- Draft Title: Death by a Thousand Tokens: How to Cut Your AI Coding Bill in Half (Without Cutting Corners)
- Length: 15 min
- Links: (slides, article, notes — TBD)
Every AI coding session quietly burns money on things that don't make your code better: bloated context windows, redundant re-reads, verbose tool output, over-eager subagents, and frontier models running on lightweight tasks. This talk walks through concrete, practical techniques — prompt/context hygiene, output filtering, model tiering, caching, and workflow design — that cut real AI development costs 60-90% with no loss in output quality. You'll leave with a checklist you can apply to your own AI-assisted workflow today.
Speaker Bio
- Name: Shawn McKay
- Mini-bio: Shawn is a Sr Staff Engineer at Amplitude leading the DevX team. Recently he started working on Frugl.dev, an automated AI session analyzer to provide feedback on how to reduce token waste.
- Picture for slides: I think we already have one
- Would you like help with your presentation? Sure!
- Social media link(s): https://www.linkedin.com/in/shmck/
- Do you agree to the VanJS Code of Conduct? YES
Every AI coding session quietly burns money on things that don't make your code better: bloated context windows, redundant re-reads, verbose tool output, over-eager subagents, and frontier models running on lightweight tasks. This talk walks through concrete, practical techniques — prompt/context hygiene, output filtering, model tiering, caching, and workflow design — that cut real AI development costs 60-90% with no loss in output quality. You'll leave with a checklist you can apply to your own AI-assisted workflow today.
Speaker Bio