Two Yale seniors just raised millions to build a social network that lives entirely inside iMessage. No app. No download. You text a phone number, tell it who you are and who you'd like to meet, and Series sends back “shares” — carousels of 10 image cards from other users you might want to know. Swipe, tap, start a conversation, never share your number. Already live on 750+ college campuses, with Day-30 retention of 82% — higher than early Facebook's benchmark. Founders Nathaneo Johnson and Sean Hargrow met as freshmen running a podcast for the Yale Entrepreneurial Society. They iterated for a full year before raising.

Recruiting still runs on a 1995 playbook: keyword-match resumes, blast irrelevant roles, take a 25% cut. Dex burned that down. Their AI talent agent has voice or text conversations with candidates — asking what motivates them, where they want to grow, what they're worth — then quietly matches them to a curated set of openings at companies that actually fit. The pitch caught on fast: 15,000+ engineers signed up, 50+ companies (ElevenLabs, Synthesia, Lovable, Granola) using it, and revenue ramped from zero to $1.8M ARR since they started charging in late 2025. Notion Capital, a16z Speedrun, and angels from OpenAI just put in $5.3M to scale.

Building a house in 2026 looks shockingly similar to building one in 1926. Slow, manual, error-prone. All3 is rewriting that. They're a London startup building autonomous legged robots (“Mantis”) that assemble buildings on-site, AI software that designs every component, and robotic factories that manufacture every part before it ships. One end-to-end system. Mantis is operational, customers signed, and the first commercial deployments hit active sites in Germany this year. The founders aren't first-timers — Rodion Shishkov and Slava Bocharov previously built Samokat into Russia's largest food delivery business and sold it for $1.5 billion in five years. Now they're aiming the same execution muscle at construction.

Two ex-Uber operators looked at the trillion-dollar U.S. home maintenance market and saw exactly what Uber found in taxis: opaque pricing, unreliable supply, no software layer. So Michael York and Michael Mizrahi built Casa — a monthly subscription that handles your home before it breaks. AI schedules proactive maintenance, dispatches vetted handymen, bills you a flat fee. The cap table reads like a Silicon Valley reunion: Forerunner Ventures led, with checks from Sheryl Sandberg's fund and Travis Kalanick personally. The pitch worked because both founders previously scaled CloudKitchens and Uber — they know exactly how to layer software onto messy real-world labor.

A year ago, Snabbit was completing 400 jobs a day. Today: more than 40,000 — a 100x leap in 12 months. They've crossed 1 million monthly jobs, operate across 140 micro-markets in India, and field a network of 15,000 service professionals. This is their fourth funding round in 15 months, bringing total capital to roughly $112M at a ~$350-400M valuation. The thesis is dead simple: India's $60B quick-home-services market wants the same speed the country now expects from food and grocery delivery — and Snabbit got there first.

AI's biggest problem isn't intelligence — it's electricity. Training and running models is melting the grid. Panthalassa's answer: don't fight for grid capacity. Build the data center in the ocean. The Oregon startup is engineering massive floating orbs that harvest wave power, run AI inference onsite, and use cold ocean water for cooling — sidestepping transmission losses entirely. Peter Thiel just led a $140M Series B at a near-$1B valuation, joined by John Doerr, Marc Benioff's TIME Ventures, Max Levchin's SciFi Ventures, and Susquehanna Sustainable Investments. Their first pilot — Ocean-3 nodes in the northern Pacific — deploys this year, with commercial deployments planned for 2027.