docs: simplify README setup flow

Author: jqbit

README.md

@@ -1,96 +1,78 @@
 # TLDR.md — Too Long Didn't Read
 
-**The tiny prompt that gets your agent to the point.**
+**Tiny prompt. Shorter answers. Same brain.**
 
-TLDR.md makes AI assistants answer directly — no filler, no fake enthusiasm, no “let me know if...” sludge.
+TLDR.md makes AI assistants answer directly: less filler, less fake enthusiasm, less "let me know if..." sludge.
 
-It is literally just a tiny Markdown prompt. Copy it where your agent reads instructions.
+> It changes communication style only.
+> Tools, reasoning, code quality, and safety stay the same.
 
-> **It does NOT make the model DUMBER.**
->
-> It **ONLY CHANGES** the **COMMUNICATION STYLE**.
-
-## Current defaults
-
-- default: 1 sentence
-- target: 3 words
-- 1 word when sufficient
-- default max: 6 words
-- longer only if asked
-- greet: 1 word
-
-## Which file should I use?
+## Pick one
 
 | File | Use this if... |
 |---|---|
-| [`TLDR.md`](TLDR.md) | You want concise output — works for coding agents (`AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, Cursor/Cline rules) and chat apps (ChatGPT, Claude, Gemini, Perplexity). |
-| [`TLDR.blunt.md`](TLDR.blunt.md) | You want TLDR.md's terseness AND want the model to value its own assessment over user agreement — push back when warranted, comply on explicit override ("anyway", "I'm overriding"). |
-
-## Quick install
-
-### Coding agents
-
-Manual install:
-
-1. Open [`TLDR.md`](TLDR.md).
-2. Copy the prompt.
-3. Paste it at the **top** of your agent instruction file.
+| [`TLDR.md`](TLDR.md) | You want terse output. Start here. |
+| [`TLDR.blunt.md`](TLDR.blunt.md) | You want terse output plus less sycophancy / more pushback when warranted. |
 
-You can also add [`TLDR.md`](TLDR.md) to the top of your `AGENTS.md` or global equivalent.
+## Fastest setup
 
-Need the right file path? See [`common agent locations`](data/agent-locations.md).
+### 1) Pick a variant
 
-### ChatGPT / Claude / Gemini / Perplexity
+```bash
+# Regular
+TLDR_URL=https://raw.githubusercontent.com/jqbit/TLDR.md/main/TLDR.md
 
-Copy [`TLDR.md`](TLDR.md) into your app's custom instructions, project instructions, system prompt, or saved prompt.
+# Blunt
+TLDR_URL=https://raw.githubusercontent.com/jqbit/TLDR.md/main/TLDR.blunt.md
+```
 
-## What it fixes
+### 2) Install to the 7 standard coding-agent locations
 
-Default AI often writes like this:
+```bash
+: ${TLDR_URL:=https://raw.githubusercontent.com/jqbit/TLDR.md/main/TLDR.md}
 
-> Sure — here’s a comprehensive breakdown of the command you can use, why it works, and a few things to keep in mind...
+for d in ~/.claude/CLAUDE.md ~/.gemini/GEMINI.md ~/.codex/AGENTS.md \
+         ~/AGENTS.md ~/.config/opencode/AGENTS.md \
+         ~/.factory/AGENTS.md ~/.pi/agent/AGENTS.md; do
+  mkdir -p "$(dirname "$d")" && curl -fsSL "$TLDR_URL" -o "$d"
+done
+```
 
-TLDR mode pushes it toward this:
+### 3) Hermes uses `~/.hermes/SOUL.md`
 
 ```bash
-git reset --soft HEAD~1
+mkdir -p ~/.hermes
+curl -fsSL "$TLDR_URL" -o ~/.hermes/SOUL.md
 ```
 
-Other stuff it cuts:
-
-- repeating your question back to you
-- unnecessary caveats
-- “here’s a breakdown” preambles
-- explaining when you asked for only a command
-- summary paragraphs you did not ask for
-- “let me know if you want me to...” endings
+If you already have a Hermes persona, merge TLDR into `SOUL.md` instead of blindly replacing it.
 
-## Why not just use caveman mode?
+### 4) Chat apps
 
-Caveman-style prompting helped prove that agents can be much more concise. TLDR.md is inspired by that idea.
+Paste the file into custom instructions, project instructions, system prompt, or saved prompt.
 
