AI Helps You Code, But Can't Replace the Human Who Understands the Problem

What it is

In an essay, Arvind Narayanan and Sayash Kappor argue that there is enough evidence to reject the narrative that once AI capabilities reach a certain threshold, it will cause mass layoffs - even in software engineering, a field uniquely suited to AI disruption. They identify three real bottlenecks that AI hasn't automated: deciding and specifying what to build, verifying and being accountable for what is delivered, and the deep human understanding of the codebase, the business, and the environment required to do both.

Why it matters

The data still doesn't support the idea that AI is causing mass unemployment. When New York added an AI disclosure checkbox to WARN Act filings, more than 160 companies filed notices in the first year - and not a single one checked the AI box. If AI-driven layoffs aren't showing up even in software engineering, a field with very few regulatory barriers, Narayanan and Kappor argue that most other professions are likely to be even more cushioned. Understanding where the real bottlenecks lie helps you use AI smarter: lean on it for drafting and generating, but don't hand over the thinking.

How to use it

1. 1Step 1: Identify which part of your task is just 'typing it out' - writing a draft, filling in a form, summarizing notes. That's where AI can genuinely save you time.
2. 2Step 2: Identify which part requires real understanding - knowing your audience, your organization's quirks, what the goal actually is. That part stays with you.
3. 3Step 3: Use AI to handle the first part quickly, then apply your own judgment to review, correct, and take responsibility for the result.
4. 4Step 4: Notice where AI output feels 'off' or generic. That's usually where deep human context was missing - your job is to supply it.

Example

Simon Willison, who linked to the essay, reflected honestly: 'I'm finding AI assistance also helps me with the deciding and verifying steps, but it's the deep human understanding that remains key to the value I provide. Give me all of the AI assistance in the world and the value I produce will still be reliant on how deeply I understand both the problems and the solutions.' Even someone who uses AI constantly finds the irreplaceable part is what he already knows - not what the AI generates.