Twenty lessons on starting and scaling an AI company, published every working day from 30 July to 26 August 2026.

Every lesson is anchored to one document. Not a metaphor and not a case study, but an actual artifact a founder produces or receives, in roughly the order it lands on the desk. The certificate of incorporation. The first model provider invoice. The pull request nobody reviewed. The security questionnaire. The term sheet. The reader recognizes the paper, so the lesson arrives already grounded.

It is written in the first person from inside a live build, and the company is deliberately not named. It is for founders and for people deciding whether to become one. It is not legal, tax or financial advice, and it says so wherever that matters.

Act I: Standing It Up

Entity, the cost of existing, equity, the cap table, and what you actually sell.

• Lesson 1: Pick the Entity Before You Pick the Logo (30 July). Artifact: the certificate of incorporation. Choose the entity that matches how you plan to get paid, not how you plan to be described.
• Lesson 2: What It Costs to Stay Alive Before You Sell Anything (31 July). Artifact: the registered agent invoice. A company has a heartbeat cost, and it starts on the day you file, not the day you sell.
• Lesson 3: The Founder Agreement Is the First Product You Ship (3 August). Artifact: the founder agreement. Equity is easy to give and impossible to take back.
• Lesson 4: The Cap Table You Can Still Raise On (4 August). Artifact: the cap table. Every early act of generosity gets priced by the next investor, and they price it against you.
• Lesson 5: Decide What You Sell Before You Decide What You Build (5 August). Artifact: the one-page product definition. A feature is not a company. A workflow somebody is afraid to change is.

Act II: The Economics

The wedge, cost of goods sold, pricing, the free tier, and burn.

• Lesson 6: The Wedge (6 August). Artifact: the ten-name customer list. You cannot be chosen by a market, only by a customer, so pick customers you can name.
• Lesson 7: The Cost Model Is the Business Model (7 August). Artifact: the first model provider invoice. In AI you do not discover your margin at year end. You design it at the start, or somebody else designs it for you.
• Lesson 8: Price the Outcome, Floor the Cost (10 August). Artifact: the pricing page. Charge for the thing that costs you money, or your best customer becomes your worst one.
• Lesson 9: The Free Tier Is a Loan You Are Making (11 August). Artifact: the usage dashboard. Every free user is a loan you make in compute and hope to repay in conversion. Know the interest rate.
• Lesson 10: Burn, Runway and the Twenty-Month Reality (12 August). Artifact: the runway model. Headcount burn you control. Compute burn your customers control. Model them separately.

Act III: Building It Without Breaking It

Generated code, undifferentiated output, evaluation, defensibility, and trust.

• Lesson 11: Vibe Coding Is a Prototyping Tool That Keeps Getting Promoted (13 August). Artifact: the pull request nobody reviewed. You can ship code you do not understand. You cannot run a company you do not understand.
• Lesson 12: AI Slop Is a Distribution Problem, Not a Taste Problem (14 August). Artifact: the churn email. If a model can produce your output, it will produce your competitor’s too. Sell the part it cannot.
• Lesson 13: Build the Eval Harness Before You Build the Demo (17 August). Artifact: the golden set. If you cannot detect a regression, you do not have a product. You have a demo that has not broken yet.
• Lesson 14: Your Moat Is Whatever You Would Have to Rebuild (18 August). Artifact: the architecture diagram. Ask what you would have to rebuild if you switched model providers on Monday. Whatever is left is the company.
• Lesson 15: The Security Questionnaire You Cannot Answer Yet (19 August). Artifact: the security questionnaire. Trust has a lead time. Start it before the deal that needs it.

Act IV: Growing It

Hiring, raising, distribution, retention, and sequencing.

• Lesson 16: When to Hire, and the Six Months Before That (20 August). Artifact: the job description you did not post. Hire against a bottleneck that has survived a quarter, never against a week that felt hard.
• Lesson 17: When to Raise, and When Not To (21 August). Artifact: the term sheet. Raising money buys time and sells a say. Be clear about which one you are short of.
• Lesson 18: You No Longer Optimize to Be Found, You Optimize to Be Quoted (24 August). Artifact: the referral log. Publish the thing a model would want to cite, not the thing a search engine would want to rank.
• Lesson 19: The Customers Who Stay Are the Ones Who Cannot Easily Leave (25 August). Artifact: the renewal report. The customers who renew are the ones who would have to change how they work in order to leave.
• Lesson 20: Scaling Is Choosing What Not to Do Next (26 August). Artifact: the roadmap with things crossed out. Growth is the easy part to want and the hard part to sequence.

The pattern across all twenty

Every one of these documents describes a decision that was cheap to make early and expensive to make late. None of them required unusual intelligence. What separates the decisions founders get right from the ones they do not is almost entirely whether they had looked at the document before the day they needed it.

New lessons appear here as they publish. The full archive lives under In Practice: Building an AI Company.

This is one of nine series. The closest companion is In Practice: The Other Side of the Table, which covers the same market from the buyer’s chair. The others are Dear Humans, How Do You Know, In Practice: AI in the Enterprise, Scaling AI FinOps in the Enterprise Jungle, What You Do All Day, While We Were Arguing and Who Pays for the Miracle. All nine are listed on the series page.