Chatham speeds up deal reviews with Codex: early timing comes in under 4 minutes

An app cross-checks the documents. Chatham built it with Codex and continues to test it on real deals, comparing its results with the work of experienced reviewers. The app gathers supporting documents, compares key terms and flags discrepancies.

How employees build these apps

Chatham gave employees an internal platform called Chatham Vibes: by Sep 10, they had created nearly 1,000 apps in four months. These apps' AI features use GPT-5.6 Terra by default. Employees enable GPT-5.6 Sol separately in the configuration.

First the outcome, then the tool. With the Process Zero method, employees define the desired outcome, the minimum inputs and a new workflow. They choose an app, model or platform afterward.

For your own development, install Codex CLI on macOS and Linux with `curl -fsSL https://chatgpt.com/codex/install.sh | sh`. Then run `codex` in the project folder, select Sign in with ChatGPT and describe the task.

What is available in Onyx

Chatham uses several models in Onyx: it routes simple analysis and nonproduction tests to more economical models and assigns complex tasks to GPT-5.6. Onyx teams use Codex from planning through review, and CTO John DeGunter reports a faster path from an idea to a working feature.

You can register for ONYX Pro on Chatham's website using the Sign up for ONYX Pro button.

ChatFIN in Onyx analyzes historical market data, helps users understand their portfolios and finds documents containing the terms of obligations, providing links to them.