Build the operating system
Keep learning, serve buyers, and connect automation with expert review.
5 lessons · Use the arrows or scroll to move between boards.
A tactic that worked can stop working.
Jason explicitly says he is not currently in the media-buying trenches. He answers from a systems perspective rather than offering a current account recipe.
shareCampaign structures change. He emphasizes adapting the system instead of attaching identity to one setup.
shareHis practical input loop includes a small group of active peers, shared learning, and a willingness to keep testing.
share
Just be testing, assume you don't know anything, assume the market's changing, assume the ad networks are changing.
The vertical is only part of the decision.
Jason sees opportunity in serving direct buyers and pairing acquisition expertise with sales expertise.
shareEfficient buyer onboarding matters. Contact intent and the path to a useful conversation also vary by vertical.
shareHis larger question is the operating model: a broad market with a team, or a focused niche with a different level of complexity.
share
how do you want to run your business?
A system earns its value by working repeatedly.
Jason sees lead and call operations as heavily driven by repeatable data tasks. Operational efficiency is a scaling lever alongside creative.
shareHis test of a useful workflow is whether it continues producing results without his constant attention, even if some steps use model judgment.
shareHe learns unfamiliar concepts by asking for simple explanations and walking through the steps. That vocabulary helps him specify the next system.
share
I can go on vacation and it continues to produce results for me.
The expert stays responsible for the finished workflow.
Jason describes an experienced ad-operations person giving the system buyer specifications and integration requirements.
shareIn his platform example, a structured prompt drives a funnel build and browser-based testing.
shareThe flow checks that the platform received the result, then he reviews it as a human. He can ask for design and content revisions afterward.
share
I'm going to look at it as a human in the loop.
A team needs the instructions as well as the code.
Jason describes broad autonomy in experiments involving things that do not matter. That low-stakes context is part of the example.
shareHis solo setup kept skills and agents in global configuration. Adding a collaborator changed how he packaged that context.
shareHe moved it to the project so a repository checkout brings the skills too. His credential-handling description is incomplete and is not a security blueprint.
share
Now we've moved to a project level so that when you pull down the repo out of GitHub, you pull down all the skills with it.