Modular Chunks
The MVP of Modular SDLC — the planning and validation layer that onboards AI coding agents. Built solo.
What we are
Modular Chunks is the MVP of Modular SDLC — a new way to build software with AI coding agents. Instead of prompting an agent and hoping, it thinks like a CTO, CPO, and CEO: it turns your idea into a plan (a graph of small, self-contained "chunks"), freezes each chunk into a contract, then onboards your agent to build and validate them one at a time — with you approving every step.
Chunks ship to whichever coding tool you already use — Cursor, Claude Code, Windsurf, Aider, Factory, GitHub Copilot, Bolt.new, Lovable, or v0. The CLI exposes an MCP server so MCP-compatible agents pull context directly.
The team
Modular Chunks is built by Nabeel El-dughailib — solo venture builder of Modular Chunks. A 2x founder, MBA mentor, and innovation advisor to Fortune companies.
Earlier, Nabeel was on the founding team of a pre-IPO startup now valued at ~$1.9B, where he built its sales, operations, and product functions and helped scale it to Series A.
The thesis
Our thesis is modular determinism: building software with an AI agent should be repeatable, not a roll of the dice. Break the work into small, verifiable chunks, freeze each into a contract, and verify by command — and the same input stops producing a different app every time.
Onboarding builds on that. We don't prompt agents; we onboard them. Prompting is like managing a new developer entirely over Slack DMs — no role, no plan, no context. Onboarding gives the agent a role, the plan, the spec, and a human gate at every step. It's how the build becomes deterministic — and why the output is reviewable, correct, and yours.
SaaS made you the operator — the human doing the work, click by click. Agents make you the supervisor: you set direction and approve each step, and the agent does the building. Modular Chunks is built for the supervisor, not the operator.
The research behind it
Modular Chunks is the product of 45 months of research — into ESG for AI models, knowledge transfer, and human reasoning. Those threads kept converging on one problem: the way we work with AI is broken at the transfer layer. We send prompts when we should be onboarding. This platform is that research, turned into something you can ship with.