Cisco gives a 27-year-old engineer unlimited AI to rewrite its old code

Cisco's president and chief product officer, Jeetu Patel, has given a 27-year-old engineer, Padmanabhan "Paddy" Krishnamurthy, unlimited access to artificial intelligence (AI) tools and 24 months to rewrite 98% of the company's legacy code, according to an interview reported at the end of October 2026. Patel said he told the engineer to "go nuts". Krishnamurthy, who joined Cisco when it bought the real-time translation start-up EzDubs that he co-founded, has proposed doing the work with a team of two and has four weeks to show the approach can work before it is scaled up.

Legacy code is software written years or decades ago that companies still depend on. It is often large, poorly documented and written in older styles or languages, which makes it slow and risky to change. Keeping such code running takes a great deal of engineering time, and rewriting it by hand is a long, costly job.

Cisco's bet is that AI coding tools have become good enough to do most of that work, guided by a very small team. A token is the unit that AI language models use to measure text, so "unlimited tokens" means the company will not limit how much the engineer can ask the AI to read, write or rewrite. In effect, Cisco is trading engineer hours for computing spending.

If it works, the implications are large. It would suggest that a task once requiring a big department can be done by two people directing AI systems, which would change how technology companies plan staffing and budgets. That is also why the story has drawn attention alongside Cisco's job cuts earlier this year. If it fails, or produces code that is hard to check, it will show the limits of today's AI tools when applied to large, critical systems.

The choice of a young engineer is also notable. Patel described Krishnamurthy as someone who had been "buried away in the organization doing something unstrategic" before being given the assignment. Patel picked a former start-up founder rather than a senior team, a sign of how much weight he places on hands-on AI skills.

For India's large software services industry, much of whose work involves maintaining and modernising clients' older systems, experiments like this one are worth watching closely. If AI can modernise legacy code with tiny teams, the nature of such work, and the skills it needs, could change.

Cisco's legacy code project: what is known Item | Detail Engineer | Padmanabhan Krishnamurthy, 27 Target | Rewrite 98% of legacy code Time | 24 months, with a four-week proof stage Resources | Unlimited AI tokens, team of two AI tools used | Not disclosed

Short-term: The four-week trial will show whether the approach works on real Cisco systems before it is expanded.

Long-term: Success would make Cisco a test case for AI-led software modernisation, likely prompting other large companies to try the same, with effects on how many engineers they need for maintenance work.

Who is affected: Cisco's engineers and customers, technology companies with large legacy systems, IT services firms that do modernisation work, and young engineers whose AI skills are becoming more valued.

Questions and Answers