Gemini Spark: Google’s Most Powerful Personal Agent
About this video
Jeff Su presents Gemini Spark as Google’s proactive, beginner-friendly AI agent, contrasting it with the reactive Gemini chatbot: chat can find and summarize Drive files but not rename or move them, while Spark can watch a folder, rename uploads based on their contents, and file them correctly; chat writes a report only when asked, while Spark builds it autonomously on a daily or weekly schedule; chat triages an inbox with inconsistent formatting and cannot archive or label, while his Spark triage workflow runs on its own, formats summaries his way, proposes labels, and archives irrelevant mail automatically. Setup: in Gemini settings under Personal Intelligence, turn on memory, and under Connected Apps enable the Google Workspace and search services toggles (enterprise accounts see only the Connected Apps tab), giving Spark direct access to Gmail, Calendar, and Drive; then create a folder called Spark OS in the root of My Drive. Capability one, ‘connect the dots,’ is cross-app context plus action: he has Spark build a spreadsheet of this year’s World Cup matches with dates and teams (requiring web search) and add predicted winners pulled from a recent meeting transcript (requiring calendar search), and in a work example Spark identifies the right ‘Austin’ from his email, reads the thread, reuses the time slot of their last meeting, and drafts a calendar invite for October 1st for his approval. Capability two is following templates: a Spark Templates folder inside Spark OS holds his weekly report template, and Spark turns a messy sales-team intake document into a leadership brief matching that template. Capability three is skills — reusable instruction sets created in chat (his example first creates a 99_temp folder inside Spark OS, then has Spark build a weekly report skill that reads the latest intake entry, formats it to the template, and saves to 99_temp, after restating the goal and showing the steps for his approval); skills are invoked with ‘/’ in a new chat, edited conversationally, shared via download, and imported by uploading a skill.md file under Skills, with his own inbox-zero skill offered as a downloadable starting point.
Key takeaways
- Turn on Personal Intelligence memory and the Google Workspace connected-apps toggle, then create a Spark OS folder in Drive root — Spark’s advantages only work with that access in place.
- Use Spark for cross-app tasks you would otherwise do manually: it can pull context from Gmail, Calendar, Drive, and web search in one instruction and act on it.
- Template repeated deliverables once (a Spark Templates folder) and point Spark at the template instead of re-explaining the format every time.
- Convert repeatable template work into a skill: make Spark restate your goal and show its steps for approval before saving, invoke it with ‘/’, and edit it conversationally later.
- Schedule skills on time-based triggers and use event triggers (such as an incoming meeting-notes email) for automatic briefs — remembering event triggers are created by talking to Gemini, not in the manual builder.
Episode details: published June 30, 2026 · 15:43 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: How I’d Learn AI From Scratch in 2026 (skip the useless 80%) · Next: Claude Design is Insanely Easy - Become a Pro in 14 Minutes! →
All Jeff Su videos · Jeff Su profile · Jeff Su (official site)
