Summary
Riley Brown delivers a hands-on review of OpenAI’s newly released GPT-5.5 (internal codename Spud), focusing on practical performance inside the Codex agent platform and what the model means for knowledge workers, developers, and businesses evaluating AI tooling costs. The model is priced at $5 per million tokens—double GPT-5.4’s rate of $2.50—but Brown argues that raw per-token pricing is the wrong lens: what matters is output quality and total token consumption per completed task, and GPT-5.5’s improved intent-understanding means it often finishes tasks with fewer retries.
Brown walks through live Codex demonstrations, including prompting the model to research and compare Codex, Claude Desktop, Manis, and Cursor across product features and revenue forecasts, then automatically generating a formatted spreadsheet, a multi-page Word document with charts, and a presentation—all within a single project folder. He also covers three new Codex updates released alongside the model, including an integrated web browser preview for live app testing (mirroring Cursor’s functionality) and a new knowledge work view designed for marketing and business use cases.
The video concludes with a live test of GPT-5.5’s browser use capabilities inside Codex, which Brown identifies as the defining competitive front between OpenAI and Anthropic: whoever builds the most reliable model for autonomous browser and computer control stands to capture significant enterprise automation spend. Subscribers can also access his previously released 100-minute Codex deep-dive for full workflow setup context.
📺 Source: Riley Brown · Published April 24, 2026
🏷️ Format: Review
