Milos Dordevic
@milosdordevic.bsky.social
~Build Your Future Today~
Backend maturity level:
Level 1 – Logs
Level 2 – Metrics
Level 3 – Traces
Level 4 – You can answer “why” without SSH-ing into a server.
Level 1 – Logs
Level 2 – Metrics
Level 3 – Traces
Level 4 – You can answer “why” without SSH-ing into a server.
February 12, 2026 at 4:22 PM
Backend maturity level:
Level 1 – Logs
Level 2 – Metrics
Level 3 – Traces
Level 4 – You can answer “why” without SSH-ing into a server.
Level 1 – Logs
Level 2 – Metrics
Level 3 – Traces
Level 4 – You can answer “why” without SSH-ing into a server.
Most “AI latency” problems are actually architecture problems.
Parallelize embedding, retrieval, and enrichment steps.
The model is often the fastest part.
Parallelize embedding, retrieval, and enrichment steps.
The model is often the fastest part.
February 12, 2026 at 4:22 PM
Most “AI latency” problems are actually architecture problems.
Parallelize embedding, retrieval, and enrichment steps.
The model is often the fastest part.
Parallelize embedding, retrieval, and enrichment steps.
The model is often the fastest part.
If scaling your API requires vertical scaling only, you designed it wrong.
Stateless services + idempotent endpoints + externalized state = horizontal freedom.
Stateless services + idempotent endpoints + externalized state = horizontal freedom.
February 12, 2026 at 4:22 PM
If scaling your API requires vertical scaling only, you designed it wrong.
Stateless services + idempotent endpoints + externalized state = horizontal freedom.
Stateless services + idempotent endpoints + externalized state = horizontal freedom.