025 AI for Operators

Justin joined us for a wide-ranging conversation that included:

Individual Operating Systems

• Voice has largely replaced typing for Justin: he dictates using Whisper and context dumps to achieve better outputs.

• Fyxer for admin work: a single tool for transcription, calendar scheduling, and email drafts; fewer systems, more context (for ~$30/month).

• Leaning on agentic browsers: Automates email triage, spreadsheet formulas, and web research. His tip to make it work: keep tasks simple, programmatic, and independent for the most reliable outputs.

Company Operating Systems

• His first major win: a cash-collection dashboard. He vibe-coded it in Replit, unified data from Xero and HubSpot; and set it up to sync hourly, which gave the CFO and customer success leaders a real-time view of accounts receivable, leading to better customer meetings and more accountability.

• Building an “Adoption Ladder”: Learn AI’s strengths and limits, then design systems. Let AI handle technical details so that non-engineers can safely build zero-to-one.

• The Company OS mindset: Build hubs-and-spokes bridges across CRM, finance, and ops; ship small internal tools that remove you from recurring workflows.

Looking Ahead

• Career calculus: Systematize yourself before someone else does; use AI to amplify judgment and become more leverageable, not replaceable.

• His outlook for 2026: Natural-language app-building with native connectors (instead of working with painful APIs), agents get stronger but accuracy is still an issue.

Practical

• Zapier CEO Wade Foster on their 2026 AI Transformation plans (lots to learn from here)…

• …and the full presentation here

Perspectives

• The Agentic AI Adoption Matrix (Consumer) - a useful framework for thinking about applying agents to different parts of a consumer market (in this case, travel).

• The New AI Consciousness Paper: Astral Codex Ten unpacks an important new paper on how we’d identify sparks of consciousness in AI systems.

• Benedict Evans on “AI Eats the World”

• Google DeepMind AGI Policy Dev Lead Seb Krier on why we’re thinking of AI’s “intelligence” and its downstream implications all wrong…

• …and Andrej Karpathy on a similar topic (clarifying his earlier ‘ghosts/spirits’ comment (why do these topics matter? because what kind of intelligence AI is will determine what the world will look like, and what operators can achieve with it)

• Three years from GPT-3 to Gemini 3: Ethan Mollick weighs in on Google’s big release.

News

• Accelerating Science with GPT-5: A breakdown of OpenAI Science’s work. The main takeaway: it’s not AI alone that’s accelerating progress, but rather human-AI teams, with scientists setting the agenda and GPT-5 adding breadth, speed, and parallelization of work.

• Meta’s Chief Revenue Officer leaving to launch startup: another blow to Meta’s AI ambitions - they seem to be significantly behind the other AI leaders.

• Altman forecasts “rough vibes” for OpenAI after Gemini 3 launch

Thanks for reading,

Tom Guthrie

If you liked this edition of AI for Operators, share it with another operator you know!