The landscape of software development has shifted, making personalization of tools astonishingly easy. Previously, engineers primarily used and customized off-the-shelf software, with few developing personal projects due to questionable ROI and maintenance pain.

Now, with the advent of AI agents, personalization is significantly more accessible. Two key prompts enable this: downloading and building software from source for local use, and instructing an agent to manage ongoing updates by rebasing local changes onto upstream releases. This dual benefit—easier initiation and sustained maintenance—transforms the economics of customization.

An example is the integration of meat.dev, a tool for minimizing code diffs, into an agent called Shelley. Instead of manual integration into complex systems like VS Code or Vim, a single prompt allowed Shelley to build meat.dev, process commits in the background, and add a user-friendly toggle. This illustrates how agents can handle the heavy lifting of source code modification, making personalization far more powerful than traditional configuration or plugin systems.

This agent-driven personalization eliminates the need for extensive plugin architectures or complex configuration files. Users can now instruct agents to modify software directly, such as changing font sizes or replacing hardcoded elements. This capability extends to small teams as well, fundamentally altering how software products can be reinvented.