Why giants like Ramp and Spotify build their own AI infrastructure
bendee983 · x · 2026-08-26
This post analyzes why major companies like Spotify and Ramp choose to build their own coding agents and AI infrastructure instead of buying off-the-shelf tools. The key driver is scale: at their size, every optimization can save millions in token spend. This justifies developing custom harnesses, replacing closed models with open ones, and training smart routers. For smaller teams, however, the convenience of buying still outweighs the cost savings of building from scratch.
More from coding & agent
- Built a caching layer for MCP to stop agents asking the same question — Ambitious_Bake_6232 · 2026-08-26
- micro1 hosts coding agent hackathon with $5,000 prize pool — AromaticFood83 · 2026-08-26
- Optimization tip: Use delegation.model to cut costs — alexcovo_eth · 2026-08-26
- Vibe-coding a demo isn't enough for real field robots — ZeYanjie · 2026-08-26
- Open-Source Agentic Trading Lab: Prototype, Backtest, and Inspect LLM Trading Agents — tom_doerr · 2026-08-26
- ClawProBench: Trace-Aware Agent Evaluation with Runtime Coverage and Frozen Tasks — YuanHang Xiao · 2026-08-26