Syed Aslam
@syedaslam.bsky.social
Designing software systems and thinking about why most of them hurt.
Rails, JavaScript, architecture, product.
Rails, JavaScript, architecture, product.
With AMS, performance meant complex caching. With Alba, baseline speed let us delete most of it.
Less caching. Fewer bugs. Simpler code.
If you’re still on AMS because “it works,” benchmark Alba on your heaviest model.
syedaslam.com/posts/escapi...
Less caching. Fewer bugs. Simpler code.
If you’re still on AMS because “it works,” benchmark Alba on your heaviest model.
syedaslam.com/posts/escapi...
Escaping the AMS Trap: How We Benchmarked Our Way to Alba | Syed Aslam
A case study on migrating from ActiveModel::Serializers to Alba, including real-world benchmarks and a custom caching solution.
syedaslam.com
January 20, 2026 at 1:13 PM
With AMS, performance meant complex caching. With Alba, baseline speed let us delete most of it.
Less caching. Fewer bugs. Simpler code.
If you’re still on AMS because “it works,” benchmark Alba on your heaviest model.
syedaslam.com/posts/escapi...
Less caching. Fewer bugs. Simpler code.
If you’re still on AMS because “it works,” benchmark Alba on your heaviest model.
syedaslam.com/posts/escapi...
Results:
• AMS with caching: ~3,500 ops/sec
• Alba (Oj backend, no caching): ~2,800 ops/sec
→ ~4-5x faster than uncached AMS
This changed our thinking.
• AMS with caching: ~3,500 ops/sec
• Alba (Oj backend, no caching): ~2,800 ops/sec
→ ~4-5x faster than uncached AMS
This changed our thinking.
January 20, 2026 at 1:13 PM
Results:
• AMS with caching: ~3,500 ops/sec
• Alba (Oj backend, no caching): ~2,800 ops/sec
→ ~4-5x faster than uncached AMS
This changed our thinking.
• AMS with caching: ~3,500 ops/sec
• Alba (Oj backend, no caching): ~2,800 ops/sec
→ ~4-5x faster than uncached AMS
This changed our thinking.
When we wanted to migrate from AMS to Alba, we found that the process was not as straightforward as we had hoped.
We benchmarked our heaviest real endpoint: an e-commerce order with ~100 shipments, ~100 line items, and deep associations.
We benchmarked our heaviest real endpoint: an e-commerce order with ~100 shipments, ~100 line items, and deep associations.
January 20, 2026 at 1:13 PM
When we wanted to migrate from AMS to Alba, we found that the process was not as straightforward as we had hoped.
We benchmarked our heaviest real endpoint: an e-commerce order with ~100 shipments, ~100 line items, and deep associations.
We benchmarked our heaviest real endpoint: an e-commerce order with ~100 shipments, ~100 line items, and deep associations.
AyatFlow does one specific thing, correct by default. syedaslam.com/posts/constr...
Constraint as a Feature: Designing an “Anti-Canvas” for Sacred Text | Syed Aslam
Why I built AyatFlow as an 'Anti-Canvas' to prioritize the dignity of sacred text over infinite design flexibility.
syedaslam.com
January 16, 2026 at 4:06 PM
AyatFlow does one specific thing, correct by default. syedaslam.com/posts/constr...