GPT 5.5 just did what no other model could

Clarevau introduces GPT 5.5, describing it as a powerhouse based on weeks of testing, especially in complex coding tasks that previous models struggled with. She highlights its higher capacity for complex work, efficiency, and the personal value she found in its enhanced intelligence and autonomy. She details her overall positive experience, emphasizing its ability to handle advanced technical problems autonomously.

The episode discusses the pricing of GPT 5.5 and GPT 5.5 Pro, noting their high costs relative to other models, but she argues that the increased capabilities justify the expense—what she calls the 'intelligence tax.' Clarevau reflects on the ROI, especially in terms of speed and the ability to undertake ambitious, complex tasks that were previously unmanageable or inefficient.

Clarevau experiments with GPT 5.5 in ChatGPT, where she finds that it tends to think extensively—sometimes too long for simple tasks. She notes its ability to generate complex apps, such as an educational app for her child, though not always efficiently. She concludes that the true power of GPT 5.5 shines more in specialized, developer-oriented tasks rather than casual, everyday use.

The highlight of her testing is the use of GPT 5.5 Pro in Codex for tackling significant engineering challenges. She describes successfully conducting threat assessments, security scans, and executing complex data migrations involving millions of unstructured rows—tasks that previously required extensive manual effort and multiple edge case handling. She emphasizes the model’s ability to work autonomously over hours, drastically reducing error rates and streamlining technical workflows.

Clarevau explores how GPT 5.5 worked independently over long periods, such as six-hour autonomous runs, to fix bugs, address tech debt, and reverse engineer proprietary hardware—like a Bluetooth pixel display—by analyzing packet captures and generating code to control hardware. These examples highlight the model’s ability to perform deep, complex, and long-term tasks without human intervention.

She discusses customizing GPT 5.5’s personality inside Codex using slash commands to replace its default 'baked potato' tone with a more engaging, friendly persona. Her overall verdict is overwhelmingly positive, praising its technical prowess, efficiency, and autonomy, while noting its suitability primarily for developer-level work rather than casual consumers.