Amp July 9, 2026 Changelog

This update, called "The Dial," completely simplifies how you work with Amp's AI coding assistants. Instead of making you guess which underlying AI model is best for your project, Amp now lets you choose a setting based purely on how difficult your task is—saving you time, money, and frustration. This matters to anyone using Amp because it takes the guesswork out of AI coding, giving you the exact amount of brainpower you need for the job.

What's new

• The Dial (Low, Medium, High, Ultra): A new, streamlined way to control the AI. Instead of picking models by name, you choose a level based on the task's difficulty:
• Low: Best for simple, highly specific tasks like quick bug fixes or basic cleanups.
• Medium (Your new default): Great for everyday coding, even when your instructions are a bit messy or vague.
• High: For complex coding where mistakes are costly. It takes about twice as long but delivers highly polished, review-ready code.
• Ultra: For massive, multi-file projects and major system changes where the AI has to figure out the path as it goes.
• Built-in "Oracles" (Double-Checkers): Every mode now automatically uses a second, highly capable AI model (an "oracle") to review the first model's work before showing it to you, ensuring much higher quality.

Improvements

• No More Manual Effort Toggling: You no longer have to manually adjust "reasoning effort" (how hard the AI thinks) using keyboard shortcuts. It is now automatically set to the perfect level based on where you put the dial.
• Smarter Defaults: The old "Smart" and "Deep" modes have been upgraded and merged into the new "Medium" default, giving you better performance right out of the box.

Breaking changes

• The old modes (*smart*, *deep*, *rush*, and *large*) have been retired. If your workflow relies on them, don't worry—you can easily install them as classic plugins to get the exact same prompts and models back.