Jason automates a "chief of staff" thread that runs multiple times a day. It scans his email, Slack, and Linear board to synthesize priorities, pre-draft responses, and even manage tasks like flight check-ins, creating a focused overview of his work.
Jason creates personalized "write me" skills by instructing his AI to analyze his past Slack messages or sent emails. This allows the AI to learn his communication style, tone, and how he tailors messages for different audiences like executives versus team members.
Jason created a meta-skill that analyzes session logs of his other AI skills. It identifies the most frequently used skills, reviews the user feedback given in those sessions, and then suggests or automatically implements improvements, creating a self-correcting system.
Instead of meticulously writing a project spec, Jason describes his high-level objective to his AI and has it generate its own `goal.md` (success criteria) and `plan.md` (implementation details). This approach often yields better results because the AI sets a plan it can execute effectively.
The primary job for humans collaborating with AI is to be dissatisfied with its output and learn the vocabulary to explain *why*. Progress comes not from coding solutions, but from clearly articulating problems with the AI's work, like a poor UX or inefficient design.
Jason uses a single Obsidian Vault as the foundational project for his AI agent. This collection of markdown files contains all his context, notes, and preferences. By starting every task from this vault, he ensures the agent always has access to a persistent, well-structured knowledge base.
While on a bike ride, Jason used his phone to remotely command his AI to find a video file on his laptop, re-export it with requested subtitle changes, post it to Slack, and then autonomously handle subsequent revision rounds based on team feedback.
When shopping online, Jason has his AI research products and then open each final recommendation in a separate browser tab. This presents him with a pre-vetted set of options ready for his final review, letting him make the high-context decision instead of reading a summary.
Jason organizes his AI work using pinned threads as dedicated, persistent workspaces for each project. He leverages the AI's strong context compaction to maintain long-running conversations without losing history, effectively turning each thread into a self-contained project environment.