DigiNews

Tech Watch by Johan Denoyer

← Back to articles

Does Speaking to Agents Like Cavemen Really Save 65% of Tokens? We Test

Quality: 8/10 Relevance: 9/10

Summary

JetBrains tests a Caveman prompt on Claude Code to measure token savings. The results show ~8.5% output-token savings with no measurable quality degradation, debunking the 65% claim; cost savings are around ~10% in ideal cases, but vary with task type.

🚀 Service construit par Johan Denoyer