Cisco Gives Young Founder Unlimited AI Tokens to Rewrite 98% of Legacy Code

Cisco has assigned 27-year-old Padmanabhan Krishnamurthy, known as Paddy, to lead an effort to rewrite 98% of its legacy code within 24 months. Jeetu Patel, Cisco’s president and chief product officer, gave Krishnamurthy unlimited AI tokens and told him to “go nuts” as he tests whether a very small team can refactor software built over decades.

The assignment stands out because Krishnamurthy was not a longtime Cisco engineering executive. He joined the company after Cisco acquired EzDubs, a real-time translation startup he cofounded in 2023. Patel said he found the former founder buried five levels below him in a role he considered unstrategic, according to Fortune. Hacker News AI described the same move as a bet on a young engineer’s ability to challenge established engineering assumptions.

Krishnamurthy won Patel’s attention during a conversation at Cisco’s San Francisco office, where he argued that a two-person team could rewrite much of Cisco’s legacy code. Patel said even highly AI-focused members of his own team had not been thinking at that scale. The immediate test gives Krishnamurthy 4 weeks to demonstrate that the approach is viable before Cisco commits to the longer 24-month target.

The project combines a founder’s preference for speed with AI-assisted software modernization. Cisco’s scale makes the goal significant: replacing 98% of legacy code would touch systems that support products, internal processes, and long-standing dependencies. Neither article details the programming languages, review process, security controls, or exact definition of “rewrite,” leaving the engineering scope unclear.

Unlimited model access removes one constraint from the experiment, but it doesn't remove the hard parts of enterprise software work. A rewrite at Cisco’s scale still requires architecture decisions, testing, integration with production systems, security review, documentation, and a plan for code that cannot be safely replaced in one pass.

The proposal also places a young engineer in a role usually reserved for people with decades inside the organization. Patel’s argument is that early-career employees ask questions experienced staff no longer think to ask. The initiative therefore tests two ideas at once: whether AI can accelerate large-scale refactoring, and whether organizational habits are a bigger barrier than technical capability. Its first meaningful checkpoint is Krishnamurthy’s 4-week demonstration, followed by evidence that the rewritten code performs reliably in Cisco’s operating environment.

Patel frames the appointment as a response to a broader problem in technology hiring. AI is reducing demand for some entry-level tasks, while companies face pressure to hire fewer graduates. His answer is to give young employees work that shapes products and systems rather than limiting them to data entry or routine support.

Patel called Krishnamurthy an “energizer bunny” and said he wants to find more young employees who challenge established processes. That approach accepts friction between recent hires and veteran staff as part of the process. He said tension can be useful when it forces teams to question habits that have hardened over time. Cisco’s experiment gives that philosophy a concrete measure: a young former founder must turn an ambitious claim about AI and code into a working modernization program.

Cisco’s next decision rests on whether Krishnamurthy can show measurable progress without creating new reliability or security problems. The 24-month objective will require more than fast code generation. It will depend on test coverage, system ownership, rollback plans, human review, and coordination with engineers who understand the company’s oldest infrastructure.

The experiment also carries an organizational message. Cisco is treating AI fluency and unconventional thinking as reasons to elevate a young employee, even when that creates tension with veteran teams. If the first 4 weeks produce convincing results, the company will have a model for finding similar talent inside a large organization. If the effort stalls, the test will still expose the limits of applying startup speed and AI tooling to enterprise codebases.