Ndoc: Former Meta AI Director Launches Unified Document

Key Takeaways

1Ndoc, founded by former Meta Superintelligence Labs director Rohit Patel, is starting in pilot mode and opening an alpha with a waitlist, with a beta expected in the coming weeks and a launch planned for next year.

2The tool combines presentations, analyses, and data into a verifiable document, exportable in .ndoc.html and promised to be open source, with connectors to existing tools.

3In the face of offerings from OpenAI, Anthropic, Meta, SpaceXAI, and Google, Patel claims lower costs due to open models and a focus on rapid execution.

Why it matters : Ndoc aims to provide a unified and open alternative to dominant office suites, emphasizing transparency, verifiability, and rapid development.

Rohit Patel, former head of Meta Superintelligence Labs, is launching Ndoc, a workspace designed to produce and explain analyses and presentations using AI. The tool is entering a pilot phase with an alpha version available to a limited number of clients on a waitlist, a beta version expected in the coming weeks, and a full launch planned for next year. The startup is betting on an exportable format that promises to be open source, claiming lower costs thanks to open models.

Pilot Launched, Alpha on Waitlist, and Usage-Based Pricing

Ndoc is starting a corporate pilot on Tuesday while simultaneously opening an alpha version to a limited number of clients who have signed up on a waitlist. Registration is free, but billing will be based on usage and tokens. Rohit Patel aims for a beta version in the coming weeks, with a full launch announced for next year. The project is supported by a team of three and currently has no funding.

Competing with ChatGPT's Space and Connected Suites

OpenAI, Anthropic, Meta, SpaceXAI, and Google already offer AI tools for businesses, integrated with Google Workspace, Microsoft 365, and other professional applications. Last month, OpenAI unveiled Space, a shared workspace integrated with ChatGPT and a direct competitor to established environments. Rohit Patel admits that Space could pose a threat. The appeal of these integrations also lies in the difficulty of changing habits and replacing entire workflows.

A Unified Document with Visible Scripts and Multi-Level Memory

Ndoc combines three components within a single interface: Stage for presenting deliverables like slideshows, dashboards, simulations, or mini-apps; Narrative for developing analyses, numbered citations, and generated figures; Model for gathering datasets and Python scripts. Patel wants the produced content to be verifiable, with access to scripts and the ability to ask the AI to explain a code or indicate the line of a formula, asserting that nothing is hidden. Users can upload data, request analyses and narratives, and specify the audience, tone, and memory. The memory is described as extending from the individual to the document, to the team, and to the company. Collaborators can view deliverables even without being signed up on Ndoc.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Connectors, .ndoc.html Format, and Promise of Open Source

To minimize disruption to existing practices, Ndoc integrates data import connectors from tools already in use. Patel also intends to rethink the file: documents can be exported as a single .ndoc.html file that can be opened in any browser. He plans to publish this format as open source to allow other developers to create compatible applications, aiming to replace fragmented workflows across Excel, Word, PowerPoint, Tableau, Replit, and ChatGPT's copy-paste functionality.

Claimed Lower Costs and Testers' Preference for Unification

Patel explains that Ndoc is built and refined with a mix of open-source models. He argues that choosing to develop from scratch significantly reduces operating costs compared to solutions relying on cutting-edge AI models. According to his initial testers, however, the most attractive element would be the vision of work being done end-to-end in the same place.

Patel's Background, Execution Philosophy, and Reasons for Departure

After eight years working on AI at Meta, including reinforcement learning and assessing the reliability of results, Rohit Patel chose to leave to launch Ndoc. He links his decision to the desire to publish an open format, believing that achieving such a consensus in a large organization would require input from hundreds of people. He advocates for a rapid execution approach, preferring to implement an imperfect plan rather than waiting for the best, asserting that execution is low-cost and that others will advance otherwise. He claims to deliver only tools he uses himself, referencing a graveyard of ideas that did not meet this criterion, and wishes to convince clients before seeking funding. He criticizes the approach of only adding AI to spreadsheets to fill cells, using an analogy between electricity and tractors to emphasize his difference.

Context at Meta: Other Recent Departures Mentioned

Rohit Patel's departure comes during a period marked by other movements at Meta, including mentions of Joel Kaplan, Andrew Tulloch, and Emily Dalton Smith.