Search
4 results for “tysongern”
-
Two systems with 90% accuracy in series give 81%. Three give 72.9%. Chains of LLM calls multiply error without adding accuracy.
A deterministic check or a human between calls breaks the chain.
-
The real tell for AI-generated writing is in the citations, not the tone. LLMs cite references to convey authority, not accuracy, often pulling in papers that have nothing to do with their topic and misinterpreting key results.
If the citations check out, a human probably wrote it. #AI #Writing #Citations #LLM
-
For decades we've known that proper code review increases a team's comprehension of a codebase. Fagan showed in 1976 that structured reading was the highest-yield defect removal activity. The increased output of coding agents makes comprehension more of a bottleneck, and raises the stakes for review. #CodeReview #CodeQuality #SoftwareEngineering
https://www.ida.liu.se/~TDDC90/literature/lab-papers/fagan76.pdf
-
Many developers use CSS frameworks to reduce boilerplate, increase quality & drive consistency. Good in theory - not always in practice.
Instead of using a #CSSframework, you should write your own custom #CSS. This is the best option!
Dive into Tyson Gern's article on #InfoQ to learn more: https://bit.ly/4aYEELL