Cisco's Jeetu Patel Gave a 27-Year-Old Unlimited AI Tokens to Rewrite Legacy Code
A.I. / news
Cisco's Jeetu Patel Gave a 27-Year-Old Unlimited AI Tokens to Rewrite Legacy Code
Fortune's account gives a 24-month goal and a four-week proof period, but no token budget, codebase size or result.
Cisco president and chief product officer Jeetu Patel said he has given a 27-year-old engineer unlimited AI tokens to refactor the company's legacy code, with a target of refactoring 98 percent of it within 24 months, according to a Fortune report published Oct. 30. Tokens are the units AI services bill by, so "unlimited" removes the usual per-person spending cap.
Patel said it at a small roundtable dinner with Fortune. "I've given him unlimited tokens because he's going to refactor code," he said. "So I'm like, you can go, you can go nuts."
Who the engineer is
Fortune names him as Padmanabhan Krishnamurthy, known as Paddy, who cofounded the real-time translation startup EzDubs in 2023 and joined Cisco in 2025. Cisco's own announcement of the EzDubs deal, written by Snorre Kjesbu, senior vice president and general manager of Collaboration, was posted Nov. 14, 2025 and updated Nov. 25, 2025 to say the acquisition was complete. That post does not mention Krishnamurthy by name, so his role comes from Fortune alone.
Patel said he found him "five levels" below himself in the organisation, "buried away in the organization doing something unstrategic". Krishnamurthy argued that a two-person team could rewrite the legacy code, Patel said, and gave himself four weeks to prove it.
What the report leaves out
| Detail | In Fortune's report |
|---|---|
| Token budget | "Unlimited", no figure |
| Size of the legacy codebase | Not stated |
| Result of the four-week trial | Not stated |
| Target | 98 percent of legacy code in 24 months |
| Team size proposed | Two people |
Fortune gives no codebase size in lines or repositories, no model or vendor used for the rewrite, and no cost. Without a baseline, "98 percent" cannot be tested, because a percentage of an unmeasured codebase is not a measurement. Cisco did not publish any of this alongside the interview.
Fortune does not say when the four-week proof period began or ended, who else is on the effort, or what the first four weeks produced. It also does not define what Cisco counts as legacy code, which is the denominator in the 98 percent target. The Cisco blog post on EzDubs, the only other document read for this piece, says nothing on the subject.
The hiring argument
The rest of the piece is about early-career hiring. Patel said roughly one-third of his team is experienced operators, one-third outside-in systems thinkers and one-third AI-native founders and engineers. He asked, "If you don't hire the early-in-career [workers]... where do fresh ideas get injected?" Fortune cites research that about 40 percent of employers plan to hire fewer graduates because of AI.
The scepticism around AI productivity claims is already documented on this site. The SRE deskilling argument shows how far one study can be stretched, and a vendor-run browser benchmark shows why a self-reported figure needs a named harness. Patel's 98 percent target has neither.
Fortune describes Cisco as a $429 billion company. Cisco has not said when it will report progress on the legacy rewrite.
Sources
More in A.I.
- 01Nanbeige4.3-33B Ships Under Apache-2.0 With a 524,288-Token Window and Self-Reported WinsNanbeige Team's 33B model is built as a teacher for smaller agents. Its own table has Qwen3.8-27B ahead on SWE-Bench Pro and GDPval.
- 02GPT-5.6 Sol Launched a Factorio Rocket on Peaceful Mode, With Hauling Done by HandBlogger Ryan Madden's single run took 13.5 in-game hours. The agent carried supplies between areas because it never linked some production lines.
- 03Bootgpt Fits a Character-Level Language Model Into a 512-Byte Boot Sector Plus 23 Sectors of WeightsThe MIT-licensed Show HN project writes fake Linux boot logs on a 386-class machine with no operating system. A rival project runs a Llama 2 model in 1,189 bytes.
- 04Mistral Large 4 Weights Are on Hugging Face, and the Card Disagrees With the Launch PostThe 1-trillion-parameter model is ungated under a custom licence, but its active-parameter count and two benchmark scores differ from Mistral's October 6 announcement.