July 28, 2026
Logs out, claws out
Show HN: Ctrlb-decompose: Strip the noise from logs before sending to LLMs
This log-cleaning tool has coders cheering, side-eyeing, and debating if AI gets smarter or dumber
TLDR: Ctrlb-decompose claims it can condense huge piles of server logs into a few simple patterns before sending them to AI, making debugging cheaper and easier. Commenters loved that it’s a small standalone tool, but skeptics challenged whether summarizing logs actually helps AI or just hides useful clues.
A new Hacker News launch called Ctrlb-decompose is promising something catnip-level irresistible to exhausted developers: take a mountain of messy server logs and shrink them into a tiny set of readable patterns before handing them to an AI. In plain English, it tries to turn millions of confusing machine messages into a short summary of what’s actually going wrong. The pitch is dramatic — huge reduction, cleaner patterns, fewer tokens, less noise — and the comments instantly turned into a mini referendum on how much people trust “AI helper” tools in the first place.
The biggest applause came from the "finally, a tool that doesn’t make me buy into a whole empire" crowd. One commenter praised its simple, modular vibe, basically celebrating that it behaves like an old-school command-line tool instead of trying to drag users into yet another sprawling platform. That “just do one job well” energy clearly landed. Another person immediately started building a buddy-comedy workflow around it, comparing it to their own file-peeking tool, which gave the thread a very nerdy cinematic-universe feel.
But then came the skeptical side-eye. One commenter threw the sharpest hot take of the thread: what if tools like this actually make large language models more confused, not less? That question quietly stole the spotlight. The creator responded with calm “this is step one” energy, hinting at bigger ambitions around tracking systems and metrics later. So the mood? Half "shut up and take my logs," half "prove it saves money first."
Key Points
- •Ctrlb-decompose is described as a tool that reduces large volumes of raw log lines into a smaller set of structural patterns with typed variables, statistics, anomaly flags, and severity scoring.
- •The tool is offered as a CLI, a browser-based application via WASM, and a Rust library.
- •Its processing pipeline performs timestamp normalization, CLP encoding, Drain3 clustering, variable extraction and typing, statistics accumulation, anomaly detection, and scoring/correlation in a single streaming pass.
- •Variable classification includes semantic types such as IPv4, UUID, Duration, HexID, Integer, Float, Enum, Timestamp, and String.
- •The article says outputs can be generated for terminal viewing, compact markdown for LLMs, or JSON for downstream systems.