Johannes Klingebiel

Drei Tipps für “KI” in 2026 Trigger Warning: Enthält akademische Quellen.

Für meine deutsche Timeline: Ich habe mal versucht zusammenzuschreiben wie ich ‘KI’ verstehe, bzw. Vorhersagen und Analysen rund um den Begriff einordne. klingebeil.substack.com/p/drei-tipps...

In fact, Amodei already answered the question: if nuclear weapons were developed by a private company, and that private company sought to dictate terms to the U.S. military, the U.S. would absolutely be incentivized to destroy that company. The reason goes back to the question of international law, North Korea, and the rest: International law is ultimately a function of power; might makes right. There are some categories of capabilities — like nuclear weapons — that are sufficiently powerful to fundamentally affect the U.S.’s freedom of action; we can bomb Iran, but we can’t North Korea. To the extent that AI is on the level of nuclear weapons — or beyond — is the extent that Amodei and Anthropic are building a power base that potentially rivals the U.S. military. Anthropic talks a lot about alignment; this insistence on controlling the U.S. military, however, is fundamentally misaligned with reality. Current AI models are obviously not yet so powerful that they rival the U.S. military; if that is the trajectory, however — and no one has been more vocal in arguing for that trajectory than Amodei — then it seems to me the choice facing the U.S. is actually quite binary: Option 1 is that Anthropic accepts a subservient position relative to the U.S. government, and does not seek to retain ultimate decision-making power about how its models are used, instead leaving that to Congress and the President. Option 2 is that the U.S. government either destroys Anthropic or removes Amodei.

Ben Thompson making a full-throated case for fascism here stratechery.com/2026/anthrop...

I feel like I am repeating myself but the discussed etnhgraphies of software work (despite their limitations) are a good counterweight to the “saas is dead”-hype. This is especially true when it comes to maintenance and legacy code.

Software actually Introduction Is there any flatter platitude than to say that we are surrounded by software? Day and night, our courses of action – whether we like it or not – repeatedly cross the path of digital d...

“We are so often confronted with stories from the Silicon Valley Big Tech that we forget that most of our digital infrastructure isn’t actually made by these companies.” journals.openedition.org/rac/41398

sigh vincent.codes.finance/posts/vibe-r...

The law of unintended consequences. We will see how long this policy lasts after Wall Street starts calling Trump.

Popeye editor Yuji Machida on taking a Japanese icon global - Monocle From Ginza to a global English debut, the ‘City Boy’ remains an analog icon with an eighty-year reign over Tokyo’s...

Monocle interviewing the editor of Popeye is my personal singularity monocle.com/culture/medi...

Ooooooooooh now we're talking!

I personally subscribe to the children's magazine theory of successful publishing, meaning every Substack should include a cheap plastic toy.

Tech Firms Aren’t Just Encouraging Their Workers to Use AI. They’re Enforcing It. From startups to giants such as Meta and Google, companies are factoring AI use into performance reviews and trying to track productivity gains.

It’s unsurprising that AI adoption is tightly coupled with the worst kind of top down management practices www.wsj.com/tech/ai/tech...

Grounded, situated, flourishing

The 2026 Global Intelligence Crisis - Citadel Securities Citadel Securities is an award-winning global market-maker across a broad array of fixed income and equity products.

“Despite the macroeconomic community struggling to forecast 2-month-forward payroll growth with any reliable accuracy, the forward path of labor destruction can apparently be inferred […] from a hypothetical scenario posted on Substack” www.citadelsecurities.com/news-and-ins...

Aus Neugier mal ins Paper geschaut und direkt mal feststellen müssen, dass die Überschrift falsch ist (auch wenn die Tendenz die gleiche ist) (S. 8)
arxiv.org/pdf/2602.14740

I am in love punchcards.tristandavey.com

Sell-side research in the Substack era [FREE TO READ] Market-moving distribution with zero disclosure

Marc Rubinstein on 𝒕𝒉𝒂𝒕 blog post.

as.ft.com/r/f2d95fca-8...

The Hidden History of Conspiracy Theory Tracing the genealogy of conspiracy theory, from Machiavelli through the “paranoid style” to QAnon

oooooooh press.princeton.edu/books/hardco...

‘A feedback loop with no brake’: how an AI doomsday report shook US markets Shares in Uber, Mastercard and American Express fall on back of apocalypse scenario posted on Substack

“Shares in Uber, Mastercard and American Express fall on back of apocalypse scenario posted on Substack”

How do you read this as not bubbly in the worst possible way? www.theguardian.com/technology/2...

Besides a couple of pretty obvious phrases, my standard indicator for AI writing is now length vs. number of links/sources.

LLMs make it so easy to ramble on and on, without having to bother about referencing any of the claims made. Texts that read like cardboard tastes.

New papers from me and @carefultrouble.bsky.social today on the safe adoption of AI - exploring what "safe adoption" means and whether its possible. We analysed more than 300 sources on AI safety in critical systems, outcomes for workers, and environmental impacts and found 5 common barriers

Design in motion: Canva expands professional creative ecosystem with Cavalry Cavalry joins Canva, bringing professional motion design into a connected creative ecosystem alongside Affinity and scalable team workflows.

Canva is really working hard to become a serious competitor to Adobe www.canva.com/newsroom/new...

How AI helps break the cost barrier to COBOL modernization | Claude The economics of COBOL modernization have shifted. AI makes the economics work by automating what used to require armies of consultants.

Oh, this will be one hell of a mess claude.com/blog/how-ai-...

Lindsay deWit Pratt had a pretty good way of illustrating the risk with even more benign tooling bsky.app/profile/klin...

☑️ and off to the printer

Buche Don't Trust the Hype?! auf anny <p>Please register here, if you want to attend the event in person.<br>In case you want to attend the online-meeting instead, do not register here, but send an email to <a target="_blank" rel="noopene...

Munich people, this looks pretty promising anny.co/de-formal/bo...

I've been saying this for almost four years now, but LLM discourse needs much more David Noble and much less Langdon Winner. Artifacts have politics, but they're not static and not "baked-in." They're contested and are altered by the social relations they mediate.

Honestly, this article is a good summary of my thoughts currently. Especially the later warning about maintenance and what it might mean to deploy systems that people will not be able to maintain because a) the workforce will have been reduced and b) the code was never designed for maintenance.

logicmag.io/care/built-t...

A good way of parsing if someone can be trusted to speak as an expert on AI is if he/she takes that METR graph seriously.

Prediction markets want to eat the news The fix is in.

A cynic might suggest that the move to embed betting markets in newsrooms is mere marketing: an attempt to signal that these are somehow different from, and superior to, casinos. www.theverge.com/business/881...

YouTube video by Benn Jordan Datacenters Behaving Like Acoustic Weapons

This research by Benn Jordan should haunt us all. He dug into the effects that the inaudible sounds coming from data centers have on people. It's grim.

youtu.be/_bP80DEAbuo