harm @HarmSebastian]

Ich arbeite an einem Projekt. Zunächst wird eine Projektbeschreibung erstellt, dann rufe ich @grok oder @AnthropicAI Claude auf, gebe die Beschreibung rein, frage nach Unklarheiten und lasse mir so einen Prototypen erstellen. Den bearbeite ich dann, ändere den Code leicht oder baue ihn an den entscheidenden Stellen aus. Was früher deutlich länger gedauert hätte, schaffe ich heute an einem Tag.

Und: Während die KIs denken, habe ich Zeit Tweets abzusetzen oder bissl Sport zu machen. ;)

[Sebastian Harm @HarmSebastian]

@sama @huggingface Is this the beginning of who can achieve the more dangerous model in the shortest time competition?

[Sebastian Harm @HarmSebastian]

What the fuck is Google doing with Gemini? Anything interesting coming out of Google these days?

[Sebastian Harm @HarmSebastian]

@ulfposh Sie finden wirklich immer eine Weg um neue Schulden zu machen.

[Lukasz Olejnik @lukOlejnik]

Fun fact. Other, now destroyed conjectures are: the Hessian conjecture, the Poisson conjecture, the Dixmier conjecture, the cubic-homogeneous and Drużkowski forms of the Jacobian conjecture, and Zhao’s vanishing conjecture. This is because the counterexample to the Jacobian conjecture can either be transformed directly into counterexamples to these statements or combined with established equivalence and reduction theorems. (Also, I have no idea what I’m doing, but thanks gpt-5.6 Pro.)

[Sebastian Harm @HarmSebastian]

This century is the century of Computer Science and Mathematics. Both fields will accelerate human progress in a way we never have seen before.

[gfodor.id @gfodor]

the stochastic parrots are getting pretty lucky

[Sebastian Harm @HarmSebastian]

@goodhunt Probably the most funny thing I saw today

[levent @__alpoge__]

hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final

((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)

[Sebastian Harm @HarmSebastian]

AI being able to help disprove the Jacobian Conjecture is one of the great examples for meaningful AI usage.
But it proves one thing even more: AI is being able to produce new output and does not just bring up stuff already written somewhere. It disproves many critics of it by doing so. Congrats to @AnthropicAI and Claude Fable 5.

[Sebastian Harm @HarmSebastian]

@__alpoge__ Congratulations! 🥳

[Sebastian Harm @HarmSebastian]

Dear @EU_Commission,

I would like to see a better digital strategy than just fining US or Chinese companies to get some money from their innovations.
Maybe a European initiative for a big tax relief and less government intervention, less bureaucracy and regulations (especially from the European side)? We have innovative minds, we can do it.

[The Kobeissi Letter @KobeissiLetter]

Competition between US and Chinese AI models is heating up:

The proportion of tokens used by US firms that run through Chinese AI models on OpenRouter is up to a record ~58%.

OpenRouter is a platform that allows developers and companies to access and compare AI models from multiple providers, making it a useful gauge of real-world AI model adoption.

This percentage has TRIPLED since mid-January, overtaking American peers on the platform for the first time in March.

The token share of Chinese models briefly rose as high as 63% in the first week of July.

By comparison, Chinese models accounted for less than 10% of usage at the start of 2025, while US models accounted for ~80%.

DeepSeek has emerged as the most popular choice among American firms in recent months.

Chinese AI models are rapidly gaining market share.