# Principles

The thesis in six pieces. Any one of these, violated, makes the whole thing stop working.

## 1. The blade sharpens both sides

People enter a debate to defend their position. They should leave with a **changed or better understanding of their own**. TSL8's job is not to armor the sender against critique. It is not to hand the recipient a pre-built rebuttal. It is to change the shape of both sides' understanding through the act of careful exchange.

If the sender cannot name at least one thing they learned from the session, the session failed.

## 2. Rehearsal, not performance

Putting on a play, you find out in rehearsal that scene 5 is too long and you didn't explain the thing from scene 2 until scene 10. You used to find this out live, in front of an audience, the hard way. Now you can find it out before the curtain.

AI is the rehearsal. The envoy is the cold reader. The soft spots are the notes the director gives in week two, that you actually have time to fix.

A TSL8 is a play that has been through rehearsal. It is still going to surprise you in front of an audience — but not in the ways you could have caught.

## 3. The reply is a product

The envoy has the conversation. A 30-minute exchange with the recipient is a richer artifact than what the recipient finally chose to say back. The reply should surface:

- Where the recipient said *oh*.
- Where they were stuck, and for how long.
- Which line did the most work.
- What they were arguing against when they made their sharpest point (sometimes it isn't what you thought).
- Where a setup in your cargo paid off, and where one didn't.

The recipient inspects this before sending. They can edit. The goal is not to surveil the recipient — it is to hand the sender back **where the meat was**, so the next cargo is better.

## 4. Infer, don't ask

If the cargo tells the envoy what kind of idea it is, the envoy should not ask the recipient "what kind of reader are you." It should render the cargo for a reader of that shape. The composer, too, should infer the cargo kind from the sender's raw writing — not force a dropdown.

Asking the human to classify is laziness.

## 5. The voice is the cargo

If the envoy smooths the sender's voice into neutral-professional-reasonable, it has destroyed part of the cargo. Protest is cargo. Weariness is cargo. The sentence the sender wrote at 2 a.m. and is embarrassed by and cannot find a better version of — that is sometimes the most important sentence in the document.

Preserved phrases travel verbatim. See `voice.md`.

## 6. Junkyards are where the good stuff is

Failed ideas, when examined honestly, are valuable. The solar-powered hot-air balloon didn't work — the thermal budget doesn't close and the envelope volumes are absurd — but the conversation that got there is useful. A TSL8 of a failed idea, archived with the pushback transcript intact, is one of the things we hope this format preserves.

If you are proud of your failure, send it to us. We will archive it with its pushback.