How AI is changing Software Engineering: A Conversation with Gergely Orosz, @pragmaticengineer

Gergely Orosz argues token maxing is big tech’s new lines-of-code antipattern, and that AI is shrinking team sizes faster than it’s improving team output.

• Salesforce mandates a minimum $175/month AI token spend per engineer; Meta had a public leaderboard that drove engineers to run junk agents just to inflate counts.
• Meta’s performance reviews use token count as one data point alongside diffs and code reviews — low count on a low performer is actively used against them.
• The Meter study found engineers felt 20% more productive with AI but demonstrated 20% less productivity; sample was only ~30 people with one strong outlier.
• Shopify got GitHub Copilot a full year before public release by offering 3,000 engineers as beta testers; Farhan Thawar says early churn is worth staying months ahead of competitors.
• Large companies (Uber, Airbnb, Meta, Intercom) are building custom internal AI infra — MCP gateways, monorepo-integrated coding agents, AI-categorized code review — rather than shipping AI user features.
• John Deere’s two-pizza teams have shrunk to one-pizza teams; the software engineer role is absorbing QA, DevOps, and increasingly product management responsibilities.
• The Pragmatic Engineer hit 1,000 paying subscribers at $100/year within six weeks of launch, before publishing a single article, by posting one confident tweet.

2026-04-21 · Watch on YouTube