The analysis
Weekly posts on Physical AI safety. Data-backed. Source-cited.
A specialist publication for the people building, regulating, and capitalising the next decade of Physical AI. Every post starts from a primary source, cites the source, and argues for one thing at a time. No speculation where data exists. No data where speculation would be more honest.
Featured
Venture funding into robotics startups in 2025 is tracked to the dollar. Venture funding into robot safety infrastructure is not tracked at all. I searched every major VC database. The category does not exist. Not because investment is low. Because the category hasn't been defined yet.
Data
Venture funding into robotics startups in 2025 is tracked to the dollar. Venture funding into robot safety infrastructure is not tracked at all. I searched every major VC database. The category does not exist. Not because investment is low. Because the category hasn't been defined yet.
Standards
Software watchdogs are enough to keep robots safe. That's the myth. The reality is that software fails with the system it's watching. Aviation, nuclear, and rail learned this decades ago. Physical AI is learning it now.
Standards
Software safety has a ceiling. Understand where it sits, and you understand why Physical AI in 2030 will look very different from Physical AI in 2020 — not because models got smarter, but because the industry finally accepted that certain guarantees cannot be made in software alone.
Data
From factory floors to living rooms to public roads. Six publicly-reported robot incidents from 2021 to 2026. Different companies, different sectors, different countries. One common thread — no hardware-enforced safety layer between the AI and the physical world.
Industry
The underrated skill in deep-tech is slow reading. Not papers. Not newsletters. The actual standards. This is a short field note on the Tuesday-morning habit that shapes everything else I publish here.
Industry
AI models keep getting smarter. Robot hardware keeps getting cheaper. Robot safety infrastructure has barely changed since the 1990s. That gap has a deadline — 20 January 2027.
We use cookies
This site uses essential cookies to function and, with your consent, analytics cookies (Google Analytics) to understand how the site is used. Learn more.