Jason compares learning an AI tool’s strengths with learning how an employee works: observe output and adjust expectations.

He uses throwaway projects so that a failed experiment teaches him something without risking an important project.

When models improve, he revisits older specifications to see which previously difficult tasks now work.

Jason’s advantage for an experienced operator is judgment: recognizing the wrong direction and correcting it.

He reports a VSL writer compressing work from roughly three months to 16 days, while still reviewing and revising the output.

John asks where junior roles fit. Jason calls it the hardest part and turns to education; the discussion does not resolve a junior hiring model.

Jason argues that new platforms and tools create openings, while an advantage tends to shrink as more competitors adopt it.

He suggests looking at a specific traffic source or vertical where deep expertise can matter, rather than copying the most visible play.

His broker example is about margin pressure: extra middlemen become harder to sustain as a market gets more efficient.

Jason compares a creative workflow with documented operating procedures: give the generation step a repeatable process.

He proposes an explicit human rating step, using a scale or thumbs-up/down to make feedback usable.

Then improve the workflow itself: add a step, strengthen a skill, or add human review where quality falls short.

Jason breaks a page into four pieces: layout, copy, imagery, and a design system.

Precise terms such as wireframe help communicate intent. Learning the vocabulary makes requests easier to specify.

He studies a reference page’s structure and section purpose, then combines that structure with content and visual references.