Loading video...

Video Failed to Load

Go Home

Just launched an autoresearch loop for speeding up your frontend. I spent ~$10k in fable api credits speeding up the top 200 github repos, took the learnings and tracked the most common wins. The loop iterates through them to find speedups.

54,527 views • 10 days ago •via X (Twitter)

15 Comments

Jacob Miller's profile picture
Jacob Miller10 days ago

Oh also, it ties into your current claude/codex/cursor subs, so it's free to use. Give it a shot: npx makefaster

Varik Verilion's profile picture
Varik Verilion9 days ago

The interesting part is turning a pile of one-off frontend wins into a repeatable search loop. The benchmark matters too. Speedups that survive across 200 repos are a much better signal than a polished demo.

Jacob Miller's profile picture
Jacob Miller10 days ago

Also I made a Show HN for this:

Jared Smith's profile picture
Jared Smith10 days ago

@sheherenow_ Legend!!! Ty for your service sir

Jacob Miller's profile picture
Jacob Miller10 days ago

@sheherenow_ Only a legend if it works - give it a shot and report back 🙏

Terry Carson's profile picture
Terry Carson9 days ago

Great info to know.

Ezra Mechaber's profile picture
Ezra Mechaber10 days ago

Love this idea. We thank you for your credit burn. Is it looped over Core Web Vitals?

Jacob Miller's profile picture
Jacob Miller10 days ago

Lighthouse score was the metric I was using. Core web vitals are user-submitted scores, which can't really be automated easily. They measure roughly the same thing though - LCP/TTI/etc, so this should have a measurable impact on your CWV score.

Ezra Mechaber's profile picture
Ezra Mechaber10 days ago

Ohhh, I'm not sure I realized that CWV was only live users.

jem 💜🩵🩷's profile picture
jem 💜🩵🩷10 days ago

he’s at it again 😮‍💨

Jacob Miller's profile picture
Jacob Miller10 days ago

I'm just out here makin' shit I want to use :)

Camaleón Raro's profile picture
Camaleón Raro10 days ago

THANK u!! I will use this!!!

Jarno's profile picture
Jarno9 days ago

The real product here might be that common wins dataset you built along the way, not the tool.

Jacob Miller's profile picture
Jacob Miller9 days ago

It 100% is. It's why I also let you optionally submit your own stats/thinking traces - ideally building up the db over time will continue to make this better.

arin's profile picture
arin9 days ago

$10k testing real repos is a much better moat than another performance checklist

Related Videos