Patrick Collison, chief executive of Stripe, wrote on X on 31 October that he is building an “improvements” feature into all of his personal vibecoded applications: “a freeform text input that’s hooked up to an agent that yolo implements whatever is submitted”.
Someone else who uses one of those apps “has already implemented 150 improvements over the course of a few days”, he wrote. The post does not give the number of days; at three days that would be 50 improvements a day, by our arithmetic.
Short explainer · Stripe CEO’s apps: 150 changes in days
A vertical news Short summarising this story. Quotes come from Patrick Collison’s post on X; our arithmetic is labelled.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on how an AI model and the harness around it split the work. It teaches the general idea only and does not discuss Collison’s post.
The key insight: As we read the post, the change Collison describes is who sends a feature to production: in his set-up a user’s typed request goes to a background agent that implements, tests and deploys it, and one user of one app did that 150 times in a few days.
Collison wrote that with the feature “the activation energy barrier to sanding any rough edge or implementing any new feature becomes extremely low”.
The screenshot attached to his post shows a “Send feedback” box. Its note reads: “A background agent on GCPVM in ~/mgr implements it, tests it and deploys it, then reports back under Feedback & fixes”.
He called the pattern “a new kind of tool — liquid software?” and wrote that “every application that doesn’t have this feature now has a slightly rigid and anachronistic flavor”.
The improvements loop shown in the screenshot attached to Patrick Collison’s post on X, 31 October 2026: a request typed into a Send feedback box goes to a background agent that implements, tests and deploys it, then reports back. Diagram by FourWeekMBA; the 50-a-day figure assumes three days and is our arithmetic.
What Was Posted on X
I’m now building an “improvements” feature into all of my personal vibecoded applications: a freeform text input that’s hooked up to an agent that yolo implements whatever is submitted.
It’s an interesting experience — the activation energy barrier to sanding any rough edge or implementing any new feature becomes extremely low. Someone else who uses one of the applications I created has already implemented 150 improvements over the course of a few days. This pattern feels like a new kind of tool — liquid software — and every application that doesn’t have this feature now has a slightly rigid and anachronistic flavor. Why can’t I just tell it how I want it to be different
Now, it’s not obvious to me how this pattern can or should scale. Does it become unmaintainable after a while (Maybe model intelligence improvement outstrips my feature appetite such that it’s fine?) How does collaborative software work in this model Can software designed this way be secure and reliable
Whatever the larger questions about broader deployment, it definitely works great in a personal context. It reminds me a little of the visions of Lisp/Lisp machines, Smalltalk, and HyperCard — systems that the user can take control of and modify. Those visions never quite worked, though: the vast majority of software today is not end-user modifiable in any useful way. So, while this “liquid” affordance is a very small touch, the way it changes one’s relation to the software feels a bit transgressive and empowering in a Lickliderian and Engelbartian way.
Patrick Collison, chief executive of Stripe, posted this on X on 31 October 2026. He describes an “improvements” text box in his personal apps that sends each request to an agent, and writes that one user implemented 150 improvements over a few days. The piece quotes the post and its screenshot.
The Older Visions He Named
Collison wrote that it reminds him “a little of the visions of Lisp/Lisp machines, Smalltalk, and HyperCard — systems that the user can take control of and modify”.
He added that “Those visions never quite worked, though: the vast majority of software today is not end-user modifiable in any useful way”.
The Questions He Left Open
The post lists open questions rather than answers. Collison asked: “Does it become unmaintainable after a while?”, “How does collaborative software work in this model?” and “Can software designed this way be secure and reliable?”.
He wrote that “it’s not obvious to me how this pattern can or should scale”, and that “it definitely works great in a personal context”.
The Business Lever: Who Decides What Gets Built
In the set-up Collison describes, a user’s request goes to an agent that implements, tests and deploys it, rather than to a product team’s queue. The cost of each change in that loop is the agent’s compute; recent pieces covered how vendors price and cut that cost for coding agents and how OpenAI is moving ChatGPT and Codex onto one shared allowance.
The Structural Read
As we read the screenshot, the loop has no human step between the request and the deploy: the agent implements it, tests it and deploys it, then reports back under Feedback & fixes.
As we read the post, Collison sets the idea against Lisp machines, Smalltalk and HyperCard, which he wrote never quite worked because most software is not end-user modifiable in any useful way.
As we read his questions, the open issues are maintainability, collaboration, security and reliability; he wrote that the pattern definitely works great in a personal context.
Patrick Collison, chief executive of Stripe, on X, 31 October 2026
“Why can’t I just tell it how I want it to be different?”
Three Implications
FOR SOFTWARE MAKERS In Collison’s set-up a user’s request reaches an agent that implements, tests and deploys it, without a product team in between.
FOR AGENT VENDORS Every request in that loop is agent work; one user of one app made 150 such requests in a few days, per the post.
FOR BUYERS OF SOFTWARE Collison asked whether software built this way can be secure and reliable, and how collaborative software works in this model; the post gives no answers.
The framework says “Orchestration, memory, tool use, and workflow execution have become the new sources of competitive advantage.”
As we read it, Collison’s improvements box puts workflow execution in the user’s hands: the request goes straight to the agent. The framework is the owner’s argument and does not discuss Collison’s post.
The apps are Collison’s personal projects, not Stripe products, and the post does not name them. It gives no measure of the quality of the 150 changes, how many days they took or how many people use the apps. We did not see the apps.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Patrick Collison wrote that one user of an app he built implemented 150 improvements in a few days through a text box wired to an agent, and asked whether such “liquid” software can stay maintainable, collaborative, secure and reliable.
A note on sourcing. Figures and quotations come from a post on X by Patrick Collison, chief executive of Stripe, and the screenshot attached to it, read on 31 October 2026. The 50-a-day figure assumes three days and is our arithmetic. We did not contact Mr Collison or Stripe.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
