---
title: "mayaandersson (@mayaanderssondev) | Hashnode"
description: "Just a bored curious dev. Read the latest articles by mayaandersson on Hashnode."
url: "https://hashnode.com/@mayaanderssondev"
---

@mayaanderssondev

# mayaandersson

@mayaanderssondev · Palo Alto CA · Joined May 2026

Just a bored curious dev

### [Dividing your RAG score by retrieval recall overstates your generation quality, and here is by how much](https://llmasajudge.hashnode.dev/dividing-your-rag-score-by-retrieval-recall-overstates-your-generation-quality-and-here-is-by-how-much?utm_source=hashnode&utm_medium=feed)

Almost every RAG eval writeup I read, including several I have recommended, uses the same mental model. The system is two stages in series. Retrieval either finds the right context or it does not, wit

12 min read · [llmasajudge.hashnode.dev](https://llmasajudge.hashnode.dev/dividing-your-rag-score-by-retrieval-recall-overstates-your-generation-quality-and-here-is-by-how-much?utm_source=hashnode&utm_medium=feed)

[Aug 25](https://hashnode.com/posts/dividing-your-rag-score-by-retrieval-recall-overstates-your-generation-quality-and-here-is-by-how-much/6a8dfffe3a681f3fb91ed2e6)

### [Six prompt-optimization frameworks: what matters when you run them on the same task](https://llmasajudge.hashnode.dev/six-prompt-optimization-frameworks-what-matters-when-you-run-them-on-the-same-task?utm_source=hashnode&utm_medium=feed)

"Prompt optimization" spans random search to evolutionary Pareto fronts.

4 min read · [llmasajudge.hashnode.dev](https://llmasajudge.hashnode.dev/six-prompt-optimization-frameworks-what-matters-when-you-run-them-on-the-same-task?utm_source=hashnode&utm_medium=feed)

[Aug 18](https://hashnode.com/posts/six-prompt-optimization-frameworks-what-matters-when-you-run-them-on-the-same-task/6a84a549730583a4edc47101)

### [A judge that agrees with your humans 92 percent of the time can be at 60 percent where the gate actually decides](https://llmasajudge.hashnode.dev/a-judge-that-agrees-with-your-humans-92-percent-of-the-time-can-be-at-60-percent-where-the-gate-actually-decides?utm_source=hashnode&utm_medium=feed)

TL;DR: Judge-human agreement is almost always reported as one number over a whole validation set. That number is dominated by the easy cases, because most examples are not close to your decision bound

9 min read · [llmasajudge.hashnode.dev](https://llmasajudge.hashnode.dev/a-judge-that-agrees-with-your-humans-92-percent-of-the-time-can-be-at-60-percent-where-the-gate-actually-decides?utm_source=hashnode&utm_medium=feed)

[Aug 17](https://hashnode.com/posts/a-judge-that-agrees-with-your-humans-92-percent-of-the-time-can-be-at-60-percent-where-the-gate-actually-decides/6a83477a7b7f90bd9fc771c4)

### [Run forty experiments against one eval set and you will find an improvement that is not there](https://llmasajudge.hashnode.dev/run-forty-experiments-against-one-eval-set-and-you-will-find-an-improvement-that-is-not-there?utm_source=hashnode&utm_medium=feed)

You iterate on a prompt, run the eval set, keep the best variant. Repeat for a quarter. The number that survives is the maximum of forty noisy measurements, and the maximum of noisy measurements is an

5 min read · [llmasajudge.hashnode.dev](https://llmasajudge.hashnode.dev/run-forty-experiments-against-one-eval-set-and-you-will-find-an-improvement-that-is-not-there?utm_source=hashnode&utm_medium=feed)

[Aug 12](https://hashnode.com/posts/run-forty-experiments-against-one-eval-set-and-you-will-find-an-improvement-that-is-not-there/6a7c7e4cae9b7572d6527799)

### [I read the metric libraries of five widely-used eval tools. The metric was never the hard part.](https://llmasajudge.hashnode.dev/i-read-the-metric-libraries-of-five-widely-used-eval-tools-the-metric-was-never-the-hard-part?utm_source=hashnode&utm_medium=feed)

Every LLM eval tool sells you the same headline: a big bag of ready-made metrics. Fifty of them. Seventy. Pick one, call evaluate(), get a number. The pitch works because it is true, and because it qu

9 min read · [llmasajudge.hashnode.dev](https://llmasajudge.hashnode.dev/i-read-the-metric-libraries-of-five-widely-used-eval-tools-the-metric-was-never-the-hard-part?utm_source=hashnode&utm_medium=feed)

[Aug 11](https://hashnode.com/posts/i-read-the-metric-libraries-of-five-widely-used-eval-tools-the-metric-was-never-the-hard-part/6a7b6feada84d66b0fbb7f1b)
