On 27 September, MTV invented an award that had never existed and gave it to Taylor Swift. Two days later the President of the United States signed an executive order renaming artificial intelligence.

One of those worked instantly and permanently. The other one did not work at all.

Both are the same kind of act: somebody with institutional standing saying a thing is so in order to make it so. John Searle called these declarations, and they are the only sentences that change reality rather than describe it. I now pronounce you married. You're fired. The defendant is guilty. They work, but only where the speaker has recognised standing and the audience collectively accepts it.

MTV has total authority over what constitutes an MTV award, so the Artist Director Honors existed from the moment they announced it and will be in the record forever. A president has enormous authority over federal agencies and none whatsoever over a global research field, which is why the executive order directing agencies to "no longer acknowledge" the terms AI and artificial intelligence changed how government documents are written and nothing else. PolitiFact checked. Industry largely carried on.

The detail that turns the order into comedy: it defines "Super Intelligence" as having the same meaning as "artificial intelligence" under existing statute. Same definition, new letters. And the serious version of the joke, which Poynter made: superintelligence already denotes a hypothetical system exceeding human cognition in nearly all domains. Assigning that word to current technology degrades the vocabulary you would need to regulate the future one.

Then two stories where declarative power mattered considerably more.

Germany's federal prosecutor arrested August Hanning, who ran the BND from 1998 to 2005, on suspicion of treasonous espionage. Allegations only, and he is entitled to the presumption of innocence. The part relevant here is how it surfaced: inside an investigation into something else entirely. The institution with the authority to name a traitor apparently went sixteen years without knowing there was anything to name.

And Thailand, where I live, has been flooding since mid-September. More than three million people affected across twenty-one provinces and Bangkok. Emergency cell broadcasts reached residents after the heavy flooding was already underway, and a deputy interior minister had to publicly ask why the existing system had not been used.

An evacuation warning is a declaration in exactly Searle's sense. It converts weather into an official emergency and changes what millions of people should do in the next two hours. It is the most consequential version of the thing this whole episode is about, and it is the one we are worst at.

Ranked by institutional power: the state, the state, the state, and then a cable channel. Ranked by how cleanly the declaration took effect, that order reverses completely.

Between the first of June and the end of August, political accounts in Saxony-Anhalt collected around 43.4 million views on TikTok. Sixty-four and a half percent of them went to a single account.

On Sunday the party behind that account took 43.8 percent of the vote, more than doubling its 2021 result and recording the strongest far-right performance in any German state election since the war.

I'm German, I teach media and communication studies, and this episode is not about how that feels. It's about how it was built.

The lead candidate did 28 million views in three months, out-performing both his own party's federal parliamentary group and its national co-leader. But the finding that matters comes from research on last year's federal campaign: the visibility was driven mainly by ordinary users rather than party accounts. Which is Katz and Lazarsfeld's two-step flow from 1955, rebuilt at scale, and it explains why suspending an account achieves precisely nothing.

Then Noelle-Neumann, because the best available explanation for Sunday was written in Germany in 1974. The spiral of silence assumed people read the climate of opinion from the room around them. That instrument has been replaced by a personalised feed. And at 43.8 percent in a single state, the spiral inverts: the people managing a fear of isolation are now everybody else.

Also in here: why a parliament writing speeches for vertical video is Postman arriving literally. Why the Brandmauer was never a wall but a seat count, and what a 39 to 39 to 5 parliament does to a doctrine that only works while the excluded party stays small. Why the state's own constitutional protection office classifying this party as right-wing extremist moved nothing at all, and what that says about institutional authority when the audience stops granting it.

And one argument with my own field, about the recurring claim that these messages reach young people who lack background knowledge. Under-24 support tripled in a single federal cycle. If your explanation is that they didn't know enough, you haven't explained anything. You've consoled yourself.

None of this required anybody to be stupid. Most of it was documented in advance, in public, by researchers, in German.

Five questions that will ruin every app you love, and a sixth one that matters more than all of them.

I teach a platform teardown framework to my Digital Strategy students. Five questions, run them on any company, get an uncomfortable answer in ten minutes. This episode is those five questions, anchored to three stories from the last six months:

Grab finally turned a profit. Thirteen years after founding, $200 million net, on $22 billion of transaction value. The "strategic subsidy" everyone defends as a bet took thirteen years to pay off, and the drivers who funded it own none of it.

Spotify's royalty pool is a transfer, not a payment. No payout under 1,000 streams a year. The roughly $40 million that would have gone to the smallest artists gets pooled and redistributed to the biggest ones. There's a lawsuit in Connecticut. There's a case of AI speed-altering taking 94% of an artist's royalties. Asymmetric pricing working exactly as the theory describes.

OpenAI rebuilt Google Shopping inside a chat window. Ads since February, a self-serve Ads Manager since May, product carousels pulling straight from retailer feeds. $2.5 billion projected this year, $100 billion targeted by 2030. Paid tiers stay ad free, obviously.

Then the turn. This month Mark Zuckerberg published 6,500 words called "The Future is for Everyone," on the same day Meta shipped a new open-weight model. My own teaching slides contain the rebuttal, and it comes down to one substitution: the word "open" meaning a property of a file, standing in for the word "open" meaning a property of a world.

Open weights don't make an open internet. Openness lives in architecture, not artifacts. And a company spending $130 billion in capex doesn't give its best model away out of generosity, it does it because free was never the opposite of owned.

