I Built a 90-Minute Founder System With Claude (Works While I Sleep)

90-Minute Founder OS

The video demonstrates a “90-Minute Founder OS”: a workflow for making progress on a side business while keeping a full-time job. Its core division of labor is Claude prepares and organizes; the founder makes decisions, talks to customers, and sends messages. The system aims to turn limited evening time into focused decisions rather than tool experimentation or administration.

Operating system and setup

• Keep business context in one place. A Notion workspace contains five lists:
  • Founder profile: Background, business idea, target customer, constraints, and available work times.
  • Evidence log: Customer statements recorded in their exact words, with the date, source, and business stage.
  • Session log: What happened in each work session, so the founder can resume quickly next time.
  • Decision queue: Items requiring the founder’s approval or rejection.
  • Later list: Good ideas that are premature and should not distract from current validation.
• Use a persistent Claude project with instructions and files. Connect Gmail and Google Calendar so Claude can prepare email drafts and plan around the founder’s actual availability.
• Create a dedicated Gmail label and a recurring calendar block for founder work.
• Set a strict approval boundary: Claude may research, draft, and organize, but must not send messages, publish, spend money, sign up for services, or book meetings.
• Start with a setup interview covering the founder’s experience, idea, target customer, evidence of the problem, schedule, constraints, and what would make the next three months worthwhile. If the founder has no idea yet, the process suggests three options based on their experience and existing contacts.

Validation playbook

Claude is instructed to work through five stages in order:

1. Real problem: Who has the problem, and how do they describe it?
2. Clear promise: Can the proposed value be explained simply?
3. First “yes”: Will someone commit to a meaningful next step?
4. Execution: Can the business deliver the solution once, properly?
5. Repeatability: Does it work again with people beyond the founder’s initial circle?

The system prioritizes the first unanswered question, rather than whichever task feels most exciting. For each stage, Claude identifies what remains unknown, proposes the smallest useful test, and prepares the work the founder needs to do personally.

• The presenter recommends aiming for 30 customer conversations as a concrete validation goal, while emphasizing that the number alone proves nothing. The important output is specific evidence about what customers actually do—not general expressions of interest.
• Record customers’ exact words and distinguish a specific recent problem or request from vague encouragement.
• Defer premature “busy work”—such as logos, websites, social accounts, and company setup—until evidence supports moving forward.

Daily and weekly workflow

• End each founder session with a clear assignment for the next workday. Each task should specify:
  • The goal.
  • What Claude should prepare.
  • Any constraints, such as using public information only or not contacting anyone.
  • Where the work and drafts should be saved.
  • What to do if information is missing: record the gap and continue with the rest.
• A scheduled weekday day shift can conduct research and prepare drafts while the founder is at work. In the video’s example, it runs at 9:00 a.m. in the cloud, even if the founder’s laptop is closed. The presenter recommends testing the first run manually before relying on it.
• An evening preparation task checks for new replies, logs them as evidence, reviews the calendar, and proposes an agenda. If the founder has less than 45 minutes, it prepares a shorter session.
• For a short or tiring session, Claude narrows the work to one decision that unlocks the next step, plus any action only the founder can take.
• A Sunday CEO brief reviews the week’s sessions, evidence, and decisions; recommends whether to continue, change direction, or shelve the idea; suggests work times for the coming week without adding them automatically; prepares Monday’s task; and updates a dashboard.

Example: Tom’s supplier-price-check idea

Tom is a Manchester purchasing manager developing a service to help independent cafés identify when they are overpaying suppliers.

• The initial concept was a monthly supplier-price check.
• Claude prepared a list of 10 independent cafés in Greater Manchester, with sources and draft outreach messages. Tom reviewed and sent any messages himself.
• Four sample replies illustrated how to assess evidence:
  • “Sounds useful” was treated as encouragement, not proof.
  • A café owner comparing invoices on Sundays described a specific behavior.
  • An owner whose accountant already handled the task revealed an existing alternative worth investigating.
  • A request to check invoices before a Tuesday order indicated a concrete need and deadline.
• Taken together, the replies suggested the problem might arise right before an order, rather than once a month. Claude therefore recommended testing a pre-order check instead of assuming a monthly report was the right product.
• Suggested follow-ups included responding to the owner with the upcoming order, testing a small check with a few owners, and learning more about the accountant’s existing role. All communications remained drafts for Tom’s approval.

Metrics, targets, and operating rules

• Time budget: Approximately 90 minutes per day is the intended founder-work allowance.
• Validation target: The presenter suggests 30 customer conversations, while stressing that the quality of evidence matters more than the count.
• Planning horizon: The setup interview asks what would make the next three months worthwhile; weekly reviews assess progress.
• Short-session threshold: Less than 45 minutes triggers a shortened agenda. The example demonstrates a 25-minute session.
• Review cadence: Weekday preparation and research, plus a Sunday 5:00 p.m. CEO briefing in the example.
• No revenue, margin, customer acquisition cost, or other financial performance results are reported. The demonstrated measures are progress in tests, evidence gathered, open questions, and pending decisions.

Sources and presenters

The presenter and host are unnamed in the subtitles. The video is from AI Founders. HubSpot is identified as the sponsor and is mentioned as the source of a free AI Agent Starter Kit and an AI agents report.

The workflow uses Claude, Notion, Gmail, and Google Calendar. The presenter also references the Founder OS/Ark OS materials and the Founders Hive community.