Meta spent 20 years building one of Silicon Valley's most respected engineering organizations. It ran on two sequential cultures: 'move fast and break things' through the 2010s, then 'move fast with stable infra' into the early 2020s. Engineers were treated as a profit center, empowered to own impact, and trusted to balance business needs with technical quality. That changed in April 2025.
In a matter of weeks, Meta leadership began dismantling that foundation. Engineers are being forcibly reassigned to AI data labeling work as a precursor to layoffs. The company suffered what sources describe as its most embarrassing outage ever. Internal processes have fractured. Software engineering has been reclassified, in practice, from profit center to cost center. The original piece documents each of these inflection points with specificity, including the internal signals that preceded each failure.
Gergely Orosz frames this as a case study in what he calls 'AI psychosis': the organizational dysfunction that sets in when leadership treats AI adoption as a culture replacement rather than a capability addition. The question the full piece answers is not just what happened at Meta, but whether this pattern is spreading to other companies. The section on self-inflicted wounds is where the real argument lives.
[READ ORIGINAL →]