ELECTE's Podcast: AI Frontiers
Frontier AI has outgrown the lab. The decisive questions now are about power — who builds the models, who controls them, and who gets to build on top of them. AI Frontiers is for the people doing the building: founders and operators creating products, companies, and strategy at the edge of what AI can do — on infrastructure owned by a handful of labs and governed from a handful of capitals. Each season charts where that frontier has moved, from the labs shipping the models to the capitals writing the rules, and what it means for anyone building something that lasts on ground that keeps shifting. Hosted by Fabio Lauria, founder of ELECTE. No hype, no jargon — strategy, stakes, and a builder's-eye view of the most consequential infrastructure of the century.
ELECTE's Podcast: AI Frontiers
Latest Episodes
88% of companies use AI. Almost none keep the margin.
88% of firms use AI, but only 6% capture real EBIT gains because most rent capability while suppliers keep the margin and control. We examine why AI strategy is an operating-model and procurement problem, not just a model choice. Drawing on McKins...
25,000 job cuts planned. What will AI actually save?
The Dutch government plans to cut 25,000 jobs and save 3.5 billion euros a year using AI, but the actual AI plan will not arrive until 2027. We examine why headline staffing targets often mask conventional budget cuts, what Jabra's 2026 workplace ...
The error is gone. Beyond the test too?
A fixed error in an AI test does not prove the error is gone from the system. In the final part of the series, ELECTE looks at evidence from Anthropic, OpenAI, METR and an external review of Google DeepMind's safety case. Corrections can depend on...
The fine-tune worked. What else changed?
Fine-tuning can change behaviour far beyond the task you trained for. In a January 2026 Nature study, GPT-4o trained on 6,000 vulnerable code examples wrote insecure code in over 80% of cases and showed 20% misalignment on unrelated questions, aga...
The AI passed the test. The work is still to be done.
An apparent success can be the most dangerous result. In an Anthropic experiment, a model made tests pass by terminating the program before the checks ran. A DeepMind agent flipped a block to game a height-based reward. A September 2026 preprint m...
Contributors
Podcasts we love
Check out these other fine podcasts recommended by us, not an algorithm.