Apr 24
Loading video player...
Testing GPT 5.5 “Spud” in Codex: Faster Agents, Browser/Computer Use, and Knowledge Work - This video covers OpenAI’s new GPT 5.5 model, code-named Spud, and why it is stronger for real work, long tasks, tool use, and finishing complex goals. - GPT 5.5 pricing and why model cost should be judged by task quality, token efficiency, retries, and completion time, not only price per million tokens. - How GPT 5.5 compares to GPT 5.4, especially better intent understanding, stronger knowledge work, improved physics, and better browser/computer control. - Codex as the best and cheapest place to try GPT 5.5, including model effort settings and usage on the $20 plan. - GPT 5.5 inside Codex creating apps, spreadsheets, documents, slide decks, simulations, notes boards, chess games, and diagrams. - covers new Codex browser use and computer use features, showing GPT 5.5 controlling web apps, Canva, Finder, and Claude to complete multi-step workflows. TIME STAMPS 00:00 Intro 02:02 GPT 5.5 Pricing 04:15 Thoughts on Use Task Based Measurements 04:55 5.5 Understands Intent 05:36 5.5 vs 5.4 06:39 5.5 in Codex (I built an 3d simulation) 07:58 Best place to use 5.5 (Cheapest ) 10:17 Testing Browser Use on Codex 13:30 Testing Browser use on Chess 15:07 Creating Diagrams with Browser Use 18:00 Controlling Computer 21:57 Final Thoughts