Gergely Kövesd

Arguments look different when characters must live with the consequences. THE FRICTION POINT, 2026. Being-in-the-world: what AI still lacks. Essays: Substack.

[Gergely Kövesd @gkoevesd]

THE FRICTION POINT, a techno-thriller: superintelligence in a jar, or machines that learn by acting in the physical world?
Wittgenstein and Heidegger, but with a collapsing bridge and one pilot wired to a thousand drones.
Out August 2026. Reviewer copies: https://t.co/88Mjp1VBPh

[Gergely Kövesd @gkoevesd]

@PAHoyeck Fight with love!

[Gergely Kövesd @gkoevesd]

Note the flat, empty floor. The fancier demos add obstacles, but the robot knew them in advance. Both are choreography. Being-in-the-world starts where the script ends.

[China pulse 🇨🇳 @Eng_china5]

China-developed AGIBOT humanoid robots feature ultra-smooth motion control. Every movement is lifelike and seamless, from subtle gestures to complex athletic acrobatics.

[Gergely Kövesd @gkoevesd]

Pre-order the ebook here:
https://t.co/BD4gsC10TB

[Gergely Kövesd @gkoevesd]

THE FRICTION POINT, a techno-thriller: superintelligence in a jar, or machines that learn by acting in the physical world?
Wittgenstein and Heidegger, but with a collapsing bridge and one pilot wired to a thousand drones.
Out August 2026. Reviewer copies: https://t.co/88Mjp1VBPh

[Gergely Kövesd @gkoevesd]

@TrueAIHound We’re being played here. The fancy movement of this humanoid are performed on a perfect clean space with zero obstacles.

[Gergely Kövesd @gkoevesd]

@PAHoyeck Wittgenstein always ate bread and cheese.

[Gergely Kövesd @gkoevesd]

@TrueAIHound @lu_blacktron Wittgenstein pointed out there’s "no private language." Human thinking and intelligence are inherently social.

[Gergely Kövesd @gkoevesd]

@TrueAIHound Descartes’ “Cogito ergo sum” assumes an isolated, thinking individual. Heidegger & Wittgenstein strongly opposed that view. Viewing a super-intelligent "thing" in isolation is still just stuck in that Cartesian mindset.

[Gergely Kövesd @gkoevesd]

@tunguz You are responsible for rogue AI.

[Gergely Kövesd @gkoevesd]

@PAHoyeck Forza Wittgenstein!

[Gergely Kövesd @gkoevesd]

@karlmehta You are mixing the terms ‘agency’ and ‘AI agent’. Current au agents don’t have a agency, they follow a goal (and evals) that is set for them.

[Gergely Kövesd @gkoevesd]

@ArtemisConsort You assume that ‘scaling is all you need’. This assumption can turn out to be wrong. It’s actually quite likely considering that a simple fruit fly can beat the real world intelligence of the current top class robots.

[Gergely Kövesd @gkoevesd]

@ArtemisConsort xAI is not grounded in physical truth. The same LLM tech like OpenAI and Claude just below their capability level.

[Gergely Kövesd @gkoevesd]

@rohanpaul_ai 250% measured in what, LOC?

[Gergely Kövesd @gkoevesd]

@ToKTeacher LLMs are brains in a jar. They have inherent epistemological limitations.

[Gergely Kövesd @gkoevesd]

@oprydai There is no such thing as a private language (Ludwig Wittgenstein)

[Gergely Kövesd @gkoevesd]

@alexisgallagher I agree that current AI doesn’t have agency. But isn’t agency one of the most important things we should be trying to understand, for our own sake?

[Gergely Kövesd @gkoevesd]

@itsolelehmann SMR or fusion is your reactor of choice?

[Gergely Kövesd @gkoevesd]

@LedermanHarvey @AnthropicAI @nyuniversity @UTAustin "The philosophers have only interpreted the world, in various ways. The point, however, is to change it." (KM)

[Gergely Kövesd @gkoevesd]

@yacineMTB Blessed times! Enjoy.

[Gergely Kövesd @gkoevesd]

@GaryMarcus LLMs are strongest where they have huge training data (e.g. coding), but way weaker when forced to use RAG. But they have an even more fundamental limitation than that.