Skip to main content
The cheapest token is the one you never send. But the real prize is not the token savings. A compressed request is a clearer request, and clarity is what actually changes the answer. Compression is not about being terse. It is about density: more signal per token, not fewer tokens at any cost. Sometimes the dense version is longer in one spot, because naming the deliverable precisely costs a line and saves a rewrite.

​
Work the press

Below is a real request the way most people write it: warm, apologetic, narrative, and expensive. Four handles, four compression moves. Pull them in any order and watch the text transform, the token needle drop, and the clarity gauge climb.
Notice the fourth move. It made the request longer and better. That is the point: compression optimizes signal per token, and sometimes the highest-signal move is adding the sentence that says exactly what you want back.

​
The four moves, named

  1. Cut the warm-up. Greetings, apologies, and hedges carry zero information. Claude needs no rapport.
  2. Facts, not story. Turn the paragraph of history into three lines of numbers and constraints. Stories are for humans; specs are for work.
  3. Summarize, don’t paste. One line saying what the fourteen-email thread concluded beats the thread itself.
  4. Name the deliverable. Format, length, audience, and what “done” looks like. The only move that spends tokens, and the best one.

​
Let Claude do the compressing

You do not have to compress by hand. Claude is excellent at packing its own suitcase. At the end of any long working thread:
The same trick works on documents before they enter a thread (“Compress this report to the 15 facts a negotiator would need”) and on handoffs to teammates (“Write the two-paragraph version a new team member needs to continue this”).

​
The test of a compressed request

Read it back and ask: could a sharp colleague act on this without a follow-up question? If yes, it is dense enough. If they would have to ask “what format?” or “which client?” or “by when?”, the compression cut signal instead of fat. Put the line back. Fluffy and long fails twice: it costs more and communicates less. Dense wins twice. That is the whole economics of this section in one sentence.