ניתוחים

פרסום מקצועי לאנשים שבונים, מסדירים, ומממנים את העשור הבא של Physical AI. כל פוסט מתחיל ממקור ראשוני, מצטט את המקור, וטוען טיעון אחד בכל פעם. ללא ספקולציות היכן שיש נתונים. ללא נתונים היכן שספקולציה תהיה כנה יותר.

תקנים

How a safety standard gets written — six stages, years of work, one vote

Most people use safety standards. Almost nobody knows how they're made. Six stages from proposal to publication. IEC 61508 took 12 years between editions. The next revision is being shaped now.

23 ביולי 2026קרא →

תקנים

The rules are being written right now. Are you in the room?

IEC TC 65. ISO TC 299. IEEE P7000. Three committees drafting the rules that will govern every robot sold in Europe. The seats are not all taken. If you're not in the room, you're subject to the outcome.

22 ביולי 2026קרא →

תקנים

Standards aren't written by governments. They're written by committees you can join.

IEC TC 65. ISO TC 299. IEEE P7000. The rules that will govern every robot sold in Europe are being drafted now. Almost no Physical AI founders show up to write them. Here's who does, how it works, and why it matters.

21 ביולי 2026קרא →

כלכלה

The underwriting gap — where the insurance models haven't arrived yet

A 2×2 map of insurance coverage vs. model maturity. Automotive is mature. Cyber is fast-growing. Traditional machinery has coverage but outdated models. Physical AI sits in the worst quadrant: thin coverage, immature models. The first robot-specific policies arrived in late 2025 — and the gap is still the story.

16 ביולי 2026קרא →

נתונים

1979. $15 million. 2025.

The first human killed by a robot was Robert Williams, age 25, at Ford's Flat Rock plant. His family was awarded $15 million. Forty-six years later, in October 2025, the first insurance written specifically for robots finally appeared — in China, not the US or the EU.

15 ביולי 2026קרא →

הנדסה

Fleet safety — why 1 robot is not 1,000 robots

Single-robot safety is predictable and testable. Fleet safety introduces failure modes that don't exist at the individual level: shared updates that fail fleet-wide, emergent congestion patterns, and coordinated behaviors nobody programmed.

14 ביולי 2026קרא →

אקוסיסטם

Three superpowers, one missing layer — what mapping Physical AI revealed

I spent a month mapping the only three ecosystems on Earth that build Physical AI at scale — Israel, Europe, Japan. Three superpowers, three philosophies, three completely different ways to build a robot. One hole runs through all of them: no independent safety layer. As of 22 June, that hole has a shape — because NVIDIA just filled the single-vendor version of it.

9 ביולי 2026קרא →

רגולציה

Directive vs Regulation — why the legal instrument change matters more than you think

A directive lets each country interpret differently. A regulation is directly binding — same text, same day, all 27 member states. That shift from 2006/42/EC to 2023/1230 changes everything for robot OEMs selling into Europe.

8 ביולי 2026קרא →

מדיניות

Israel already has regulatory sandboxes. Here's what's missing.

Israel's AV sandbox is operational. Healthcare AI sandbox launched. General framework published. But no sandbox exists for Physical AI safety — where a robot OEM can test a novel safety architecture under regulatory supervision.

7 ביולי 2026קרא →