750 kW. Then 400. Then 200. The Software Decides.

Formula 1’s first season under its 2026 power-unit regulations has produced a precise, uncomfortable demonstration of what happens when you hand energy-management authority to an algorithm and tell the driver — and the audience — to live with it. As Ars Technica reported this week, the new 1.6 L turbo V6 plus 350 kW MGU hybrid architecture means an F1 car can deliver anywhere between 200 kW and 750 kW to the rear wheels at any given moment, depending entirely on what the onboard software decides about battery state. The driver’s right foot is no longer the final arbiter. The car’s brain is.

McLaren’s Lando Norris called it “soul destroying.” The FIA convened an emergency technical working group this month — with representatives from all ten constructors — after Suzuka’s Japan Grand Prix exposed two hard consequences: fast corners neutered by mandatory lift-and-coast, and speed differentials of up to 70 km/h between cars at different charge states. That second point produced a concrete crash: Franco Colapinto’s depleted Alpine slowed sharply; Oliver Bearman’s Haas had no warning and hit it hard. The FIA acknowledged the meeting and “a commitment to making tweaks to energy management regulations,” per its official statement cited in the Ars Technica piece.

←TODAY: F1’s 2026 hybrid ECU decides power output mid-corner; the FIA has no real-time override and no coordinated inter-car protocol.
→3012: Every building system in Zurich-3012 runs on a negotiated energy mesh — HVAC, lifts, façades — where an invisible scheduler arbitrates load. The question is who wrote the arbitration rules, and who can audit them.
Fulcrum: Autonomous energy control only works when the governance layer — permissions, audit trail, override hierarchy — is designed before the algorithm ships, not patched in after the first crash.

The Control Mechanism Nobody Designed Deliberately

The structural problem here is not horsepower arithmetic. It is that the FIA’s 2026 technical regulations specified what the hybrid system must do — store 4 MJ, deploy up to 350 kW, cap super-clipping at 200 kW — without specifying who controls when. Each manufacturer implemented their own energy-management software. The result is that the permission layer (who decides power delivery) defaulted to proprietary ECU logic, with no inter-car coordination protocol and no driver-readable state signal. A core FIA sporting principle — “the driver alone controls the car” — is now structurally violated by the regulations the FIA itself wrote.

The Ars Technica report notes the new formula was designed in a different market moment: before Volkswagen Group, Toyota, and others began retreating from EV commitments. Audi, Cadillac, and Honda signed on to the 2026 formula partly because the electrification ratio looked commercially relevant. The technical choices that made the system legible to road-car marketing — a Prius-scale 1.1 kWh battery, PHEV-style engine charging — are precisely the choices that made the racing ungovernable. Commercial incentives shaped the architecture; governance was left for later.

On an Engineer’s Desk This Week

If you work in BIM coordination, building automation, or parametric façade control, you are already managing analogous systems. A BMS (Building Management System) arbitrating between PV yield, battery storage, heat pump demand, and EV charging does exactly what the F1 ECU does: it allocates limited energy across competing loads in real time, using rules its integrator wrote, with minimal transparency to the occupant or operator. The analogy breaks down in one important direction — nobody crashes at 300 km/h — but the governance failure mode is identical: the algorithm makes binding decisions under conditions the specifier never stress-tested, and the audit trail is thin.

Swiss SIA norm 386 (building automation) and the EN 15232 energy efficiency classification for HVAC controls both define what automated systems must achieve. Neither mandates an override hierarchy, a logged decision trace, or a human-readable state display for the operator. That gap is where the F1 problem lives in architecture, too.

Atelier: In PAZ’s HIM (Human-in-the-Middle) framework, every autonomous building subsystem requires three explicit design decisions before it ships: who holds override authority at each priority tier, what triggers a mandatory human notification, and how the system’s current state is legible to a non-specialist. The FIA is learning this the hard way at 320 km/h; design it before you close the BEP.

The Trade-Off, Stated Plainly

Giving an algorithm real-time control over energy allocation is efficient — it reacts faster than any human, and it optimizes across more variables simultaneously. The cost is that its decisions are opaque, its failure modes are emergent, and patching the governance layer after deployment is structurally harder than any other kind of retrofit. The FIA is now discovering that “constructive dialogue” with ten competing manufacturers about changing software behavior mid-season is slower than a pitstop and more expensive than a new front wing.

Read the FIA’s post-meeting statement, then pull your next BMS specification and locate the section on override hierarchy and decision logging. If it is not there, that is the section you need to write — before the system goes live, not after the first incident report.

Source: Ars Technica