Over 44 days of dictating, I produced more than 203,000 words and almost no finished work. What I did produce was a new vocabulary: 'speechos', outpourings, and a felt sense of what the next interface with AI might actually be.
The French Government has recently acquired all of the shares of Bull in the hopes of rebuilding a sovereign AI stack from the ground up. Bull CEO Emmanuel Le Roux explains why this time might be different from the last time this occurred and how he hopes to inspire European leadership to shape the future of AI
AvePoint’s Dana Louise Simberkoff argues that shadow AI could be viewed as a cultural stress test rather than a technology failure.
AI rockstar Andrej Karpathy has launched a self-learning agent that shines a light on the boundary problem. Unremarkable at first read, but a deeper analysis shines a light on the biggest challenges and even bigger opportunities in agentic AI.
The ARC-AGI-3 challenge launches today as the first interactive reasoning benchmark which stumps current frontier LLMs. Karl Friston, one of the most cited neuroscientists alive and Chief Scientist at VERSES AI, argues that his active inference framework can compete where LLMs fail because it does something they structurally cannot: encode uncertainty and be curious. The questions Friston is asking may outlast the answers Big AI is selling.
Salesforce recently completed an ambitious plan to migrate the Informatica Help system to the Salesforce AgentForce experience in 24 days ahead of its 30-day ambition. Salesforce VP of Data Science Irina Malkova walks through the essential role that human curiosity and experimentation played in making this ambitious project go smoothly.
Concr CEO Irina Babina and CTO Matthew Griffiths unpack how Bayesian foundation models can excel at uncertainty management to build trust in medical AI digital twins. Their insight has important implications for user experience design that augments rather than replaces human context and expertise.
Qoob CEO AJ Javan argues that building data centers for AI inference and fine-tuning might be a more practical opportunity than building bigger. These could be smaller, more distributed, and more sustainable.
AI is often framed as a technology or meme that inflates aspiration into valuation. Both can be useful, but miss something important unfolding in the world right now. Something that could be better informed by an individual and collective felt sense of various lenses into AI as a trance.
AI chips, models, and data are generally seen as the limiting factors for AI growth. But Caspar Herzberg, CEO at Aveva, argues that electrical grids are emerging as the binding constraint to scaling AI.
Druid AI CEO Joe Kim weighs in on why the future of enterprise agents needs a governance-first mindset that supports composability and extensibility.
OpenAI has signed on Peter Steinberger, the pioneer of the viral OpenClaw open source personal agentic development tool. Qwibit Co-founder Gavriel Cohen, who developed NanoClaw, a more code-efficient and secure alternative, weighs in on implications for the enterprise.
Google’s DORA first report on AI coding maturity provides a valuable starting point and a bevy of helpful metrics. It also points to some interesting paradoxes that require additional context and might benefit from an expanded HR role.
