⚡ Caveman Converter — paste your bloated prompt, get compressed version
Your prompt (bloated)
→
Caveman version
Input: 0 tokens
Output: 0 tokens
Saved: 0 tokens (0%)
⚡ Quick Rules Applied by Converter
Remove: can you / please / I need / I want
Remove: Could you help me with / I was wondering
Remove: please / kindly / thank you
Replace: that → →
Replace: and → +
Remove: articles: a / an / the
10 Core Token Reduction Rules
-
R01
Kill filler phrases HIGH IMPACT Remove: "Can you please help me..." → just say what you want Remove: "I was wondering if..." / "Could you kindly..." / "I'd like you to..."
-
R02
Drop articles + prepositions HIGH IMPACT ❌ "Write a function that connects to the database" ✓ "fn: connect database"
-
R03
Use RTK symbols instead of words HIGH IMPACT → = results in / leads to / gives + = and / with / plus | ~ = approx | ? = question
-
R04
Paste errors directly — no wrapper text MED IMPACT ❌ "I'm getting this error when I run my code: [error]" ✓ "fix: [paste error]"
-
R05
Limit code pasting — only relevant snippet HIGH IMPACT Don't paste 200 lines when the bug is in 10 lines ✓ "line 45-60 broken: [snippet]"
-
R06
Use "short" / "brief" / "1 line" to control output length MED IMPACT ✓ "explain JWT. short." — saves tokens on Claude's response too
-
R07
Refer to previous message with ^ or "above" LOW IMPACT ❌ "Based on the code you wrote earlier for the Flask app..." ✓ "^ add error handling"
-
R08
Use colons to define task type MED IMPACT fix: / add: / explain: / rewrite: / debug: / review: ✓ "rewrite: [code] → cleaner"
-
R09
Avoid repeating context Claude already has HIGH IMPACT If you gave context at chat start, don't re-explain every message ✓ Use a context block ONCE at top, then reference it with "^ context"
-
R10
Specify format to avoid verbose responses MED IMPACT ✓ "list only" / "code only" / "yes/no" / "2 sentences" ✓ "no explanation, just code"
RTK — Reduce Token Keywords Symbol System
→results in / gives / leads to / convert
+and / with / plus / also
^previous / above / that / the above
~approximately / around / similar to
?question / clarify / is this / check
>>transform / refactor / convert into
!important / must / critical / required
@at / using / via / with (tool/service)
#number / count / total / quantity
✗not / don't / avoid / remove / no
✓yes / keep / use / include / do this
vscompare / difference / versus
Common Abbreviations
fnfunction
dbdatabase
cfgconfig / configuration
envenvironment variables
errerror
reqrequest / required
resresponse / result
authauthentication
pyPython
jsJavaScript
apiAPI endpoint/call
msgmessage
Real Examples — Deepak's Stack
Can you please help me write a Python Flask route that receives a WhatsApp webhook message, extracts the user's phone number and text, and then stores it in Google Sheets using gspread?
py flask route: recv WA webhook → extract phone+text → store @gspread
~65% token reduction
I'm getting an error when I try to deploy my Flask app to Railway. The error says that the port binding is failing. Could you help me understand why this might be happening and how I can fix it?
fix: Railway deploy err → port binding fail. Flask. why + fix?
~70% token reduction
I need you to review my existing Python code that sends messages to WhatsApp using the Meta Cloud API and tell me if there are any issues with how I'm handling the authentication token and if the error handling is good enough
review: py code @Meta WA CloudAPI → auth token handling + err handling ok?
~68% token reduction
Can you write a JavaScript function that takes a salary amount in Indian Rupees and calculates the take-home salary after applying the new tax regime for FY 2026-27, including all the standard deductions?
js fn: INR salary → takehome @new tax regime FY26-27 + std deductions
~60% token reduction
Your Personal Context Block — Paste at Start of New Chats
[CTX] Deepak. SAP consultant Hoskote/BLR.
Projects: HoskoteClassifieds(Flask/Railway/WA-bot/GSheets/Cloudinary), salarybit.in(tools/blog/adsense), VahanClear, CreditVaidya.
Stack: Python, Claude API, Groq, Railway, GSheets, Razorpay, Cloudinary, Meta WA CloudAPI.
[RULES] Short Q = short A. No intros. No summaries. Code only if asked. Assume ^ ctx always.
Paste this ONCE at the start of a new chat. Saves ~50-80 tokens per message by not re-explaining context.
Task Starter Templates
Debug / Fix
fix: [paste error or broken code]
lang: py/js | context: [1 line what it should do]
Write New Code
fn: [what it does] → [output]
lang: py | use: [library if any] | ✗ [what to avoid]
Review / Audit
review: [paste code]
check: security + perf + edge cases. short findings only.
Explain Concept
explain: [topic]. 3 sentences. no jargon. eg included.
Continue / Iterate
^ add: [what to add]
^ fix: [what's broken]
^ rewrite → [improvement goal]
Pre-Send Checklist — Run this before every prompt
Compression score
0%
- Removed "Can you please / I need / I want / Could you help me"
- Dropped unnecessary articles (a, an, the)
- Used symbols where possible (→, +, ^, ~)
- Pasted only relevant code snippet, not entire file
- Used task prefix (fix: / add: / explain: / review:)
- Specified output format (short / code only / list / yes-no)
- Not re-explaining context already given earlier in chat
- Used ^ to refer to previous output instead of repeating it