THEFUTURE
When the CEO Says Don’t Trust the Product
In this issue: A Ukrainian journalist talks AI workflow optimization between Russian attacks. Google’s CEO tells the BBC not to “blindly trust” AI, then launches Gemini 3 everywhere. Isabel Lerch from NDR on AI hallucinations. Plus: Building games in minutes.
Where I’ve been: There’s this thing my colleague says when journalists get too intense: “We’re not saving lives here.” Usually, he’s right.
But at the Baltic Media AI Forum in Riga, that mantra felt different: Kateryna Noshkaliuk from Ukrainian media network Rayon talked about AI tools and workflow optimization. Standard conference stuff, except for the part about Russian attacks knocking out their power.
Her talk was called “The Role of AI Technologies in Wartime.” When the Q&A started, before anyone asked anything, they thanked her. For defending Europe. For doing journalism under fire.
After a full day of examples from across the Baltics and beyond, I closed the conference talking about how media can position itself in an AI-mediated ecosystem. Great event, and huge respect to all the amazing colleagues!
What we’re talking about: Perfect timing! Google releases Gemini 3, and right before that, the BBC interviews CEO Sundar Pichai. He says what’s usually hidden in fine print: Don’t “blindly trust” AI – these models are “prone to errors.”
Wait, what? The guy whose company is about to flood the internet with more AI tools is telling us not to trust them? His solution: Use Google Search too. How convenient that Google is integrating Gemini 3 into Google Search and has launched an API to feed AI models web facts. Because nothing says “confidence in our product” like immediately suggesting you double-check it with our other product.
But enough snark—Gemini 3 is impressive. My AI colleague Rignam Wangkhang from CBC casually built a game in no time. Coding with Gemini works not just in chat or the command line, but also through a new IDE called Antigravity.
What else I’ve been reading:
And now: At Spiegel, we’ve already wrapped our annual AI Week – workshops, discussions, the works. NDR, the public broadcaster, has theirs coming up. When Isabel Lerch invited me to speak on a panel about AI’s impact on media and society, I made her a deal: I’ll show up if she does a mini-interview for the newsletter. So in this issue, meet this fantastic data journalist!
Four Questions with Isabel Lerch
Isabel Lerch is a data journalist with Germany’s public broadcaster NDR.
What's one fact about AI that everyone should know?
AI hallucinates! AI hallucinates! AI hallucinates! And always will!
How can we better understand the current AI hype?
As a LinkedIn-Addict, I am definitely a huge fan of certain channels, content creators and AI experts. For example, I really enjoy the hands-on newsletter by German AI expert Jan Eggers (“KI-Häppchen”). I also enjoy reading posts by AI expert Elisabeth L’Orange. She is really good at explaining new trends in the AI world, always keeping up with the pace and asking the right questions. A newsletter I really enjoy reading the AI newsletter by AI expert, speaker and podcast host Gregor Schmalzried as he always takes on a more philosophical and intellectual approach on AI topics. And lastly, I highly appreciate the summaries provided by the Handelsblatt “KI-Briefing” – a weekly dosis of the most important AI news regarding business trends, investments and AI companies.
Is there something you wish you'd built yourself?
Yes! Newspapers such as Zeit, Spiegel or Handelsblatt are allowed to create calculators for their audience. For example: Spiegel created “Gehaltsrechner 2025” – a super useful tool which is, I assume, very popular among the audience. It attracts attention and promotes interaction – and it is just an overall very exciting journalistic tool to use. But unfortunately, working for a public broadcaster, we as a data-driven team are not allowed to use these kind of tools.
Are we taking AI seriously enough in the media?
Yes – but this is not necessarily represented in the financial means we put into the monitoring, knowledge-building and implementation of AI. This is a general problem in media – there is so little money at the moment that it is difficult to push a reasonable chunk into the development of AI. So for media outlets, it is super difficult to balance out the investments needed to keep successful, existing formats and products alive and develop AI skills, knowledge and tools at the same time. Simply because it does cost a lot of money and time.
A very good post from vasilenko93 on Reddit:
This is THEFUTURE.
