In this issue: OpenAI’s Pulse turns ChatGPT into a recommendation engine that nudges you every morning. Is this enshittification or just catching up with the competition? Switzerland’s radically open AI model Apertus. Journalist Santina Russo on why transparency matters. Plus: My thoughts on AI strategy at SPIEGEL.
What we’re talking about: OpenAI is introducing a discovery feature called Pulse in ChatGPT. It collects topics from previous chats and presents users with cards containing updates and ideas the next morning. So no one has to stare at an empty chat window anymore.
The passive chatbot tool is becoming a recommendation engine with something new every day. In the future, users won’t be asking the chatbot—they’ll be the ones getting asked, reminded, and encouraged. Users are meant to get used to it and develop routines.
This creates dynamic, personalized spaces perfect for business: partner offers, e-commerce, advertising. For now, Pulse is reserved for paying Pro users—but the vast majority of ChatGPT users don’t pay yet, even though they still generate costs.
Is this already the enshittification that all platforms eventually embrace? First, put up a sleek, clean interface, get the masses excited, and then finally have to think about making money? On the other hand, Google, Perplexity, and many others already have discovery spaces and features built into their products, opening up various possibilities for meaningful media partnerships.
The New York Times’ Zach Seward on AI newsroom strategy: “We’re not trying to be AI boosters. In fact, quite the opposite. I think there’s a lot of caution. A lot of time we spend cautioning people about uses of AI, both [in the] legal and editorial senses.” (Sara Guaglione, Digiday)
And now: There’s so much talk about AI dominance—the US here, China there, Europe practically left behind, sure there’s Mistral, but the investors… and then Switzerland shows there’s another way: completely open. The AI model is fittingly called Apertus, and it can partially compete with Llama. More importantly: training data, weights, everything’s open. Right in the middle of it: journalist Santina Russo, who spends part of her time embedded at the Swiss National Supercomputing Centre.
Three Questions with Santina Russo
Santina Russo is a science journalist—part-time at the Swiss National Supercomputing Centre CSCS and part-time a freelancer.
Is there a feature of AI that you’re jealous of?
I am a writer by trade so I’m generally jealous of the sheer speed with which AI produces text – less of the skill of the writing though, at least until recently. Only recently I realized how well AI can create original content when asked with a well crafted series of prompts, especially when you let the AI assess its own output critically, and use that assessment to improve its output. While I know that AI cannot really be creative based on the core of what it can do, which is to merely put one word after another, it can simulate creativity rather convincingly.
What’s the one fact about AI that everyone should know?
Be aware that almost all providers of large language models keep the training data and how the models were fine-tuned a secret. Notably, this applies to the most popular LLMs from the US and China. Which means that the AI models may be and almost certainly are purposefully biased in terms of political and societal view. Example: just this summer Donald Trump issued an executive order promoting radical freedom of expression in American AI models, with which he meant that the training data should be free of what he calls wokeness and political correctness. Instead, factually wrong claims, e.g. denying climate change, should be reproduced unfiltered. Not what I would call any sort of reasonable basis for a high quality AI. In contrast, Apertus, the model recently released by ETH Zürich, EPFL and CSCS in Switzerland, is purposefully open and transparent with everything concerning training, fine-tuning and post-training of the model. To my knowledge, it is the first LLM to fully comply with the EU AI Act.
What’s a good hobby to pick up?
Tennis. First of all it’s fun. And it’s interesting because it’s not just physical exercise but also a mind game, especially when you play matches. For me, playing matches often involves the challenge to let go of my edginess, frustration or anger at myself, and instead face each new point with a positive attitude. I kind of hope that this mental exercise will make me more relaxed not only on the tennis court, but also in everyday life. Apart from that, Tennis is the healthiest sport there is, as several studies found: people who play recreational tennis live longer than those who practice any other sport.
What’s our approach to AI at SPIEGEL? Lennart Schneider invited me onto his podcast, Subscribe Now, and I really enjoyed the conversation. It runs 81 minutes, which is admittedly quite long—but I’d like to think it offers nuance rather than flashy soundbites. If you’re interested in learning more but don’t understand German, you could upload the podcast to Google NotebookLM and explore topics using a mind map.
What can be done about hungry bots? The taz—Germany’s left-leaning, alternative publication I wrote for many years ago—interviewed me about crawlers that scrape knowledge from across the internet. Moritz Müllender thesis: The internet is “degrading” into a chatbox. I cautiously asked: What if a chatbox is actually more practical for users than a website overloaded with pop-ups? That’s exactly the one quote that made it into the article. And as a punchline, they cite studies claiming that critical thinking declines among AI users. Touché, dear taz.
Another technical protocol: The excellent and thoughtful radio show Breitband covered RSL, a new protocol for bot access to websites that theoretically could enable fair compensation. I talked to Pia Behme and remain cautiously skeptical—I find long-term customer relationships and the ability to transfer and link subscriptions more interesting than micropayments.
Coming up next: I’ll be speaking at Google’s Baltic News Publishers Day on October 16th in Riga and at the News Product Alliance Summit 2025 in Chicago. If you’re also there, feel free to say hi and let’s nerd about AI. Or go for a run.
This is THEFUTURE.
Subscribe to THEFUTURE
Get this newsletter in your inbox. For free.
You want to join THEFUTURE. Please check your mail for confirmation.
