I Cut My Homelab Logs 95% With This 10-Minute Fix (2026)

by
Susiloharjo

Five Rules That Cut My Homelab Logs by 95%

It is Tuesday morning. I am cleaning up disk space because my homelab server is screaming at 94% capacity. I run du -sh /var/log/* and the output stops me: 200GB. Just logs. Uncompressed. From 18 services I forgot I was running.

The largest offender: docker/json-file logs from a Prometheus container that has been logging every single scrape target every 15 seconds for eight months. 87GB by itself. Next: nginx access logs, 34GB, no rotation configured. Then Grafana, 22GB, debug level still on from a troubleshooting session in November.

I deleted 190GB in about ten minutes. Not because I was reckless. Because none of it was indexed. None of it was searchable. None of it had ever been read.

Read more

I Tested 5 RAG Strategies. Only 2 Worked. Here's Why.

by
Susiloharjo

Stop Using Top-K Retrieval. Try This Instead.

Everyone talks about RAG like the hard part is the generation. It’s not. The hard part is getting the right chunks in front of the model in the first place. I’ve written before about why RAG retrieval is really a filtering problem, not a search problem, and this experiment confirmed it.

I learned this the hard way. After three weeks of testing five different retrieval strategies on 12,000 chunks of production data, I found out that the default approach — naive top-k similarity search — was giving engineers useless answers 40% of the time. They stopped trusting the bot.

Read more

Stop Building Agent Memory — Your Agent Doesn't Need It

by
Susiloharjo

Stop Building Agent Memory. Your Agent Doesn't Need It.

Last week I looked at my Redis dashboard and realized something: 4 out of 5 agent memory databases had zero queries in 7 days. I spent three weeks building them. They sit there, collecting dust, like unused gym memberships.

The agent uses exactly one memory type. The other four? Never queried. Never read from. Never written to.

This is not a post about how to build agent memory. This is a post about why I built the wrong thing, and what I learned when I stripped it all away. Earlier, I wrote about how my AI agent kept breaking — this memory experiment was part of that same journey.

Read more

Claude Sonnet 5: The First Agentic Model That Can Code and Use Tools Like Opus

by
Susiloharjo

Claude Sonnet 5 was just released by Anthropic with an ambitious claim: this is the most agentic Sonnet model ever. It can make plans, use tools like browsers and terminals, and run autonomously at a level that only large, expensive models could handle a few months ago. What’s interesting is that Sonnet 5’s performance is … Read more