Superduperbert (𝔦, 𝔦)

Another day, another startup telling us what's coming. Flapping Airplanes is valued at 1.5B with investment from Google and Nvidia. From day one, they made it clear: the algorithms aren't getting shared. Their thesis: AI is wildly data-inefficient and that the path to AGI is finding fundamentally different learning algorithms. They're targeting >1,000x improvement. Eighteen months ago the founder was openly publishing through Stanford. Now he's sitting on $1.5B of stealth IP. This is what science looks like in 2026. And remember, this shift is new. Research like this used to get published openly. Now, companies raise a huge amount and need to capture value, so everything stays private. Sequoia (their lead investor) specifically mentioned their reason for investing: "we are 2-3 research breakthroughs away from an 'AGI' intelligence." Thoes "breakthroughs" are algorithms. If those algorithms stay private, whoever owns them owns the future of AI. We HAVE to ensure they stay open.

The only sectors I care about in equities for the coming decade: AI, robotics, compute and energy infrastructure. AI needs compute to function. Compute needs energy infrastructure to run. And robotics is AI moving into the physical world.

I think the next AI leg may rotate from semis into enterprise software. Trimmed $AMD $NVDA $INTC, rotated into $NOW, $CRM and speculative $PATH $NOW cleanest agent infrastructure play, owns the workflow layer where enterprise AI gets orchestrated $CRM levered to monetizing customer facing AI agents across sales, service and revenue workflows $PATH higher risk asymmetric bet on automation evolving into agentic execution Thesis: the next big AI spend may be enterprise deployment, not just model compute. - AI is moving from chat interfaces to agents that execute work - Enterprise adoption is shifting from experimentation to workflow automation - The bottleneck is no longer model capability, it’s integration and change management - Legacy systems make agent deployment hard, and that creates opportunity for software incumbents - Compute and token budgets may become a new enterprise resource constraint - The highest value use cases are moving from cost cutting to revenue generation - Demand should rise for interoperable, agent agnostic software layers - The future likely looks multi agent, not single platform - Ironically, AI is increasing operational complexity before it reduces it - Engineers increasingly move from coding software to operating agent systems The market may be shifting from asking: Who enables AI training? To asking: Who monetizes AI deployment? $MSFT and $ORCL also on my radar.

World Health Organization says there are now 5 confirmed cases of hantavirus.