Cisco sets a 24-month goal to refactor 98% of its legacy code

Patel, Cisco’s president and chief product officer, assigned Paddy to work on refactoring the company’s legacy code.[1] Patel gave him unlimited tokens and encouraged him to use them freely.[1] Paddy had said he could rewrite the code with a two-person team.[1] Patel set himself four weeks to see whether Paddy could prove the approach.[1] That short evaluation period is distinct from Cisco’s broader stated objective: refactoring 98% of its legacy code within 24 months.[1]

Paddy is 27, and Patel has described him as an “energizer bunny.”[1] Paddy cofounded a real-time translation startup in 2023, then joined Cisco in 2025 after Cisco acquired EzDubs.[1] Before the assignment, he had been “buried” in the organization doing work described as unstrategic.[1] Patel and Paddy began discussing the project after Patel visited Cisco’s San Francisco office.[1] The evidence describes the proposed team size as two, but does not detail its roles or the codebase’s scope.[1]

Patel has said he looks for young hires who will challenge the system and the status quo.[1] He also said he deliberately seeks tension in teams to sharpen debate about ideas, while distinguishing productive disagreement from conflict between people.[1] Patel treats every intern as a reverse mentor.[1] He asked Baya to email him monthly about what she had learned about AI; she did not miss an update, later interned at Cisco, and redesigned something experienced staff had not solved.[1] Cisco brought Baya in to teach its people to think differently.[1]

Cisco has also described a separate AI-assisted development effort: twelve engineers shipped a rebuild in about five and a half months, producing output Cisco compared with that of about 20 engineers.[2] Cisco said that rebuild, PubHub, was required by a security mandate and involved a new technology stack.[2] Those details offer an example of AI-assisted development at Cisco, but the evidence does not say PubHub was part of Paddy’s assignment or that its results predict the outcome of the legacy-code effort.[1][2] Professionals assessing the new project should keep its four-week proof period distinct from the 24-month target.[1]