AI forecasts compress four different clocks into one, Rodney Brooks argues

Why it matters

AI valuations often price economy-level outcomes while products remain in research or early deployment. Brooks's framework separates technical progress from adoption and labor-market claims.

Rodney Brooks, the roboticist who co-founded iRobot and built two more robotics companies, argued in an August 1st essay that technology forecasts routinely collapse four separate clocks into a single, misleading timeline.

Those clocks govern research, hype, deployment and economy-wide change. Brooks's framework explains how a genuine laboratory advance can become a marketing category within months, require decades to reach broad commercial use and take longer still to alter employment or productivity across an economy.

The argument carries the weight of Brooks's path through research and commercialization. He directed the MIT Artificial Intelligence Laboratory and later MIT CSAIL, co-founded iRobot, and served as its CTO from 1990 to 2008. He then founded Rethink Robotics, which built the Baxter and Sawyer collaborative robots and operated from 2008 to 2018. Brooks is now co-founder and CTO of Robust.AI, a warehouse robotics developer.

That record includes products that reached customers, technologies that took years to mature and a robotics venture that closed after a decade. Brooks is writing from the long middle between a convincing demonstration and a durable business.

Four clocks behind one forecast

Brooks puts new research ideas on the first clock. He estimates that they commonly need 10 to 20 years to produce solid laboratory demonstrations, with harder problems taking much longer. Once an approach works, improvements can arrive every six months, creating the appearance that the entire field is moving at that speed.

Neural networks are his central example. Brooks traces the line from the computational neuron models published by Warren McCulloch and Walter Pitts in 1943, through threshold neuron work in 1960, convolutional networks and backpropagation, to the 2012 deep-learning advance associated with Geoffrey Hinton and his collaborators.

The dates make Brooks's point: a concentrated run of visible progress rested on research accumulated over generations. The latest model release reveals little about how quickly a regulated industry, factory network or labor market can absorb the underlying technology.

Hype moves on Brooks's second clock. He points to "AI agents" going from limited public visibility in mid-2025 to a dominant theme in San Francisco bus advertising by August 2026. That observation is anecdotal, rather than a measure of agent adoption. It captures the speed at which a technical label can become a fundraising and sales category.

Brooks groups agents with earlier waves around blockchain, the metaverse, IBM Watson, expert systems and nanotechnology. Some of the underlying work survived and found uses after attention moved elsewhere. The marketing category had a much shorter life than the engineering.

Deployment has its own bottlenecks

The third clock starts when a product works and ends when it is widely used. Brooks says even software often takes 20 years or longer to reach broad deployment because customers must change workflows, integrate systems and accept operational risk. Hardware adds manufacturing, maintenance, support and physical supply chains.

His Linux example needs a precise reading. Microsoft added preview support for Linux images on Windows Azure in June 2012, matching the year Brooks identifies as Microsoft's adoption point. Microsoft's formal shift continued afterward, including joining the Linux Foundation as a Platinum member on November 16th, 2016. Adoption was a sequence of product, support and institutional decisions, rather than one event.

Waymo provides the sharper hardware case. Brooks first saw a self-driving vehicle operating on a German freeway in 1987 and rode in a Google X predecessor to Waymo in 2012. He took a commercial Waymo ride in San Francisco before publishing the new essay, more than three decades after that first demonstration.

The ride also exposed the operational layer that demos leave out. Brooks wrote that the vehicle failed to take him to his house and dropped him elsewhere after he spent more than 20 minutes with customer support.

Waymo has moved well beyond a laboratory project. On February 2nd, Waymo said it had raised $16 billion at a $126 billion post-money valuation. Waymo also reported more than 400,000 weekly rides across six major US metropolitan areas, 15 million rides during 2025 and more than 20 million lifetime rides. Those are company-supplied figures, and they describe a service that remains small beside the total volume of US passenger transportation.

Brooks calls Waymo the US leader while using it to reject the idea that technical leadership equals economy-wide deployment. Waymo's scale, capital requirements and customer-support problems show how much work begins after autonomous driving becomes technically credible.

The capital-market consequence

Brooks's fourth clock covers technologies that reshape an economy. Forecasts that jump from a research result to mass labor replacement within a few years skip the factories, integrations, service organizations and purchasing decisions that make economic change real.

That compression affects where capital goes. In a September 26th, 2025 essay on humanoid dexterity, Brooks wrote that venture capitalists had questioned why Robust.AI should scale wheeled warehouse robots when two-legged humanoids were expected to replace most human work within two years. His four-clock framework is an answer to that funding logic.

Robust.AI's commercial path is deliberately narrower. A 2020 financing announcement said Robust.AI had raised a $15 million Series A led by JAZZ Venture Partners, bringing its total funding at the time to $22.5 million. In December 2025, DHL and Robust.AI announced a five-year alliance that began with 15 Carter robots in Mexico. The partners claimed Carter had increased picking productivity by more than 60% in North American deployments and said DHL planned to deploy hundreds across the Americas.

Those figures come from the partners, and the rollout remains modest beside the general-purpose labor replacement promised by humanoid robotics pitches. It is also a recognizable deployment plan: prove a bounded use case, integrate it into warehouse systems and expand facility by facility.

Brooks is not arguing for less technology investment. In an April 17th, 2025 essay, he connected long deployment cycles to the need for sustained funding. His warning is about attaching the wrong promised outcome to each stage of progress.

A laboratory result can justify more research. A working product can justify a deployment budget. Reliable use across customers can support a scaling case. Claims about mass unemployment or economy-wide productivity require evidence accumulated over a much longer clock.