AI coding agents have made reverse engineering cheap: an observation from Simon Willison

Simon Willison, a developer and co-creator of the Django web framework, published a post on July 20, 2026 arguing that AI coding agents have radically lowered the cost of reverse engineering home devices: tasks that never paid off before because of the effort involved are now taken on in passing.

What has changed in the economics of code

Coding agents have collapsed the cost of two things at once: writing a one-off automation and failing at the attempt to write it. According to Willison, what used to stop people was the price of the effort, not technical impossibility. Cracking open a smart bulb, plug, or robot vacuum and wiring it into your own workflow was always feasible. The only question was whether the result justified the time spent.

Willison points to a stream of stories from people who are using agents to reverse engineer and automate the devices in their homes. He sees this as a clear illustration of how cheaper code changes the very framing of the decision: the question shifts from "can I do this?" to "is it worth it?"

Why didn't this pay off before?

Reverse engineering home devices was possible before agents. The obstacle was ROI. Any experienced programmer knows that undocumented, unstable APIs break easily and change in future versions. Spending a weekend on a custom integration meant signing up for an endless maintenance cycle: the manufacturer ships a firmware update, and everything stops working again.

It was this future maintenance burden, rather than the initial difficulty, that most often killed the idea at the start. When fixing code is expensive and writing it takes real effort, the rational choice is simply not to begin.

"Coding agents completely change this equation,"

Simon Willison writes on his blog.

Code as a consumable

The main shift is in treating code as a consumable rather than an asset to be protected. Now that code is cheap, the price of failure has dropped too: trying and not succeeding stings less, because rewriting everything from scratch costs little.

The psychological weight of future maintenance fades along with it. In Willison's words, the prospect of maintaining a piece of work for years, or throwing it away and starting over, stops being an argument against taking on the task.

"The thought of having to maintain the code in the future, or throw it away and start again, carries far less psychological weight,"

Simon Willison notes.

What this means

Cheap code changes not so much the speed of development as the list of tasks people take on at all. The "is it worth it?" threshold moves down: a large number of small automations that were once unprofitable, from smart devices at home to personal scripts, move from "someday" to "I'll do it tonight." Reverse engineering stops being the preserve of enthusiasts with plenty of spare time and becomes an ordinary tool.