Toffer
Winslow

Toffer Winslow's product writes a Java unit test about once every two seconds. That is the number he leads with, and it is the number that gets him into rooms, but it is not really the point. The point is who is in those rooms. Diffblue's customers include four of the ten largest banks in the United States, plus Citi, ING, Cisco, AstraZeneca, and the Bank of New York Mellon - which is to say, a roster of institutions whose entire operating premise is that unproven software is a liability, and whose lawyers get nervous when anyone says the words "generative AI."

So Winslow's daily work is a kind of translation. Most of the noise in AI-for-software right now comes from a handful of enormous companies pointing very large language models at every problem at once. Diffblue does the opposite. Its technology, spun out of Oxford University's computer science department, uses reinforcement learning to produce tests that are deterministic - the same input yields the same output, every time - and it runs on-premises, so no customer code is shipped off to train someone else's model. For a bank, that combination is the whole sale. Winslow's job is to keep saying it until the risk committee believes him.

He took the CEO seat because the pieces were already in place. "Diffblue's deterministic AI technology, committed employees, blue chip customers, and visionary investors are a rare and enviable foundation on which a great company can be built," he said on arriving. He is a builder who prefers to join after the foundation is poured - which, over three decades, has become something of a specialty.

Diffblue did not begin as a company. It began as a research project inside the computer science department of Oxford University, a place more accustomed to publishing papers than closing deals with Citi. That heritage is why the technology is what it is - reinforcement learning rather than a large language model, determinism rather than probability - and it is also the reason a company like this needs someone like Winslow at the top. The hard part of turning academic research into a business is rarely the research. It is the enterprise sales floor, the security questionnaire, the procurement cycle at a bank that measures risk in decades.

Winslow is a Harvard MBA with an undergraduate degree from the University of British Columbia, based in Boston, running a company headquartered an ocean away in Oxford. His management style, according to people who have worked for him, bends to the person: coaching when coaching helps, and rolling up his sleeves to execute when execution is what the moment needs. It is an unglamorous description, which suits an unglamorous product. He is not selling a revolution. He is selling the tests that prove the revolution works.

In 2025 Diffblue extended its flagship Cover platform with three new capabilities - Test Asset Insights, LLM-Augmented Intelligence, and Guided Coverage Improvement - a quiet signal that even a company built on the anti-LLM pitch is willing to borrow from LLMs where it helps, so long as the determinism and the on-premises guarantees hold. That is the Winslow move in miniature: less ideology, more focus.