Bennett Jones

A principal-level engineer with 13 years of experience posted on Reddit this week that he thinks about quitting software engineering almost every day. Not because AI took his job. Because AI changed what his job asks from his brain. His line: "The cost of writing code in effort / time was a throttling middleware." Writing code used to be slow. That slowness wasn't just friction. It was pacing. It gave your brain time to process architecture decisions, sense when a pattern was wrong, think through implications before the next choice arrived. AI removed that pacing layer. And what's left underneath is a continuous stream of high-stakes judgment. Architecture calls. Quality assessments. Risk trade-offs. All day. No recovery intervals. He said he's making ten whiteboard-level decisions before his second cup of coffee. Decisions that used to happen once or twice per sprint. BCG just published a study in Harvard Business Review on this exact phenomenon. They surveyed 1,488 workers and found that high AI oversight was associated with 14% more mental effort, 12% greater fatigue, and 19% more information overload. They called it "AI brain fry." Workers described a fog that forced them to physically step away from their screens. This isn't a coding problem. It's a knowledge work problem. Every profession that adopts AI tools effectively will hit this same wall — the production layer compresses, the judgment layer concentrates, and nobody built the cognitive infrastructure to handle sustained decision-making at that density. We spent two years celebrating AI's ability to remove the production burden. We haven't spent two minutes thinking about what happens when the production burden was also the only thing giving our brains time to recover between hard decisions. Your brain was always a judgment machine. AI just made it the only machine that matters. And we have no idea what it costs to run it all day.

On my way to Prague, and very happy to be participating in #PAD2026 . I get a chance to combine two of my favourite things: international arbitration and my Czech heritage. It’s not often that my family name is presented with its diacritics — Valášek. As for its proper pronunciation, let’s get into it over a proper beer.🍺 Here’s the important information: I’ll be joining Ondrej Cech , Monika Diehl and Shaparak Saleh for a panel discussion on “Trade Disputes and Geopolitical Developments: Implications for Enforcement”. I’m looking forward to contributing to the discussion. Dekuji moc za pozvání, Miloš Olík ! A brzy na viděnou. Registration: https://lnkd.in/dqsissMi

Yesterday. Chernihiv Ukraine. 16 year old boy killed One of many attacks on civilians yesterday by russia. Does anyone in the west even notice anymore?

Great to see a transaction of historic proportions come together with such strategic collaboration.

And people like JD Vance celebrate denying Ukraine aid to defend themselves against this…..

Forty-four percent of Gen Z workers admit to sabotaging their company's AI rollout. Not quietly resisting. Sabotaging. Feeding proprietary data into unapproved tools. Deliberately producing poor output to make AI look bad. Refusing training entirely. The conventional response: blame the generation. More training. Stricter mandates. Threaten consequences. But buried in the same study is a number that reframes the whole thing: 26% of employees doing the sabotaging say the company's AI strategy is poorly executed. That's not resistance. That's feedback. And the Wharton School just confirmed it. Their latest enterprise AI study found a widening gap between executive enthusiasm and manager reality. Two-thirds of executives have become "much more positive" about AI. The managers running the actual workflows feel unsupported and overwhelmed. Sixty percent of executives say they'd cut employees who refuse to adopt AI. But only 14% of enterprises have a clear AI strategy at all. Three numbers. Same slide deck. And the picture they paint isn't a workforce problem. It's a measurement problem. Most organizations track adoption — seats, logins, prompt counts. A study of 1.4 million AI interactions found that 90% of employees use AI regularly, but only 5% use it in ways that create differentiated value. That's an 18-to-1 ratio between activity and sophistication. And no adoption dashboard in the world can see it. The organizations that fix this will stop measuring whether people use AI and start measuring how they use it. The ones that don't will keep watching their dashboards climb while their best people quietly opt out. I made a video breaking down the research and the scoring framework behind this — link in the comments.

As a Civil Servant, you don't often get to brag about what you've been working on - but I can today. The contract between GBE-N and RR SMR is now signed - a major milestone on the path to delivering our first SMRs. As part of the DESNZ team that has led this work, I've seen first-hand the huge amount of effort across the department and partners to get us here. There's lots more still to do - but this one is worth celebrating. ⚛️