~/series/
Solution Architecture in 30 Days
28 lessons published · new lesson added daily.
1
Day 1
Why More Agent Output Didn't Speed Up Shipping: The Review Bottleneck
2
Day 2
Why Passing Health Checks Doesn't Mean Requests Are Working
3
Day 3
Why prompt caching cuts tokens 90% but not your bill
4
Day 4
Why more AI-generated PRs increased cycle time
5
Day 5
Why adding people or agents can make a team slower
6
Day 6
How retries can double-charge customers in agent pipelines
7
Day 7
How the CAP theorem shows up in multi-agent systems
8
Day 8
When two agents give contradicting answers and nothing errors
9
Day 9
The Replica Race That Makes Users Think Their Save Vanished
10
Day 10
Why Your Shard Key Will Eventually Need to Change
11
Day 11
How a bad partition key leaves coding agents idle on a hot shard
12
Day 12
Why a 95% cache-hit ratio hides slow pages for logged-in users
13
Day 13
Why an API gateway can't see your multi-agent LLM costs
14
Day 14
How a rate limiter that queues requests spreads latency downstream
15
Day 15
How a message queue hides a dead consumer
16
Day 16
How event-driven architecture hides runtime coupling between services
17
Day 17
CQRS
18
Day 18
Saga Compensation Taxonomy
19
Day 19
Circuit breakers
20
Day 20
Idempotency for LLM agents
21
Day 21
Service discovery
22
Day 22
Load shedding for agent systems
23
Day 23
Backpressure
24
Day 24
Bulkhead pattern
25
Day 25
Blue-green deployment with state
26
Day 26
Reversibility and canary releases
27
Day 27
Run records for LLM agents
28
Day 28
C4 diagrams