AI Engineer Paris 2026 Opening Keynotes: Mistral, Langfuse & Sizzy | Day 1
The AI Engineer Paris 2026 main stage featured discussions on the historical impact of AI as a general-purpose technology, the evolving landscape of AI engineering towards sophisticated "software factories," and Mistral AI's strategic vision for building frontier AI, ensuring agentic safety, and addressing data sovereignty through advanced models and infrastructure.
The AI Engineer Paris 2026 main stage commenced with welcome remarks and insightful talks spanning AI infrastructure, coding agents, software factories, and the development of frontier AI.
Clemens Rawert (Langfuse): What Can Economic History Teach Us About the AI Moment?
Clemens Rawert initiated the main stage by framing AI within the context of economic history, particularly its role as a potential General Purpose Technology (GPT). He posited that long-run economic growth is primarily driven by productivity increases, which are, in turn, fueled by technological and organizational advancements. Rawert highlighted the “AI productivity puzzle,” drawing parallels with historical technological revolutions such as the Industrial Revolution.
AI as a General Purpose Technology
Rawert defined GPTs as technologies that are pervasive across all economic sectors, improvable over long periods, and possess “complementarities”—meaning they raise returns in applications, which then increases the value of the original technology. Historical examples include the steam engine, railways, electricity, and computers. AI, he argued, has a strong claim to be both a pervasive GPT and a technology that enhances the very process of invention itself, similar to the printing press or the microscope.
Historical Parallels and the Productivity Paradox
- The Steam Engine: Invented in 1712, but its significant impact on productivity was not observed until the 1830s, over a century later. This delay was attributed to the need for radical improvements in the core technology and the development of crucial complements like new machine tools, railways, steamships, and a newly skilled workforce.
- Electrification: Thomas Edison’s first power plant in 1882 preceded widespread factory electrification by 30-40 years. The bottleneck wasn’t the technology’s improvement but the fundamental redesign of factories. Moving from a central steam engine (group drive system) to individual electric motors (unit drive system) for each machine revolutionized factory layout, allowing for workflow optimization rather than power transmission constraints.
- Computers and Information Technology: Robert Solo’s 1987 observation that “you can see the computer age everywhere but in the productivity statistics” illustrated the initial disconnect. Productivity gains only became evident in the mid-to-late 1990s, requiring extensive organizational changes, new skills, enterprise software, database technologies, and supply chain redesigns. This adoption pattern is often described as a J-Curve relationship, where initial investment and adaptation cause a temporary slowdown before accelerating productivity.
Investment Booms and AI’s Future Potential
Rawert compared current AI infrastructure investment to historical capital expenditure booms:
- Railway Mania (1840s Britain): Investment reached 5-7% of GDP.
- Dot-com Bubble (late 1990s Fiber): Investment reached 1-1.5% of GDP.
- AI Infrastructure: Already at approximately 1.8% of US GDP this year, projected to reach 3% by 2028. This places AI investment on par with some of the largest historical investment waves. Rawert noted that even if these booms resemble bubbles, the resulting infrastructure often outlives the speculative phase and becomes productive later (e.g.,