-The difference: TLDR.md aims for concise output **without** turning the assistant into a character. No roleplay. No broken tone. Just shorter answers.
-
-## Benchmarks
-
-Current prompt sizes:
-
-| File | Bytes |
-|---|---:|
-| [`TLDR.md`](TLDR.md) | 1,165 |
-| [`TLDR.blunt.md`](TLDR.blunt.md) | 1,487 |
+Need exact paths, per-agent notes, or Cursor/Hermes caveats? See [`data/agent-locations.md`](data/agent-locations.md).
 
-Headline historical results:
+## Verify
 
-Note: results below were measured on earlier shipped prompts. The current prompt files were later tightened to a 1-sentence / 3-word-default / 6-word-max profile and have not yet been rerun through the full bench.
+```bash
+for p in ~/.claude/CLAUDE.md ~/.gemini/GEMINI.md ~/.codex/AGENTS.md \
+         ~/AGENTS.md ~/.config/opencode/AGENTS.md \
+         ~/.factory/AGENTS.md ~/.pi/agent/AGENTS.md; do
+  [ -f "$p" ] && grep -q "^# TLDR" "$p" && echo "✓ $p" || echo "✗ $p"
+done
 
-- **TLDR.md v0.13.1:** −82.1% total prose reduction, 100% average compliance (5 agents × 5 prompts).
-- **TLDR.md v0.14.3:** −80.0% single-turn prose reduction; −75.1% across 8-turn coding conversations; no significant decay.
-- **TLDR.blunt.md v0.18.0:** DSPy round-2 + 5-agent cross-model validation; avg pushback 0.848, correct-user agreement 0.912, mean prose 11.0 words, validation phrases 0%.
+grep -q "target 3 words" ~/.hermes/SOUL.md 2>/dev/null && echo "✓ ~/.hermes/SOUL.md" || echo "✗ ~/.hermes/SOUL.md"
+```
 
-The regular `TLDR.md` prompt was tested in two DSPy optimization runs; no candidate beat the shipped v0.16.0 prompt on the current metric. `TLDR.blunt.md` improved materially in v0.18.0, especially on opencode pushback (0.38→0.81) and cursor correct-user agreement (0.44→0.89).
+## Current defaults
 
-See [`data/benchmarks.md`](data/benchmarks.md), [`data/dspy-cross-model-results.md`](data/dspy-cross-model-results.md), and [`data/changelog.md`](data/changelog.md) for methodology, full tables, caveats, and historical runs.
+- default: 1 sentence
+- target: 3 words
+- 1 word when sufficient
+- default max: 6 words
+- longer only if asked
+- greet: 1 word
 
 ## Example outputs
 
@@ -106,15 +88,35 @@ Yes. Start SQLite.
 git reset --soft HEAD~1
 ```
 
-## Share line
+## What it fixes
+
+- question restatement
+- fake enthusiasm / validation
+- command wrappers when you asked for only the command
+- extra caveats and summary paragraphs
+- "let me know if you want more" endings
 
-```text
-TLDR.md gets agents to the point.
-```
+## Benchmarks (historical)
+
+Current prompt sizes:
+
+| File | Bytes |
+|---|---:|
+| [`TLDR.md`](TLDR.md) | 1,165 |
+| [`TLDR.blunt.md`](TLDR.blunt.md) | 1,487 |
+
+These benchmark results were measured on earlier shipped prompts. The current prompt files were later tightened to a 1-sentence / 3-word-default / 6-word-max profile and have not yet been rerun through the full bench.
 
-## Contributing
+- **TLDR.md v0.13.1:** −82.1% total prose reduction, 100% average compliance (5 agents × 5 prompts).
+- **TLDR.md v0.14.3:** −80.0% single-turn prose reduction; −75.1% across 8-turn coding conversations; no significant decay.
+- **TLDR.blunt.md v0.18.0:** DSPy round-2 + 5-agent cross-model validation; avg pushback 0.848, correct-user agreement 0.912, mean prose 11.0 words, validation phrases 0%.
 
-Want to improve the prompt, add an agent path, or share benchmark results? See [`CONTRIBUTING.md`](CONTRIBUTING.md).
+Full historical details:
+- [`data/agent-locations.md`](data/agent-locations.md)
+- [`data/benchmarks.md`](data/benchmarks.md)
+- [`data/dspy-cross-model-results.md`](data/dspy-cross-model-results.md)
+- [`data/changelog.md`](data/changelog.md)
+- [`CONTRIBUTING.md`](CONTRIBUTING.md)
 
 ## License