In a post dated 3rd October 2026, developer Simon Willison argued that pay-by-usage services and APIs should offer 'default hard budget caps'—limits that return errors and cut a service off after a set monthly spend, rather than 'soft caps' that only send a warning email. Willison wrote that coding agents and personal agents reduce the friction of spinning up code that can incur costs through paid APIs, hosted applications, or billed storage and compute, and that users risk waking up to a surprise bill of several hundred or several thousand dollars. He acknowledged a counterargument—that businesses may not want applications throwing errors when a budget is exceeded—but wrote that most would prefer errors to a surprise $10,000+ bill. He proposed hard caps as the default, with an opt-out checkbox for those willing to accept subsequent charges. Willison noted the service he most wants this from is AWS, and reported that AWS launched spending limits that pause a project when it reaches its spend limit, per a 16th September announcement, though a related settings page states the experience is releasing to a limited number of customers. He also noted Google Cloud launched a similar 'Spend Caps' feature in July.
- Willison distinguishes 'hard caps' (cut off, return errors) from 'soft caps' (warning email only)
- He proposes hard caps as default, with an opt-out checkbox for continued charges
- AWS announced spending limits that pause a project on 16th September; still releasing to limited customers
- Google Cloud launched 'Spend Caps' in July
What it means for you
If you use cloud services (like AWS or Google Cloud) or any pay-per-use AI API, you can rack up large bills fast—especially now that AI coding tools make it easy to deploy things that quietly cost money. A 'hard' spending limit shuts the service off when you hit a dollar amount, instead of just emailing you after the damage is done. Both AWS and Google Cloud have recently added versions of this.
Try this
Log into any cloud or API account you use for projects and set a monthly spend limit now—Google Cloud has 'Spend Caps' and AWS is rolling out spending limits. Pick a dollar figure you'd be comfortable losing in a worst case.
Who should care
Anyone running personal projects, test apps, or small-business workloads on usage-billed platforms like AWS, Google Cloud, or metered AI APIs—especially if you're using AI coding assistants to deploy things you don't fully monitor.
Skip this if
You only use flat-rate, fixed-price AI tools (like a monthly ChatGPT subscription) and never touch pay-per-use cloud services or APIs.
Sources: Simon Willison — read the original