Meta Drops AI Usage Metrics from Engineer Reviews Amid Gaming, Pushes Agent Tool Hatch

Per The Information and WIRED, Meta this week told employees internally that AI token usage metrics will be removed from engineer performance reviews, effectively ending the informal 'tokenmaxxing' incentive scheme. The direct trigger was employees gaming the usage metrics. Meta stressed this does not mean encouraging less AI use — the company still claims 93% of code changes are AI-generated — and is now pushing its internal agent tool Hatch.

Confirmed

• Meta notified employees that performance reviews will no longer depend on AI tool usage (WIRED, relayed by @Hesamation and @nordicinst)
• The metric was removed because employees began gaming it (The Information, relayed by @rohanpaulai)
• About a year ago Meta introduced 'AI-driven impact' scoring, with high-usage employees earning labels like 'AI Native' and 'AI First'
• Around twenty-plus employees have sued Meta over the scheme
• Meta is now promoting its internal agent tool Hatch

Why it matters

The episode shows that single-metric AI adoption KPIs are easily gamed once tied to performance. Meta's pivot also signals a shift in enterprise AI rollout strategy: from mandating quantified usage to driving organic adoption via better agent tooling. The related lawsuits highlight the labor-dispute risks of aggressive AI KPIs.

2026-09-03 ~ 2026-09-03 · 5 related posts

• Meta Ends AI Usage Quotas in Performance Reviews While Rolling Out Agent Tool Hatch — nordicinst · 2026-09-03
• [source] Meta drops AI token counts from engineer reviews after employees game usage metrics — rohanpaul_ai · 2026-09-03

3 near-duplicate retellings: rohanpaul_ai · Hesamation · Hesamation