Ndoc, a three-person startup founded by Rohit Patel, a former director at Meta Superintelligence Labs, launched an AI-native office suite on Oct. 27 that folds documents, slides and spreadsheets into one inspectable file, according to TechCrunch. The company is pitching the product as a replacement for Microsoft Office and Google Workspace, and it arrives only weeks after OpenAI introduced Space, a shared workspace inside ChatGPT.
One file instead of three
Each ndoc contains three tabs: Stage, Narrative and Model. According to Patel's Medium post, published Oct. 28, the narrative runs on Markdown, the stage on React (a popular web interface framework), the data on Parquet and SQL, and the analysis on Python.
The idea is to collapse the familiar trio of document, deck and spreadsheet. "ndoc starts from a different premise: there should be two things, not three. Just you and the ndoc, working together on the idea itself," Patel wrote.
He was blunt about the incumbents. "The office suite we all use was built for a world without AI, and today's AI tools are bolted on top of it," he wrote.
Inspectable by design
Users can click any number to see the run or table behind it, TechCrunch reported. They can also read the underlying Python and ask the AI what a script does or which line holds a given formula.
"There's nothing hidden behind the curtain here," Patel told TechCrunch on Oct. 27, referring to the visible code and figures.
Ndoc also keeps memory at the user, document, team and company levels, according to TechCrunch.
An open export format
Every ndoc can be exported as a single ".ndoc.html" file. In his Medium post, Patel described it as "a single self-contained file that runs interactively offline in any browser."
His personal site states that the format specification is open under the Apache 2.0 license. TechCrunch, for its part, reported only that the format is intended to be open-sourced, so the exact status of the release is worth confirming.
Models, access and pricing
Ndoc was built and fine-tuned on a mix of open-source models, TechCrunch reported. Patel claims this lowers running costs compared with tools built on frontier models, meaning the largest and most expensive systems from the leading AI labs. That claim has not been independently verified.
An enterprise pilot is under way, and an alpha is open through a waitlist, both since Oct. 27. Sign-up is free, and customers pay for usage, billed by tokens, the small units of text that AI models process, TechCrunch reported. As of Oct. 29, no public price list existed.
A beta is due "in the coming weeks," according to TechCrunch. Moneycontrol reported on Oct. 28 that a full commercial launch is planned for 2027.
No customers have been named, and the company has published no logos or case studies.
The founder and the money
Patel spent about eight years at Meta building AI and most recently served as a director at Meta Superintelligence Labs, TechCrunch reported. His Forbes Councils profile lists a PhD in game theory from Northwestern's Kellogg School of Management and bachelor's and master's degrees in statistics from the Indian Statistical Institute. The profile also lists him as an advisor to BayPine and places him in Miami.
Ndoc has disclosed no outside funding. Patel told TechCrunch he plans to raise money after proving that users love the product. The site pushinweight.ai claimed the company had landed a $215 million round, but the claim is uncorroborated, and no credible primary source supports that figure.
The other two members of the team have not been identified publicly.
A crowded field
The incumbents are not standing still. Microsoft is embedding Copilot across Word, Excel and PowerPoint, according to TechRadar. Google, per its Workspace blog, is rolling out agentic Gemini features across its apps, with Skills features arriving between Oct. 5 and mid-November.
OpenAI launched Space on Sept. 29 and 30. TechCrunch said it feels like "a whole lot like ChatGPT's own office suite."
Patel is betting that a file anyone can open, audit and carry offline will matter more than another assistant attached to an old format. The beta, due within weeks, should offer the first broad test of that wager.
