AI Across The Product Lifecycle Podcast

The episode contains solid practitioner insights about AI's impact on productivity, the shift from code-writing to product-thinking bottlenecks, and the integration of symbolic AI with LLMs for optimization problems. However, much of the discussion retreads familiar ground (AI makes you faster, bottlenecks shift), and there is considerable filler around company origin stories, reindustrialization policy, and conference plugging that dilutes insight density.

“the bottleneck has shifted from pure tech, meaning writing code, which was essentially a manual job. Uh, and the bottleneck shifted to product thinking.”

“we can deliver in two months what we could deliver in two years, uh, you know, a couple of years ago.”

The guest framing of hardware engineering as needing a 'Claude Code moment' is reasonable but not particularly original. The observation about LLMs being probabilistic while manufacturing is deterministic is somewhat fresh. However, most core ideas - AI as a tool, shifting bottlenecks, the need for data infrastructure - are well-worn in tech discourse. The comparison to Excel's impact on accounting is a common analogy.

“we have to get rid of that idea...the factory of the future will be extremely technical with heavy investment, heavy automation. Probably 10, 15, 20% of the cost of the factory will be into software.”

“I think the real, the bottleneck is the fact that in both supply chain and engineering manufacturing, we're extremely deterministic, I mean, obsessively deterministic. And LLMs are unfortunately obsessively, uh, uh, probabilistic.”

Both guests are founding CEOs of venture-backed startups solving real manufacturing problems at the bleeding edge of AI application. Matthias has deep operational experience deploying at factory floors, and Thibaut has worked at McKinsey and clearly understands supply chain complexity. They are practitioners, not just theorists. However, neither appears to have C-suite experience at Fortune 500 manufacturing or to have achieved massive scale yet, which prevents a higher score.

“I'm the CEO and co founder of Cognix, which I define, uh, as an AI engineering platform for hardware.”

“I'm the CEO and founder of oplit, which is an AI supply chain platform for industrial companies.”

The episode lacks concrete numbers, named customer examples, and specific metrics. Claims about 3-5x faster code shipping, 10x more difficult problems solvable, and 'two months vs. two years' are offered without details or proof points. The digital maturity spectrum (1-5) is introduced but never grounded in specific company examples. One reference to McKinsey consulting on factory scheduling is the only concrete work experience mentioned.

“we ship probably 3 to 5x faster in code.”

“we can solve 10x, uh more difficult uh problems and we can optimize what we could not optimize before.”

The host asks reasonable setup questions and occasionally probes deeper (e.g., on organizational structure post-AI, on digital maturity spectrum), but rarely pushes back or challenges claims. Questions about how to attract talent and policy recommendations feel tangential. The host misses opportunities to press on specifics: no follow-up on the 3-5x productivity claim, no pushback on when the 'Claude Code moment' will actually arrive, no real challenge to the 'by 2030' supply chain claim. The conversation feels more like a friendly interview than a rigorous interrogation.

“Um, Matthias, you can pronounce your name correctly so I don't mess it up and you can tell us what Cognix is doing.”

“So if I rephrase what you said, you really want to have supply chain as code.”