did some general polish and cleanup

made some small fixes

refactored the throughput optimization visualization

made throughput visualization nicer

added some learnings to agents.md

did a big restructuring of the infrastructure section and added major new results of the vllm inference benchmark

use figref throughout the blog post

add component for automatic figure number generation in references

add first version of throughput optimization visualization

add vllm inference benchmark report and some new rules to agents.md

added skill to take screenshots for better debugging with agents

Add auto-numbering for figures

add a visualization of the infrastructure using a mermaid diagram

improve the writing, mainly of the setup

made some writing improvements

make plots look nicer

add summarize results

improve coloring

move transition sentences after the charts

prevent glossary to cause line breaks

remove some repetitions and reorder content

added links to prompts in the appendix

moved the plot more to the right as to not cut off the display names

move main experimental plot to the introduction as a teaser and reformulate experiments for nicer flow

support different setups so that we can reduce the number of charts a bit

Made dataset naming more consistent throughout

make baselines striped in bar chart and dashed in line chart

removed naming section since I have better display names now

update agents.md file with guidelines for bibliography

add more citations

Fix title

display source datasets as accordions

added more references

changed links to references, removed old references and cleaned up todos

fix footer to merge references and footnotes from all chapters into one list

removed redundant defaultView and added dclm and fw-edu-hq baselines everywhere with stable colors for less mental load

wrap prompts as text blocks

wrap lines in plain text blocks for readability

replace static plots with dynamic d3 embeds

switched to just one csv file with the complete results

make view switch simpler

made plot d3 component more general and added finephrase vs baseline comparison

replaced static baselines plot with dynamic d3 plot

remove old notion import

fix title hierarchy

fix setup issue

import content from notion

vibe-wrote initial AGENTS.md file for better writing with LLMs

update

update