EXPERIMENT LOG

Open Lab

Hypotheses, trials, and results you can replicate. Focus on process, not perfect showcases.

2024-12-05
Validated

Edge caching reduces LCP by 40% for personalized static pages

Before
2.9s
After
1.7s
Delta
-41%
LEARNINGS
  • •Geo-segmented caching yields better P95 than fixed TTL
  • •Silent revalidation avoids cold starts during campaigns
a1c4f2db7e91c3
Open in Stackblitz
15MIN
2025-01-12
Running

Circuit breakers reduce errors in bursts > 2k RPM

Before
0.68% errors
After
0.18% errors
Delta
-73%
LEARNINGS
  • •Fallback to internal queue keeps UX acceptable without impacting billing
  • •A saturation dashboard reduces MTTR more than extra logs
c9e3ad4
View PoC
5MIN
2025-02-02
Validated

GLSL underlay animations don't impact CLS or TTI

Before
0.12CLS
After
0.03CLS
Delta
-75%
LEARNINGS
  • •Keeping the canvas as underlay with pointer-events-none removes jank
  • •A 2 draw-call budget keeps the main thread free
d1f2aa9e3bcf78f42d0af
View code
30MIN