www.seeingthesystem.com/p/buying-back-our-slack

This sentence says Coinbase did nothing to help engineers adapt to AI tools. Contemporary reporting on the same episode contradicts that.

TechCrunch reported that Coinbase bought enterprise licenses for GitHub Copilot and Cursor for every engineer. It also quoted Armstrong telling engineers they did not have to use the tools every day yet "until we do some training," which indicates training support was part of the rollout.

Fortune likewise reported that, beyond the mandate, Coinbase hosts monthly "AI speedruns" where employees who are using AI effectively teach the rest of the company. That is a concrete institutional support mechanism for adoption.

Those facts do not make Coinbase's approach gentle or adequate, but they do mean the claim that Coinbase "did nothing" is factually overstated.

This claim is historically inaccurate. Public sources show the software role title "product engineer" existed years before Perplexity.

Asana published a company engineering post in April 2016 titled "Product Engineering: Reimagining role boundaries at Asana." In it, Asana explicitly says: "At Asana, we have product engineers, who, as the name describes, do both" product and engineering work across the product lifecycle.

A second Asana post from October 2019 describes an employee who "started as a Product Engineer" at the company, again showing the role title was already established well before Perplexity.

Perplexity may be a prominent recent example of the role, but it did not originate it.

This sentence overstates and misstates what the Whitehall studies found.

The first Whitehall study reported that employment grade (a measure of occupational/social position), not job control, was associated with persistent differences in total mortality and many specific causes of death over 25 years.

A later Whitehall II analysis did find that low control at work mattered, but the Lancaster summary of that research is much narrower: "among the factors examined, low control at work made the largest contribution to the socio-economic gradient in coronary heart disease frequency." That is not the same as saying job control is the single strongest predictor of mortality overall.

So the article collapses two different findings into one stronger claim than the Whitehall literature supports: (1) a mortality gradient by employment grade/social position, and (2) low job control as one contributor to a specific coronary-heart-disease gradient.

This sentence overstates what happened at Coinbase.

Both TechCrunch and Fortune reported that Armstrong said some of the engineers who had not onboarded by Saturday had good reasons—for example, being away on a trip—and were not fired. Armstrong said only some of the people without a good reason were fired.

So the article's wording, "fired everyone who had not complied," is incorrect. The reporting describes a smaller subset, not every non-compliant employee.