Notes from building backend systems: what I designed, what broke, and the reasoning I wish I had written down the first time.
Topics
Not tutorials. Working notes: what I tried, what it cost, and what I would tell myself before starting.
Agents, LLMs, RAG and AI architecture — what I build, test and break.
Distributed systems, databases, messaging, caching and scaling.
Working references from Java, Spring, PostgreSQL, Kafka, Redis and Kubernetes.
Full write-ups from problem to trade-offs, with the diagrams to match.
Latest
Newest first. Some are finished thoughts, others are still moving.
Every week the same few volunteers spent two hours or more chasing sermon slides, copying Bible verses and fixing line breaks. So I built them an agent that does it on Thursday evening, for about seven cents, and still lets a person say no.
How the Sunday slides agent proves it never changed a verse, why Claude only gets to propose design changes, and how the design pass went from 50 cents to 8 cents a build.
Two Railway services, one cron, a webhook, a watchdog and a cost log. The real numbers from the Sunday slides agent's first weeks in production, including the Thursday it failed.
What actually happens when the same message arrives twice, and the smallest set of changes that makes a consumer safe to retry.
The classic interview system, built small and measured: where the reads go, what the cache is for, and the one decision that made everything else easy.
I assumed the model would use the tools sensibly. It did not. What changed once I started designing tools the way I design an API: narrower inputs, typed arguments, and error messages written for the model rather than for me.
Three index changes that looked obvious in review and cost more than they saved once real traffic arrived.
The smallest useful implementation, the two parameters that matter, and how wrong the false-positive rate gets when you guess them.
Two words used interchangeably in meetings, and the operational difference that makes them not the same thing at all.
Cache-aside looked simple until two writers raced. What I changed, and what I would measure before adding a cache next time.
Not a benchmark. A comparison of the questions each one makes you answer: about ordering, replay, consumers and who owns the offset.
Executors, CompletableFuture and virtual threads — written down so I stop rediscovering the same three pitfalls.
Next
I am always up for a conversation about architecture, trade-offs, or a bug that refuses to die.
Start a conversation