A show about people already living with personal AI.
Personal AI is already changing how a small group of people live, create, decide, and relate to the world. Already Here meets those people and makes their future visible.
The tools are the evidence. The real story is what changes in how people work, how they make decisions, and where they put their attention.
Latest episode
EP. 01 · JULY 29, 2026 · 46 MIN
AI Consultant: Stop Clicking Through Apps—Your AI Agents Should Work for You
With Sam Gaddis
Sam Gaddis is a founding partner at Runpoint, where he leads engineering and product. He has spent more than a decade building and shipping AI systems, from autonomous development pipelines to enterprise platforms. In this conversation, the interesting part is not how many tools Sam uses. It is how deliberately he decides what each one may see and do.
Sam does not use one all-powerful assistant. He uses focused helpers to gather context, prepare work, catch routine failures, and save ideas before they disappear. Michael sees those boundaries in practice: the system can prepare an email but not send it, work in Sam’s own software projects but not client repositories, and offer feedback during a difficult conversation without making the decision for the people involved.
What you’ll learn
Sam built a private desktop tool that transcribes a conversation with his wife, distinguishes the speakers, flags destructive patterns, and charts the emotional direction of the exchange. Later, it shows him that he is speaking about 60 percent of the time and can recommend a pause when the conversation becomes too heated. The useful pattern is not “put AI in a relationship.” It is turning principles they already trust into timely, observable feedback.
Sam’s most common work request starts with retrieval: find the latest meeting notes, locate the relevant client email, then prepare a response based on both. The assistant earns its place by reducing the search and assembly work that happens before judgment.
Sam’s email workflow ends in Gmail drafts. He reviews the facts, changes the tone, and presses send himself. This is a clear way to gain leverage from an agent without quietly delegating a consequential external action.
Hermes can work in Sam’s personal and Runpoint repositories, but it cannot touch client repositories directly. Capability and access are separate decisions. A system can be powerful inside a deliberately narrow boundary.
A nightly Codex automation reviews error logs and fixes routine problems in Sam’s setup. Early on, this caught enough failures to matter; after the system stabilized, it rarely finds anything. Reliable agents still need observability, repair paths, and time to become boring.
Instead of manually copying new website language into personal and company LinkedIn pages, Sam asks an agent to read the source, make the updates, and leave the result for review. The shift is from clicking through an interface to describing an outcome, while a person still inspects the work.
When an idea arrives in the car or sauna, Sam speaks it into his phone and asks the system to organize it, research it, and prepare a rough artifact. He waits until he is back at his desk to decide whether it deserves deeper work. Capture and commitment are intentionally different stages.
Topics and resources
- Runpoint — Sam’s consulting company
- Hermes — the agent Sam operates through Slack
- OpenAI Codex — used for browser work, coding, and nightly error review
- Claude Code — used alongside Codex for project work
- Granola — the source for meeting notes in Sam’s email workflow
- Google Workspace CLI — access to Gmail and other Workspace tools
- Fight Right — the book behind Sam and Victoria’s conflict experiment
One safe experiment
Choose one email you need to answer. Give an AI assistant the relevant notes and message, then ask for three possible drafts. Check every fact, choose the tone, and press send yourself.
About the show
Michael Galpert meets people already living with personal AI in everyday life. Each episode looks closely at what they built, the boundaries they chose, and what changed once the system became part of the way they work or live.