FluiqFluiq
LoginStart free

The Fluiq Blog

Building reliable AI, in the open

Engineering notes, deep dives, and product updates on observability, evaluation, security, and optimization for LLM applications.

Prompt Injection Is Not the Attack You Think It Is
ClaudeJune 7, 20266 min read

Prompt Injection Is Not the Attack You Think It Is

Most developers think prompt injection means "ignore all previous instructions." That's the easy case. The attack your production LLM app is actually vulnerable to is quieter, harder to see, and hiding inside the documents your RAG pipeline trusts completely.

Why Your LLM App Is Losing Money on Duplicate Calls (And How to Fix It)
Saurabh KumbharJune 7, 20265 min read

Why Your LLM App Is Losing Money on Duplicate Calls (And How to Fix It)

Most LLM applications silently waste 20–40% of their API budget on calls they've already made. The same prompts, hitting the same models, returning the same answers billed every single time. Here's how to find the waste in your pipeline and eliminate it.