Three days ago, Alexandre LeBrun did something almost unheard of in AI: he told the truth.
In a TechCrunch interview on July 16, 2026, the CEO of AMI Labs—Yann LeCun’s world model startup—flatly refused to describe his work as “AGI” or “superintelligence.” His observation? The industry has shifted from one vague label to another without agreement on what either means. He described them as unhelpful hype rather than meaningful definitions.
Here’s the twist: LeBrun’s system reportedly shows capabilities that “rival or exceed many so-called AGI models.” He’s not hiding a weak product. He’s making a calculated stand against the hype machine.
AMI Labs is building world models—systems designed to learn from physical reality and predict the next state of the world using internal representations that incorporate causal understanding. These models are complementary to LLMs, not replacements. The company is pre-product but is already courting industrial partners in Asia, including robotics, manufacturing, and electronics firms. SBVA (SoftBank Ventures Asia) is acting as a strategic partner to connect AMI with Asia’s industrial ecosystem, facilitating proof-of-concept initiatives with major corporations. Real factories. Real homes. No magical thinking.
Who wins if LeBrun is right? Anyone tired of the AI hype cycle. By refusing to inflate claims, AMI builds trust in risk-sensitive sectors like manufacturing and healthcare. Yann LeCun, a longtime AGI timeline skeptic, gets a champion who actually walks the walk.
Who loses? Every company that slaps “AGI” on their marketing. OpenAI, Anthropic, DeepMind—they’ve all been key proponents of the narrative that superintelligence is imminent. If world models deliver reliable physical-world performance without the buzzwords, those brands look fraudulent.
The danger for LeBrun is that substance gets ignored in a GPU-fueled gold rush. Most pre-revenue startups would chase any hype they could get. He’s betting that long-term trust beats short-term headlines.
But here’s what makes this story bigger than one startup: LeBrun isn’t just fighting competitors—he’s fighting the very vocabulary the industry built. Regulators could start cracking down on unsubstantiated AGI claims. Investors might shift from narrative-driven bets to data-driven ones.
LeBrun’s quiet rebellion might just redefine how we talk about AI. Or it could leave his company invisible while louder players cash in. The answer will come when his world models meet the messy, unpredictable chaos of the physical world. And that’s a test no buzzword can pass.