The sixth question is: can you leave? Not would you like to. Can you. Today. Taking your work, your audience and your history with you. If the answer is no, everything the first five questions found isn't a business model. It's a position of power over you.

Long-form version of the second half: "The Manifesto and the Machine" on Substack. Framework at realityinbeta.com.

I started teaching a new semester this week. Digital Strategy, week one, the same three theories I open with every year. Ten minutes in I realised I wasn't teaching theory anymore. I was reading the news out loud with citations attached.

Three frameworks, three stories happening right now:

Platform economics. OpenAI built the most talked-about product of the decade, has hundreds of millions of users, and is reportedly losing billions a year. Microsoft has the distribution. Every startup on earth is building margin on top of infrastructure OpenAI runs at a loss. Creating value and capturing value turn out to be completely different jobs, and OpenAI has so far only done one of them.

Network effects. Everyone announced they were leaving X. Repeatedly. With screenshots. It's August 2026 and X is still where news breaks. Bluesky is growing genuinely well and has changed nothing about X's position. Because leaving a platform was never a decision about the platform. It was a decision about everyone who hadn't left yet.

Enshittification. Cory Doctorow's four-stage model, currently being performed by the entire AI industry in synchronised formation. ChatGPT went from free and unlimited in 2022 to usage caps and tiered pricing in 2026 and nobody was surprised. Meanwhile Google Search hit stage three so slowly that we all just adjusted our expectations downward without noticing: AI Overviews answer your question using publisher content so you never visit the publisher.

None of this tells you what to do. That's not what theory is for. It tells you what you're looking at, which turns out to be the harder part.

Two stories this week that look completely unrelated. A presidential phone call to a football governing body. A celebrity wedding at Madison Square Garden with zero authorised photos. Both ended the same way: the people trying to control the story lost control of it completely.

FIFA reversed Folarin Balogun's World Cup red card ban after Trump personally called Gianni Infantino. No explanation given. First time in 64 years a player avoided a suspension after being sent off at a World Cup. The USA then lost 4-1 to Belgium anyway. Meanwhile FIFA's own refereeing chief told the press that "nobody can claim FIFA can be influenced by anyone, not even by the FIFA president." He said this. During this tournament.

Also: Argentina vs Egypt. Egypt lead 2-0. Argentina come back. Egypt's goal disallowed by VAR on a borderline call. Egypt's coach says they were cheated. Half the football world agrees. The other half says it's conspiracy thinking. Both halves are arguing about it now, which is its own data point about FIFA's credibility.

Taylor Swift married Travis Kelce on July 3rd. Total media lockdown. No official photos. What happened next: AI-generated wedding images flooded the internet within hours, convincing enough that outlets needed Google DeepMind's SynthID tool to debunk them. Her wedding became the first celebrity event experienced primarily through AI misinformation. Then came the ICE contractor guest controversy. Then Trump. Then the response that became its own story.

This episode is about what happens when institutions and individuals try to own their narrative completely. Spoiler: it usually makes things worse.

The World Cup is back. 48 teams, three host countries, billions of eyeballs, and — somehow — Saudi Aramco as an official sponsor. Sure.

This week we go good, bad, and ugly on the 2026 FIFA World Cup through a media and communications lens. The good: why the World Cup's claim to global shared attention is actually different from every other event that gets called "the last global media event" (yes, we know we say that every time). The bad: what it means for the world's most international tournament to be hosted by a country that has spent four years making it structurally difficult for international people to enter. And the ugly: sportswashing as a communications strategy, FIFA's governance that got quieter rather than cleaner, and what Guy Debord would make of a match ball that is also a brand.

The football is still in there. Finding it just requires looking past quite a lot of other things first.

I'm back from Berlin. Three days at re:publica — Europe's biggest digital culture conference — and I have thoughts.

The good news: the room was packed. People sat on the floor and stood in the doorway to hear a talk about digital dissent in Southeast Asia. People genuinely care about the open internet, democracy, and what happens when platforms eat themselves. There was even a hater in my Q&A, which means at least someone was paying attention.

The complicated news: caring about the right things and building a coalition big enough to actually change them are two different skills. This episode is about the gap between them — and why the people who understand the game best sometimes struggle most to grow the team playing it.

Also: Cory Doctorow, enshittification, and why I'm still wearing the entrance wristband a week later.

This past week three things happened. The Devil Wears Prada 2 opened globally to $234M — a film about an industry that sacrifices craft for speed, made by sacrificing craft for speed. The first half and second half feel directed by different people, because the production more or less was. Ted Turner died, aged 87 — the man who invented 24-hour news in 1980, closed the gap between events and their broadcast to zero, and accidentally created the model that turned urgency into a default register rather than a signal.

And re:publica, Europe's largest digital culture conference, announced its 2026 theme: Never Gonna Give You Up. The Rickroll. Inverted. You know it's Rick Astley. That's the point.

In this episode, Sascha connects all three — using auteur theory, Neil Postman's Amusing Ourselves to Death, Walter Benjamin, and the political economy of media ownership — to ask what it looks like when media gives up on coherence, depth, and craft. And what it looks like when it refuses to.

Recorded before heading to Berlin to speak at re:publica on gamification, dissent, and the digital frontlines of Southeast Asia.

FUNK !T — media and communication theory for what's actually happening this week.