I was going to write about something else this week, but then on Thursday, this happened:
this is not the actual tweet Jack sent, but it mi
So instead, I’m writing about AI again. At this point the ground has been well-tread, but if you want a primer on all the problems (especially if you think, but AI is useful to me personally!) I strongly recommend Eryk’s piece here:
What remains urgently in dispute are the boundaries of utility: what usefulness means, for whom, and under what conditions? At what cost and from whom are benefits derived, and how are benefits and risks distributed?
The company Jack Dorsey is talking about is Block, formerly known as Square. And while people paying attention to the company’s structure (operating three redundant orgs) and lack of leadership (Block’s stock price is down 70% from its peak), internet commentators latched onto Dorsey’s claim that the layoffs were due to AI-driven efficiency.
The commentariat rolled out the same claims as always: of course AI tools make you more productive, so if you don’t learn AI skills, you will be replaced by people with AI skills. However, it is difficult to reconcile these claims with reality: among the victims of the layoff were many high-performing AI enthusiasts.
Not that the connection between AI and productivity is exactly well-established. Keeping the APIs functional is a part of the job LLMs still can’t do, and reliance on AI agents has caused some high-profile outages lately. The successes, meanwhile, seem to be illusory: Cursor’s claim about vibe-coding a browser falls apart under scrutiny, and the Anthropic-funded study claiming that AI replaced 11% of workers was actually referring to simulated tasks (in other words, Claude marked its own homework). Accenture has been reduced to logging employee hours in AI tools, because the productivity of people who use it is otherwise indistinguishable from people who don’t.
Seniority makes you a target
A common claim from LLM proponents has been that only senior, experienced users get these benefits, because you need a high level of competence to know where to use the AI and how to check its work. Early-career workers are the ones most likely to have their skill formation negatively impacted by AI (this thread by Jennifer Moore dives deep into the mechanics of this).
So you’d expect Block to ditch these low-performing juniors and retain only high powered 10x developers who can harness the power of AI, right?
Take a guess. Then read Jürgen’s article, and guess again.
Those affected in Block’s layoff came from the ranks of senior developers and PMs. Because layoffs are always a cost control measure. And in a world where leadership sees every role as fungible anyway, cheap junior devs banging away on ChatGPT are just as good as the expensive people. Maybe better, as models don’t handle complex tasks nearly as well as churning out boilerplate code for less-demanding users:
Developers have experienced much higher error rates — more "hallucinations" — when models are prompted to solve more complex problems.
We see this effect manifest itself in a benchmark test of how good various models are at making web UIs accessible: in multiple cases, expert guidance actually produces a greater number of defects, because the model simply cannot comply with what is being asked of it.
So getting rid of the pesky experts who get paid more to slow down the process with unnecessary things like “checking if the code does what it’s supposed to” is a win-win. In a race to the bottom where the value is solely blasting through to the next set of deliverables, your skill and craft and taste are maladaptive.
Cultivate skills to bounce back from a layoff
But it's no picnic for juniors either. The rate of hiring early-career workers was slowing long before ChatGPT stumbled into the spotlight, due to a confluence of related crises caused by chasing short-term wins. Hedgpeth’s advice (to take ownership of your career development) may be the reason that seniors bounce back at a faster rate: they can fall back on the presentation skills they’ve honed over the years, the networks they have nurtured, and the unique intersections of interests that make them stand out in a crowded market.
The best time to give this some thought is before it becomes existential. Despite (or perhaps because of) the AI hype cycle entering its disillusionment stage, the layoffs show no sign of stopping. But after those layoffs come waves of rehiring, as companies experience AI buyer’s remorse.
And when that happens, the people filling the roles won’t be the ones who became the most reliant on AI tools. The people who will stand out will be the ones who learned how to write.
— Pavel at the Product Picnic