QuantOS

Inspiration

Quantitative research often produces promising charts without producing disciplined decisions. Strategy ideas live across notebooks, spreadsheets, chat threads, and slide decks. Assumptions are hard to compare, costs are added late, and weak ideas can consume weeks of engineering effort before anyone asks whether the edge survives real execution.

I built QuantOS to make research behave like a governed experiment portfolio: every hypothesis has a clear owner, evidence record, economic model, and decision path. I also built it as a weekend-builder dad with a full-time job and no quant firm behind me. Codex and GPT-5.6 made it possible to turn a large product idea into a working, coherent system in a compressed schedule.

What it does

QuantOS organizes strategy candidates into comparable research groups and moves them through an evidence-constrained investment committee workflow. Researchers can inspect assumptions, provenance, experiment status, modeled P&L, trade timing, exit logic, and the reason a strategy should advance, continue collecting evidence, enter review, or be rejected.

The demo follows WalletEdge. Its headline result appears attractive at +23.9 basis points, but after latency and realistic execution costs the deployable edge falls to only +2.3 basis points. QuantOS makes that collapse visible before a team commits another modeled $18,000 of engineering effort. The point is not to generate a louder signal; it is to make the decision traceable and economically honest.

How we built it

Codex with GPT-5.6 was the primary development environment and engineering partner. I used it to translate research-governance concepts into product architecture, build the landing page and dashboard, implement portfolio and decision-lab workflows, create evidence-review interactions, develop deterministic economics calculations, add provenance and rejection records, improve novice-facing explanations, test the application, and prepare it for deployment.

The application uses Next.js, React, TypeScript, Vite/vinext, Node.js tests, local structured evidence data, and Cloudflare-based hosting. The repository includes setup instructions, sample data guidance, architectural context, and a detailed account of where Codex accelerated the work and where key product decisions were made.

Challenges

The hardest problem was not generating UI. It was translating expert quantitative reasoning into an experience that a new user can follow without weakening the underlying rigor. The live demo needed clearer selection states, explicit guidance between workflow stages, slower evidence-review pacing, understandable economics, and a conclusion in which the decision rationale remained visually dominant.

A second challenge was separating persuasive storytelling from deterministic computation. QuantOS uses AI-assisted workflows for synthesis and explanation, while preserving explicit inputs and reproducible calculations for the economics that drive decisions.

Accomplishments that I am proud of

• A complete landing-page-to-dashboard product experience rather than a static concept
• A governed research portfolio with evidence, economics, provenance, decisions, and a strategy graveyard
• A novice-readable demo that explains why an apparently positive result may still make no economic sense
• A public working deployment, documented repository, automated tests, and a sub-three-minute narrated demo
• Building the system part-time as an independent builder while balancing work and family

What I learned

AI-native development changes the feasible scope for an individual builder. The biggest gain was not raw code generation; it was the ability to move continuously between product thinking, implementation, testing, design critique, deployment, and communication. GPT-5.6 and Codex helped maintain context across those layers while I retained control of the research rules and final decisions.

What's next for QuantOS

The next step is to connect QuantOS to live experiment infrastructure, richer market and execution data, multi-user review, and auditable approval policies. The longer-term vision is an AI-native research organization in which small, focused teams can explore more ideas while being more disciplined about which ideas deserve capital.

Built With

• next.js