Revolut cut its AI coding bill by 90%: 256K context enough, cache 90% of tokens
alvelda · x · 2026-11-10
David Abitbol, who runs AI for 2,000 engineers at Revolut, calls long context "a scam" and cut the company's AI coding-agent spend by 90% in under a year, with coding time up 60% and lead time down 70%. Key lessons: 256K context is all engineers need; over 90% of coding tokens should come from cache; running models at 4-bit instead of 8-bit saves 40% GPU memory with no quality loss; the same 56 GPUs went from 10+ seconds to 1.6s first-token latency while demand grew 76%; his advice is to resist adding hardware. Some engineers previously spent $1,000–$10,000/month; Revolut now pays $0.30 per million tokens, roughly a tenth of Claude Code or Codex pricing. The same week, Goldman Sachs pulled Claude Code from traders and Apple cut some engineers' daily AI budgets. The linked paid guide walks through spend audits, routing tasks to the cheapest capable model, flat-vs-token plan math, and 5 Claude skills to automate the savings.
More from coding & agent
- NODES session: Test-time strategies like confidence decoding and schema filtering for reliable Text2Cypher — JeremyCMorgan · 2026-11-10
- Stacking More MCP Servers Made My Agent Worse, Reddit User Says — Glittering-Bus-8500 · 2026-11-10
- Anthropic engineer's AI-built browser FPS nears COD quality, sparking 'one year to Sea of Thieves' claim — ChrisGPT · 2026-11-10
- Tech debt in the AI era: paid in full whenever the next frontier model ships — doodlestein · 2026-11-10
- Claude-coded Valheim mod rebuilds Ocarina of Time's Kokiri Forest from your own ROM — Any-Explanation3814 · 2026-11-10
- App Store Connect CLI 6.5 adds subscription retention messages and offers — rudrank · 2026-11-10