AI engineering moves fast. New tools, frameworks, agents, models, and workflows appear constantly. Some of them are genuinely useful. Others are mostly noise. How can we know if this new thing is just hype or actually has value to it?
For this I’d like to prepose the hype detection framework.
Measure Value
First, how do we measure value in the case of AI engineering? Is it speed? Quality (product, code, etc)? Cost? Sustainability?
Current State
Second, what does our current workflow look like? Where are the bottlenecks? Do not replace a working part of the routine because the adaptation cost will likely outweigh the net gains.
Actively look for better ways to perform the tasks causing friction in the current workflow.
Adaptation
Finally, adaptation is never free. Incremental change is the safest bet.