The most critical knowledge in a multibillion-dollar financial institution is often locked inside the head of a single person. That person, likely in their late sixties, understands the arcane logic of a COBOL program written before they were born. When they retire, the risk of a catastrophic system failure rises. Hypercubic, a seed-stage startup from Y Combinator’s Fall 2025 batch, is betting that AI agents can capture that expertise before it walks out the door [Y Combinator, 2025].
A digital twin for the retiring expert
Hypercubic’s wedge is a product called HyperTwin, which the company describes as an agentic AI that creates a digital twin of subject matter experts to preserve their workflows [Preqin, Nov 2025]. The goal is not just to document code, but to encode the tacit knowledge, tribal history, and troubleshooting instincts that keep legacy mainframes running. A companion tool, HyperDocs, automates documentation directly from COBOL codebases. The combined offering is sold as a B2B SaaS subscription to enterprises in sectors like financial services, aerospace, and government, where an estimated 70% of Fortune 500 companies still rely on mainframe systems [Y Combinator, 2025].
The team and its early backing
The company is led by co-founders Sai Gurrapu and Aayush Naik. Gurrapu, who serves as CEO and CTO, is a former Apple engineer [Preqin, Nov 2025]. Naik is listed as CTO and co-founder, with a background as a seasoned software leader focused on robotics [GetProg.ai, 2026]. Their venture is backed by Y Combinator, where Garry Tan is listed as the primary partner, and the Pioneer Fund, which participated in an undisclosed seed round in November 2025 [Preqin, Nov 2025] [Y Combinator, 2025].
| Role | Name | Background |
|---|---|---|
| CEO & CTO | Sai Gurrapu | Ex-Apple engineer [Preqin, Nov 2025] |
| CTO & Co-Founder | Aayush Naik | Seasoned software leader, robotics engineer [GetProg.ai, 2026] |
Where the proof must materialize
The ambition is clear, but the path to validation is steep. The core risk for Hypercubic is moving from a compelling concept to a proven, deployed solution inside a risk-averse enterprise. No named customers or live deployments have been disclosed in the public record. Success will likely be measured in quarters, not months. The next twelve months will be critical for the company to demonstrate three things:
- Technical validation. A published benchmark or detailed case study showing HyperTwin accurately capturing and replicating a complex expert workflow.
- Initial lighthouse customer. A named, credible enterprise in finance or aerospace willing to discuss their pilot deployment.
- Regulatory awareness. Demonstrating a mature approach to security and auditability will be essential.
The patient population: legacy system stewards
The disease state Hypercubic is treating is institutional amnesia, a systemic risk born from decades of technological debt and an aging workforce. The patient population is the global cohort of enterprises, banks, insurers, airlines, and government agencies whose core transaction processing still runs on COBOL mainframes installed in the 1970s and 80s. Hypercubic’s bet is that an AI agent, trained as a digital twin, can become a more reliable and permanent steward of that logic than a retirement-bound human or a static document.
Sources
- [Y Combinator, 2025] Hypercubic: AI to maintain and modernize COBOL | https://www.ycombinator.com/companies/hypercubic
- [Preqin, Nov 2025] Hypercubic Inc. Asset Profile | https://www.preqin.com/data/profile/asset/hypercubic-inc-/787961
- [GetProg.ai, 2026] Aayush Naik - CTO Co-Founder at Hypercubic | https://www.getprog.ai/profile/9693267