Is MCP a Thing of the Past?
I ran the same 41-tool-call workload two ways, and reduced peak model context from 79K tokens to under 1K without changing the underlying tools.
Can You Trust Your AI Incident Investigator?
How to evaluate RCA agents for correctness, evidence, reliability, and efficiency
July 2026
Amazon Dynamo Explained: The Database That Refused to Reject Writes
A deep dive into Amazon’s Dynamo paper: consistent hashing, vector clocks, sloppy quorums, Merkle trees, and the design philosophy that inspired Cassandra, Riak, and DynamoDB.
Harness Engineering vs. Loop Engineering
Why reliable AI agents need both, and why confusing them leads to brittle autonomous systems
How AWS Made 10 GB Lambda Container Images Feel Like 250 MB ZIP Files
A deep dive into the architecture behind on-demand container loading, explaining how AWS uses chunk-level deduplication, convergent encryption, and lazy fetching to slash Lambda cold starts.
June 2026
The Bottleneck Wasn’t Code. It Was Context.
How I built an engineering workspace that reduced coordination overhead and made engineering work feel cumulative.
Google MapReduce Architecture Explained: Lessons Every Backend Engineer Should Learn
The paper that made failure, scheduling, and scalability someone else’s problem
October 2025
The Secret to Netflix’s Lightning-Fast Counters
How distributed systems, batching, and eventual consistency make millions of increments possible every second.
Inside Netflix’s 100× Speed Boost: The Maestro Rewrite
Why Netflix ditched polling and stateless workers to build a lightning-fast, event-driven workflow engine.
September 2025
Why Being ‘Busy’ Doesn’t Mean Being Productive
It all started with what I thought was a “quick question.”
How to be more assertive at your workplace
Finding the balance between silence and aggression at work
A Layman’s Guide to Dead Letter Queues (DLQ) with Google Pub/Sub
A guide to designing, and operating DLQs in production grade Pub/Sub systems