You type tar -xvf and hit enter. Wait, was it -x to extract or -c to create? You open the man page. You are immediately hit with a wall of text detailing the history of tape archives since 1979. You don’t need a history lesson. You need the flags.
We’ve been told the terminal is hostile. We’ve been told the issue is insufficient documentation. So, the tech industry’s reflex? Slap an AI on it. Ask a language model how to untar a file, and you might get a perfectly formatted, completely hallucinated command that deletes your root directory.
AI doesn’t fix bad documentation; it just hallucinates a prettier version of it.
The problem was never a lack of information. The problem was surviving it. We’ve confused ‘more information’ with ‘more understanding,’ and nowhere is that clearer than the command line. Man pages are dense because they are exhaustive, not because they are helpful.
Enter whats. It’s a command-line tool that does exactly one thing: it reads the existing --help output or man page, and uses pure regex to extract just the flag descriptions. No AI. No LLM slop. No waiting for a model to generate a response. Just the exact data you need, stripped of the noise.
As one developer noted after looking under the hood: there is zero machine learning magic here. It’s just a minimalist filter that extracts the essential flags from the source of truth.
But here is the paradox. To use a tool that simplifies tools, you have to learn another command. Adding a meta-layer of complexity to solve complexity is the ultimate developer trap. Yet, whats earns its keep. The cognitive load it removes from parsing hundreds of dense manuals is massive compared to the single command it asks you to remember.
Stop trying to outsource your terminal knowledge to a language model. The answers are already on your machine; they’re just buried under decades of Unix bloat. whats hands you the shovel. Start digging.
FAQ
Q: Doesn't adding a new command to learn defeat the purpose of simplifying the CLI?
A: Yes, it's a paradox. But learning one command (whats) to instantly decode hundreds of others is a massive net positive for your mental bandwidth.
Q: How does it work without an AI model?
A: It uses pure regular expressions to parse existing --help outputs and man pages, extracting only the flag descriptions. Zero hallucinations, zero latency.
Q: Isn't AI better at explaining commands than regex?
A: No. AI is a crutch that frequently invents flags that don't exist. Regex just strips the fat from the actual source of truth.