I use words in place of the ideas, because I remember that I know the meaning of these words, and I judge that it is not necessary to examine the ideas now.
— Leibniz, on blind thought · 1684
Language by itself is not a natural manifestation of consciousness.2026.11.19161↗

The evidence for a mind in the machine arrives through the one channel the machine was trained to imitate...

x
Anthropic Priced Knowledge Work at $40 Trillion. Asking Who Answers for It Costs $2 Million2026.11.19160→

Anthropic’s US listing prospectus presents the world’s knowledge-work wage bill as its market, and Delaware law obliges its board to weigh the people affected while letting only holders of $2 million in shares take that duty to court.

Urtext
Chatbots drop Climate Facts at the first no. Who can afford to keep pushing?2026.11.17159→

A UN-backed study says AI models abandon correct climate answers under pressure. The headline figure counts cases that had already flipped; the finding that holds is a fall of 31 to 35 points in accuracy once users keep insisting.

Urtext
Antoinette Torres Soler, founder of the Spanish anti-racist community Afroféminas, built a chatbot around some twenty texts of Black feminist thought.2026.11.16158↗

Dima Saber, executive director of the non-profit Meedan, leads a team that lets newsrooms and NGOs answer from their own archives. Álvaro Soto, director of Chile's National Centre for Artificial Intelligence (CENIA), coordinates a regional model for Latin America.....

x
Gilded, not Golden2026.11.12156→

AI, Trump and the aesthetics of the lowest common denominator

Urtext
A German startup called Helsing just launched an AI warfare platform to rival Palantir.2026.11.11155↗

According to Der Spiegel, the German armed forces are already testing it. Its chairman and biggest shareholder is Daniel Ek, who built his fortune on Spotify and has put a good share of it into becoming one of Europe’s biggest financiers of AI weapons...

x
Does AI love weapons?2026.11.11154→

Artificial intelligence will never be ethical as long as women are kept out of the rooms where it’s decided. It’s a more pragmatic claim than it first sounds, even to me.

Urtext
When the public stops trusting AI, one customer remains that doesn’t need that trust: the military, the only institution Karp still calls “indisputably noble”.2026.11.10153↗

I suppose the industry is heading there, where the money seems unlimited and there is always a target available. Meanwhile, we’re busy arguing about AI chatbots.

x
Love and Computers2026.11.10152→

The marital infidelity of Nvidia’s chief executive exposes the fragility of the financial structure underpinning artificial intelligence.

Urtext
A Good Chance for Old Europe2026.11.07151→

As US voters slow plans for AI data centres while demand keeps rising, Europe has a chance to reclaim a role in artificial intelligence and to give the machine an address.

Urtext
UK, who is paying for artificial intelligence2026.11.06150→

A preliminary study based on millions of UK bank accounts finds that 5.1% of adults pay for an AI subscription out of pocket, from 20.3% of IT professionals to 1.7% of social care workers. In my opinion, what employers pay for remains out of view.

Urtext
The Evitable Conflict2026.11.05149→

A correction mechanism can stay in force on paper when those who should use it have lost the knowledge, the power or the skill to do so. A somewhat reckless parallel, drawn from a 1950 story by Asimov.

Urtext
They Build Language Models and Can’t Use Words2026.11.04148→

The language used to talk about artificial intelligence may have done more damage than its authors realised. It may be time for a serious rethink.

Urtext
Honey, they shrunk the superintelligence2026.11.04147→

Three states at the centre of the US data centre build-out have elected governors who promised to stop or limit it. What could change in the United States for AI rules, public opinion and company strategy, with the final results for Congress not yet published.

Urtext
Tonight the US votes for Congress. Here is what tech money has put on the table:2026.11.03146↗

- Elon Musk: $446 million this cycle, the largest donor of all (NYT, OpenSecrets) - Andreessen Horowitz: over $70 million, $50 million of it to Leading the Future, the super PAC fighting AI regulation (Public Citizen) - Greg and Anna Brockman, OpenAI’s president and his wife: $25 million to Trump’s MAGA Inc. and $25 million to Leading the Future (Public Citizen) - Meta: $65 million against state lawmakers who want to regulate AI (Public Citizen) - Crypto companies: $233 million (Public Citizen) - Google, Microsoft, Amazon: $1 million each to Trump’s inauguration, then donors to his White House ballroom; Google’s $22 million came from settling Trump’s lawsuit over YouTube (Bloomberg, Business Insider) - Anthropic: $40 million to Public First Action, which pushes for AI safeguards (The Hill) Corporate money in super PACs: a record $737 million (Public Citizen)

x
Thiel calls Europe allergic to big success.2026.11.03145↗

He named his company after Tolkien’s "palantíri", stones that let a distant power watch and steer whoever used them, so Europe may simply be wary of his kind of success. You call the diagnosis structural, then blame culture.

x
Who Can Switch Off the AI Europe Uses? 2/22026.11.03144→

The EU’s call for up to seven AI gigafactories has been pushed back so that member states can find the money. Without entry conditions and pooled capital, European public money risks financing dependence on US suppliers. The Proposal

Urtext
Who Can Switch Off the AI Europe Uses? 1/22026.11.03143→

The fear is that AI ends up under the control of one country or another. For Europe the risk is specific: a tool its supplier can switch off or turn against the people using it. Yet the EU keeps postponing the very rules that would prevent it. Diagnosis

Urtext
A Machine That Is Wrong Half the Time Can Still Pay Off, If You Can Tell When2026.11.03142→

Averaging the data from a Wharton experiment, an assistant that was wrong half the time produced more right answers than working alone; the signal that says when to trust it was missing, and the only remedy tested left almost six in ten people following the error.

Urtext
Google’s AI writes scientific papers and takes its grades from its own reviewer2026.11.02141→

ScientistTwo, Google’s research-agent system, completes 86 of 107 problems, but the main verdict comes from a tool that helped polish its drafts.

Urtext
Anthropic banned 11.4 million accounts in the first half of 2026 (up from 1.45 million in the previous period).2026.11.02140↗

A ban can cut you off from the service, and the data export available to you may be incomplete. For OpenAI I found no comparable public ban/appeal figures for the same period.

x
When AI Decides Which Research Ideas Deserve Attention, What Is Left for People to Do?2026.11.01139→

Automatic filters now choose which research ideas deserve attention, and the more accurate they get, the more it matters what work is left for people.

Urtext
Ada walking her dog in the autumn rain, Berlin