Romano Roth

5/ AI is a new pair partner, not a reason to stop pairing your people. That is still where the breakthroughs come from. The book: buff.ly/FdEjM6f Study: buff.ly/wVEju8Y

4/ But the breakthrough ideas, the top 10%, only came from a human pair plus AI. And people shaped the AI output, they did not copy it. About 19 prompts each.

1/ In my book I called working with AI a form of pair work. A pre-registered field experiment with 776 people just backed that up.

5/ We unpack all of it in Episode 2 of Prompt & Proper, our German-language podcast on AI in the enterprise. New episode every two weeks. buff.ly/IKYG936

4/ The macro math: over 700 billion dollars into AI this year, 1.4 trillion with infrastructure, against roughly 150 billion of revenue. This bubble will pop. Our answer: an AI gateway that routes every request to the right model. Cheaper models trail frontier by about a month.

1/ Your feed says agentic coding makes developers 10x more productive. We measured it across real projects: 20 to 30 percent over the whole value chain. Documentation goes 80 percent faster. Complex work? 20. The factor 10 is marketing.

5/ Is your AI number adoption or transformation, and would your people tell you the truth about which? Interview: buff.ly/MWxceyQ

4/ I've been making this argument for years: the next competitive edge isn't AI itself, it's the organisation around it. The model was never the hard part. Culture, process and trust are.

1/ Your AI adoption number looks great. That might be exactly why nothing has changed. Zapier's transformation officer drew the line most leaders skip: adoption and transformation are two different games.

5/ Before you trust an AI because its explanation sounds reasonable: are you reading its reasoning, or just its output? Paper: buff.ly/MiIwaMk

4/ As a Chief AI Officer I read this as good news. If a model's explanation can differ from its real reasoning, trusting that explanation is trusting a story it tells about itself. Trust needs verification that reaches below the answer.

1/ AI does not always mean what it says. And now we can partly see the gap. Anthropic just built a lens that looks inside the model, past what it prints on the screen.