Vision Can Lift Thor's Hammer. Most Humans Can't. There's No Equivalent Test for the Humanoid Robots Real Companies Are Now Shipping Into Actual Homes.
1X Technologies is delivering its $20,000 humanoid home robot, Neo, to early adopters right now -- backed by OpenAI, Tiger Global, and Samsung at a roughly $10 billion valuation. Figure AI is running a BMW factory pilot at a $39 billion valuation, backed by OpenAI, Microsoft, NVIDIA, and Jeff Bezos. Tesla is targeting 10,000-plus Optimus units by the end of 2026. The humanoid-robot race isn't a hypothetical anymore, and the industry has a real answer for how capable these systems are. It has no equivalent answer for whether they're worth trusting -- no test anyone's built for that, the way a fictional hammer tested worthiness instead of strength and let the newly-built synthetic being lift it on his first try.
Read →
Fei-Fei Li's World Labs Raised $1.2 Billion From NVIDIA, AMD, and Autodesk to Build "Spatial Intelligence." Its New Model Turns One Photo Into a Minute of Walkable 3D World. NVIDIA Is Now Backing Nearly Every Layer of the Physical AI Stack at Once.
World Labs, the startup Stanford's Fei-Fei Li founded in 2024, unveiled Atlas on September 1, 2026 -- a single model that takes one or more photos and generates a full, camera-controllable 3D world from them, outputting up to a minute of navigable video at 1440p. The company has raised $1.2 billion, with NVIDIA, AMD, and Autodesk all named as backers -- two direct GPU rivals investing in the same downstream AI company at once. NVIDIA is also the company already holding two of the three layers in this outlet's own physical-AI stack reporting (Omniverse simulation, Jetson edge silicon). A stake in the leading standalone world-model company is a fourth foothold in the same buildout, not a new one.
Read →
"Technology Private Equity, in Its Current Form, Is Dead." The Lightspeed Partner Who Said That Works for the Firm That Wired Over a Billion Dollars Into Anthropic -- the Company Doing More Than Almost Any Other to Kill It.
Isaac Kim spent twenty years buying software companies, most recently at Elliott Investment Management, before Lightspeed brought him on in 2024 to build a technology buyout practice. In February 2026 he said the entire model is finished: AI-native companies now hit $100 million in revenue in 12 to 24 months, a climb that used to take software a decade, and the old private-equity playbook -- buy it, lever it, cut costs, hope the multiple holds -- assumed the product would stay relevant long enough for that financial engineering to work. AI broke that assumption. He said it again in September. His own firm is Anthropic's largest venture backer, having wired over a billion dollars into its Series E alone. The person diagnosing the death of an asset class works for the fund with the single largest financial stake in the company doing the killing.
Read →
Oluwadi