discovered this by accident while trying to stretch my free tier.
was burning through messages embarrassingly fast. long prompts. detailed context. full sentences. please and thank you. the whole thing.
then one day i was tired and just typed:
"fix bug. line 47. null error."
it fixed it.
same quality. one fifth of the tokens.
i sat there staring at it like i'd discovered fire.
the caveman theory in one sentence:
Claude is not your colleague. it does not need pleasantries. it does not need full sentences. it needs information. just information. nothing else.
before caveman theory:
"hey Claude, i hope this makes sense but i've been working on this project and i'm running into an issue with the function on line 47, it keeps throwing a null error and i'm not sure what's causing it, could you take a look and help me figure out what's going wrong?"
57 words. full credits burned. Claude reads the pleasantries and processes zero useful information from them.
after caveman theory:
"line 47. null error. fix."
4 words. same output. same quality.
53 words of your credits just evaporated into politeness.
the full caveman framework:
no greetings. Claude doesn't need good morning. it doesn't have mornings. skip it entirely.
no apologies. "sorry if this is a weird question" — five words of pure credit waste. just ask the question.
no filler context. "i've been working on this for a while and" — Claude doesn't care. it needs the what not the backstory of the what.
no closing remarks. "thanks so much this was really helpful" — you're paying per token to say thank you to software. stop.
verbs only where possible. "summarise." "fix." "rewrite shorter." "find the bug." "make it casual." complete sentences are for humans talking to humans.
use symbols not words. instead of "can you compare option A versus option B" just type "A vs B?" Claude knows what that means.
real examples from my last week:
instead of: "could you help me make this email sound more professional and formal while keeping the core message intact"
caveman says: "email. more formal. keep meaning."
instead of: "i need you to summarise this document and pull out the key points that are most relevant to a business audience"
caveman says: "summarise. business audience. key points only."
instead of: "what do you think would be the best approach to structuring a landing page for a SaaS product targeting small business owners"
caveman says: "SaaS landing page. small business. best structure."
the one exception:
complex creative work. writing with a specific voice. nuanced emotional stuff.
caveman theory breaks here. those tasks need real context because vague input produces vague output.
caveman is for tasks where the instruction is clear and the only waste is ceremony.
which is honestly about 70% of what most people use Claude for daily.
the uncomfortable math:
if you're on free tier every wasted word is a message you don't get to send later.
if you're on paid every wasted word is money.
nobody told you this when you signed up. the product doesn't benefit from you being efficient with tokens. you figured it out or you didn't.
the meta irony:
this entire post explaining caveman theory is the opposite of caveman theory.
a caveman would have just posted:
"talk Claude like caveman. short prompt. save credit. good output. try it."
and honestly that would have been enough.
what's the most bloated prompt you've been writing that caveman theory would destroy in four words?
[link] [comments]