Founder Story

He still can't write code. His five-person ad agency now runs on about 270 tools he built with Claude. Every owner-run firm needs its first five

Jacob Malherbe couldn't write code in February 2026 and still can't. He says his five-person agency, which buys ads for law firms, now runs on about 270 small programs he built with Claude, from the inbox to the phones to the ad copy (self-reported). The opening is building the first five, and keeping them alive, for owners who never will.

Published Nov 12, 2026

The Wow

What someone built

An inbox agent that sorts and unsubscribes, an AI receptionist, a campaign dashboard that rewrites ad copy for each image, client report dashboards and sales-research agents, all on Cloudflare, each started as a one- or two-sentence note to Claude.

Founder Story

License
A personal post and site; nothing to install. Quote and link only; build your own tools for your own clients
What you need
  • A coding agent that can deploy what it writes
  • A hosting account in the client's name
  • One owner willing to name their most annoying weekly chore
To use it
Nothing to install. Read his post for the sticky-note method and the failure list, then look at the free tools on his site.
A business you could build on it
First five tools for an owner-run service firm

Why it's remarkable

His whole spec is a sticky note: what should happen, and what should happen when it fails. He also published what broke: two outages from one empty AI account and a daily job that failed without a word.

The Opportunity

The business it reveals

Sell owner-run service firms a fixed-price sprint that turns their five worst weekly chores into working tools with failure alerts, then a monthly plan that watches, fixes and prunes them.

Who would pay

Owners of small agencies, bookkeeping practices, property managers and trade offices who already pay for contractors and software but will not build with an agent themselves.

The Flip

2 ways you could turn this into a business

Silent-failure watch for owner-built tool fleets

Who pays
Owners and small teams who already built a handful of agent tools and have been burned by one stopping quietly
How it would earn
A monthly plan per firm, our hypothesis about $200 for up to 30 tools: a heartbeat on each tool, separate keys and spend caps, alerts to the owner's phone, and a monthly prune
Why it might work
The founder's own failure list is the product spec: two outages from one empty AI account and a daily job that failed with no error. Heartbeat monitors and spend dashboards exist, but someone still has to set them up for each tool and act when they fire.
What already exists
Healthchecks.io (free for 20 jobs, list price), Cronitor ($2 a month per monitor plus $5 per user, list price), and the founder's own Leash spend monitor (free beta, $29 a month plan listed); these are tools, not a person watching.

Why we'd try "First five tools for an owner-run service firm" first: It sells the result owners want, working tools around their own chores, at a fixed price they can say yes to, and every sprint naturally turns into the monthly watch plan.

Build It

How you could build it

A repeatable five-tool sprint for one vertical: an intake call that lists the owner's weekly chores, five two-sentence job notes, five tools deployed on the owner's accounts, a failure alert on each, a one-page registry, and a 30-minute team handover.

First step: Ask three owners you know in one vertical for their single most annoying weekly chore, write each as a two-sentence note with what should happen when it fails, and build one of them this week.

Agents and tools

  • A coding agent you already use and trust to deploy: ChatGPT (Codex), Claude (Claude Code), Gemini (Gemini CLI) or Grok Bot. The founder uses Claude.
  • Cloudflare Workers or similar hosting in the client's name
  • A business-rules file per client that the agent reads before every job
  • A heartbeat monitor such as Healthchecks.io or Cronitor
  • A spend view per key, such as your model provider's dashboard

Setup

  1. Pick one vertical and write down the ten chores you hear about most.
  2. Run a 45-minute intake call: list the owner's weekly chores and pick five that do not need anyone's judgment.
  3. Write each as a sticky note: what should happen, and what should happen when it fails.
  4. Have your coding agent build and deploy each tool on the client's account, with a separate key and spend cap.
  5. Use each tool like a customer, fix it in plain English, then add a failure alert and a registry line.
  6. Hand over with a short screen recording per tool and offer the monthly watch plan.

What the agent does

  • Writes, deploys and fixes each tool from the plain-English notes
  • Drafts the registry and a one-page guide per tool
  • Summarizes failures and spend each month

What you do

  • Chooses which chores to automate and which steps need a human approval
  • Tests every tool before the client sees it
  • Owns client access, prices and support when an alert fires

Your first customer: An owner you already know in one vertical, offered their single worst weekly chore as a working tool within a week before you quote the other four.

Test demand first: Send a one-paragraph offer to 20 owners in one vertical: name your worst weekly chore and get a working tool in a week, with a written fixed price for five. Success: at least three reply with a chore within a week and one pays a deposit. If nobody does, try a different vertical or a smaller first offer.

Your starter instructions

Paste this into the AI agent you use (ChatGPT, Claude, Gemini, Grok Bot or another) as your first message. Copying it starts work in your own agent; it doesn't create or deploy a business by itself.

Get a launch plan

Running it

What stays human: Interviewing the owner and writing the job notes, testing every tool like a customer, deciding what touches money or clients and needs a human approval step, setting up accounts and budgets in the owner's name, training the team, and answering when an alert fires.

Startup cost (estimate)
Under $100 a month before your time (estimate) (Estimate from list prices: an AI coding plan, Cloudflare Workers Paid at $5 a month per account, and a free monitoring tier such as Healthchecks.io Hobbyist. Model usage for tools that "think" is billed per run and grows with use; Jacob says the AI bill is real and he uses the cheapest model that does the job)
Running cost
Mostly your hours plus per-run model fees on the client's own account (estimate) (Rough arithmetic: each client pays hosting and model usage on accounts in their name; your cost is the build and support time per tool, not measured yet)
How much the agent runs
People run it; AI helps with tasks.

Pricing your own version? Use the free startup-cost worksheet.

The evidence

What the source shows: Jacob Malherbe's own post on X and the same post on his site (Nov 10, 2026), his agency's site and his public tools page, which we read on Nov 11, 2026, plus list prices for the alternatives.

What's ours: The five-tool sprint, the silent-failure watch plan and their prices are our applications.

Worth planning around

  • The tool count and what each tool does are his own description; we have not seen the tools running inside the agency.
  • He does not say how much time or money the tools save.
  • The $250M figure is ad spend managed for law-firm clients since 2015, not agency revenue.

Sources

Sources reopened Nov 11, 2026. Spot a mistake? Tell us and we'll correct it in public.

Keep reading

The Daily Paycheck: one remarkable AI business at a time, and exactly how you could build around it. Free; unsubscribe anytime.

Get new AI business discoveries by email.