There is a post on @Shopify's engineering blog admitting that their first automated judge agreed with humans barely better than a coin flip. They published the whole road from there to near-human agreement, including the ways their own models tried to cheat. Publishing your failure modes is a costly signal. The radar shows where it lands on the seven dimensions.
Next time a dashboard shows you an AI growth multiple, ask one question: what is the absolute base? If there is no denominator, you are looking at a story, not a metric. That is the denominator test. Even @Shopify, the most AI-native large company on the market, keeps its denominators private. Yours should at least be visible to you.
@solopribuilds Well, it really depends on where you want to start. You can build something for less than $500 at home, but if you want to start a company you'll need millions
I just released the last article on Shopify vs AIMF - AI maturity framework and I’m already working on the next one!
Next up is BBVA vs AIMF. Research started 3 hours ago and 250 agents are already running in the background to analyze data against the methodology.
Every few weeks, when a new AI model ships, the CEO of one of the largest commerce platforms in the world runs it through a folder of prompts he has curated for years, each one paired with the answer he expects. Decisions that reshaped @Shopify came out of that folder.
That ritual says more about AI maturity than the memo that made the headlines. I scored the whole public record on the seven dimensions of AIMF® (AI Maturity Framework): four dimensions at +4, a production system documented more openly than any company I have tracked, and one number nobody outside can check.
Episode two of the readout series, with every score explained and the sources linked:
https://t.co/h2S0nI4UF3
In April 2025 an internal memo was about to leak, so the author published it himself. It made AI usage a baseline expectation for everyone at @Shopify and set one rule people still quote: before asking for more headcount, show why AI cannot do the job. A year and a half later the doctrine shows up as arithmetic, with headcount flat while revenue keeps climbing.
How mature is @Shopify, really, on AI? Tomorrow you get the full answer.
I scored the whole company on the seven dimensions of AI maturity, from public evidence only: the filings, the earnings calls, the engineering blog. Underneath the famous memo there is a system almost nobody talks about, and it is the best part of the story.
See you tomorrow.
Every AI vendor contract caps liability, and every cap binds the two companies that signed it. The person the agent actually failed signed nothing with anyone. Europe just wrote that into product law: toward the harmed person, liability cannot be excluded by contract. Who answers when the agent fails, link by link:
https://t.co/R903FgKirr
I spent the past week inside @Shopify's public record: the earnings calls, the annual reports, the engineering blog, the developer docs. The famous AI memo is in there, and it is the least interesting page in the file. What sits underneath is a better story. Monday.
When an AI agent makes a mistake, who answers for it?
An airline in Canada argued its chatbot was 'a separate legal entity responsible for its own actions'. The tribunal rejected that in a paragraph and made the airline pay. A public database counts over 1,600 court cases of AI-invented citations in legal filings. A rejected job candidate is suing the vendor of an AI hiring screen directly, and the court is letting the case proceed.
This week's post follows that question along the whole chain: the lab that trains the model, the builder, the deploying company, the employee, the customer. Two questions decide every verdict: what could each one see, and what could each one change. Full piece in the first reply.