GPT-6 Astra is a banger - here’s everything I’ve built

The episode kicks off with an excited introduction to GPT-6 Astra, highlighting its early rollout for Daybreak customers and the speaker’s early access to test the model. Astra is portrayed as a revolutionary AI that surpasses previous models like GPT-5.6 or Fable in numerous tasks, offering significant improvements in understanding, speed, and functional capabilities like coding, math, and knowledge work. The host emphasizes Astra's ability to use software and browser functionalities, making it much more autonomous and effective at automating complex workflows and saving time.

The discussion dives into Astra's key strengths, particularly its enhanced computer and browser use, which enables it to navigate complex user interfaces, interact with web applications, and perform multi-step tasks. Examples include managing CRM workflows, generating marketing assets, and conducting QA in web environments. The model is also praised for its improved mathematical reasoning, coding skills, and ability to handle unstructured data for insights, as demonstrated through the ChatPRD product intelligence feature.

The host shares a range of creative and practical projects developed using Astra, illustrating its versatility. These include building custom hardware interfaces like a CLI for a Bluetooth device, automating thumbnail creation in Flora, designing a 3D character in Blender, creating a 3D journey for kids, and designing a retro AIM-style chat app. Each example underscores Astra’s ability to perform complex tasks in a single shot, including hardware hacking, 3D modeling, and software automation, which were previously very difficult or impossible.

A key theme is the resurgence of UI-driven interaction, facilitated by Astra's capacity to interact with complex web UIs and click buttons, representing a shift back to UI as the primary user interface. This has major implications for SaaS and MCPs (multi-channel platforms), where button-based workflows are becoming more accessible through AI automation. The host envisions a future where traditional CLI or MCP interactions might be replaced or supplemented by more intuitive, button-based AI interactions.

The podcast concludes with considerations about Astra’s speed, pricing, and practical value. It is described as a robust daily driver despite some slowdowns, with a cost of around $10 per million tokens. The host recommends trying Astra for ambitious tasks like automating workflows, building 3D games, and hardware hacking, emphasizing its ability to significantly accelerate personal and professional projects. The episode invites listeners to share their creations and stay tuned for future comparisons with other models.