We've Been Using GPT-6 Astra for a Few Weeks, Here's What We Think...

<p>Theo &amp; Ben break down OpenAI's latest model, Astra, and why it's their new benchmark for AI coding, computer use, and multimodal work. But it's not all good: they explain why its UI, stopping behavior, and agent reliability still create friction in real software workflows. From DEFCON puzzles to coding-agent PRs, we're comparing Astra with Fable and asking what a trustworthy OpenAI model should do next.</p><p><br /></p><p>Thank you to PostHog for sponsoring today's episode!</p><ul><li>PostHog, all-in-one suite of product tools: <a href="https://nerdsnipe.link/posthog" target="_blank" rel="ugc noopener noreferrer">https://nerdsnipe.link/posthog</a></li><p><br /></p></ul><p>Listen wherever you get your podcasts:</p><ul><li>Spotify: <a href="https://nerdsnipe.link/spotify" target="_blank" rel="ugc noopener noreferrer">https://nerdsnipe.link/spotify</a></li><li>Apple: <a href="https://nerdsnipe.link/apple" target="_blank" rel="ugc noopener noreferrer">https://nerdsnipe.link/apple</a></li><li>Elsewhere: <a href="https://nerdsnipe.link/listen" target="_blank" rel="ugc noopener noreferrer">https://nerdsnipe.link/listen</a></li><p><br /></p></ul><p>Sources available on our Substack:</p><ul><li><a href="https://nerdsnipe.substack.com/" target="_blank" rel="ugc noopener noreferrer">https://nerdsnipe.substack.com/</a></li><p><br /></p></ul><p>Timestamps</p><p>00:00 Meet Astra</p><p>04:03 Best Model Ever, With Catches</p><p>06:33 Reasoning and 3D Benchmarks</p><p>20:00 Astra Rebuilds Ping.gg</p><p>30:04 Instruction-Following Problems</p><p>44:47 The Uncommitted Fix Debate</p><p>01:09:37 The PR Babysitting Failure</p><p>01:22:41 Why Fable Still Wins</p><p>01:30:41 Multimodal and Computer Use</p><p>01:39:29 Astra vs. Fable Fleet Data</p>