Why Developers Should Learn to Love Regulation

Grade One — Wed, 02 Sep 2026 17:27:06 +0000

Tech culture is allergic to regulation. We want to move fast and break things. Auditors are roadblocks to shipping code. Compliance is seen as a tax to be optimized away. But there is a counter intuitive truth about the future of work that has become starkly apparent when analyzing the job market since 2024. Regulation is the only thing standing between you and automation.

I spent several months mapping out 46 computing careers using grade-1 data sources (regulators, central banks, verified payroll) in an effort to see which jobs were disappearing and which ones were safe.

It had nothing to do with how clever your code was, but rather how many of the four barriers described below were part of the job description.

The Four Barriers to Automation

If you are a frontend developer who builds a website for a marketing firm, your work is highly exposed. When you make a mistake and break the site, the stakes are low. Somebody can just open up a ticket and have it fixed in an hour. If you're an avionics software engineer who writes code that runs an aircraft, you know that every single requirement in your design document has a test associated with it in the DO-178C standard. If your code fails, people are going to die.

A generative model could write both sets of code, but it can't accept liability if the plane crashes.

These are the institutional barriers to AI substitutions, and as a heuristic, if you're not employing at least two of these mechanisms in your workflow, you are likely to see a quick uptick in automation. Here's how each one works:

Liability

In this context, it means somebody has to sign off on the work and accept responsibility for errors. Even if a company wanted to, they can't put an algorithm in jail. If a mistake would require a large fine or lawsuit, there has to be a named individual that is allowed to absorb that liability.

Regulation

This means somebody has to approve changes before they go into production. Regulators work at the speed of bureaucracy. If there's a federal agency that requires a certain testing and certification process (medical devices, IEC 62304), the company has to endure the cost and delay. They can't short circuit the process just to save time and money.

Restricted Data

Some data can't be shared with a third party. HIPAA covered entities, national defense, and proprietary algorithms at hedge funds have to assume that if the model can't see it, it can't comprehend it.

Closed Systems

Legacy systems (mainframe banking applications) or private platforms (e.g. proprietary trading algorithms at hedge funds) aren't available for the model to observe.

The Human as a Liability Absorber

It is only now, as companies begin to understand the risks, that courts begin to correct the loophole.

In late 2023, the UnitedHealth Group was attacked with a federal class-action lawsuit (Estate of Lokken v. UnitedHealth Group) after an algorithmic system called "nH Predict" began automating decisions to discharge post-acute care patients covered by Medicare Advantage. The case was ordered unsealed last month after a federal judge determined that the contract between the parties assumed that human clinical staff would make the decisions, not an algorithm.

Last year, a Canadian tribunal ruled (Moffatt v. Air Canada) in favor of a passenger who was misled by a customer service chatbot that claimed Air Canada had a bereavement fare policy. Air Canada argued the chatbot was an independent legal entity, but the court found that the corporation was liable for the algorithm's actions as if it were an employee.

These cases reveal the liability risk posed by algorithmic systems, and how regulation fills that risk by punishing the institution that deploys the technology when it goes astray. This encourages organizations to retain humans as liability absorbers. The human is placed in the seat as the weak link in the chain in order to appease the regulator.

The Strategy

Rather than competing on code creation, which is obviously going to be automated, try to acquire context that can't be downloaded. Instead of seeking to work on a more interesting stack, consider what industries have the best compliance mechanisms and learn those rules. The regulatory moats that seemed like annoying obstacles last year are exactly what will protect you this year.

A job is a set of barriers to entry, not a salary.

This framework was reverse-engineered from a 561-page book that mapped out 46 computing careers using grade-1 data sources. The research methodology can be found at gradeone.press.

DEV Community: Grade One

Why Developers Should Learn to Love Regulation

The Four Barriers to Automation

The Human as a Liability Absorber

The Strategy