Personal Agents, Clearly Explained

About this video

Jeff explains personal AI agents in plain English by contrasting three categories. Chatbots like ChatGPT, Claude and Gemini chat are good for one-off tasks but only see what is shared in that task — for example, a single health report with no connection to past reports or doctor’s notes. DIY agents like ChatGPT Work, Codex, Claude Code and Gemini Spark can remember and update previously shared information: given access to a health folder, they save a new report, compare it with five years of reports and notes, and update a health memory document. He describes his own setup, ‘Jeff OS,’ as over a hundred text files across more than 30 folders that the AI writes and organizes; a governance file, for example, ensures every reply ends with a ‘what’s next’ section. One example flags a brother’s severe shellfish allergy when planning a surprise birthday at a restaurant. Personal agents — he names Meta’s Muse, ChatGPT Dots and xAI’s GrokBot — are, in his framing, the proactive, ready-out-of-the-box version: they notice things like an upcoming health checkup and offer help before being asked, and run on dedicated cloud computers so they can work when a laptop is closed. He notes trade-offs: personal agents are newer and less capable, offer less control, and risk ecosystem lock-in, while DIY setups are portable across tools. On privacy, he says it is a personal preference, that he is wary of Meta after Cambridge Analytica, and that privacy-focused options like OpenClaw and Hermes run locally but are intimidating and cannot run the most powerful models.

Key takeaways

• Chatbots handle one-off tasks with only the information in that task; DIY agents remember and update files across tasks, and personal agents add proactive help before you ask
• DIY agents work by saving rules, preferences and memory in plain text files (e.g. memory.md) and reading them when a task calls for them
• Personal agents are the beginner-friendly, out-of-the-box version of the same file-based architecture, with guided setup such as goal interviews in Muse
• Proactivity distinguishes personal agents from scheduled tasks: they can check throughout the day for job postings, price drops or renewals and notify you as events appear
• Choose a personal agent for ease, a DIY agent for power, control and portability — accepting personal-agent lock-in and privacy trade-offs

Episode details: published October 13, 2026 · 15:21 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are Jeff Su’s own claims, shared for education — they are not typical or guaranteed.

← Previous: Stop Organizing Files Yourself.

All Jeff Su videos · Jeff Su profile · Jeff Su (official site)