Uber Cuts AI Costs 52% While 10xing Usage via Agentic Workflow
alvelda · x · 2026-08-31
Uber published a technical deep dive on operating an automated "software factory" where agents now handle 70% of pull requests. Despite usage spiking 10x in six months, the total AI bill remained flat, and per-session costs dropped by 52%.
Key strategies include:
- Orchestrator-Worker Split: Top-tier models handle planning; lightweight models execute specific tasks to maximize leverage.
- Extended Caching: Context cache extended to 1 hour (from 5 mins) with aggressive summarization (400k tokens) to avoid re-processing history.
- Just-in-Time Tooling: Over 1,000 internal MCP tools are served via a gateway and loaded on-demand, preventing 70k token dumps at startup.
- Code-mode over Chat: Models write scripts for SQL/batch jobs and return only results, cutting 90% of token overhead from Q&A loops.
- Knowledge Graph Nav: A 24M-node graph reduces debugging time from 20 mins to 38 seconds.
Related event: Uber Reveals 3,600 AI Agents Now Handle 70% of PRs While Costs Fall 52%(5 posts)→
More from coding & agent
- Claude Code plugin /ambition rewrites your task into the version a more ambitious person would ask for — jakedahn · 2026-10-16
- arXiv paper "Self-Evolving Agent Harness" formalizes harness self-evolution with admission rules — rohanpaul_ai · 2026-10-16
- Self-improving agents should only accept harness edits that beat the current version on fresh tasks — rohanpaul_ai · 2026-10-16
- Claude eng lead's 'be more ambitious' tutorial turned into a /ambition Claude Code plugin — jakedahn · 2026-10-16
- Bland launches iMessage business agents: send a link, get a branded agent in 30s — EXM7777 · 2026-10-16
- Perplexity open-sources SPACE, its Firecracker microVM agent sandbox — perplexity_ai · 2026-10-16