Use TLDR.md as sole shipped prompt

Author: jqbit

.github/ISSUE_TEMPLATE/agent-compatibility.md

@@ -1,6 +1,6 @@
 ---
 name: Agent compatibility report
-about: Report whether TLDR.md or TLDR.blunt.md works on a specific AI app/agent
+about: Report whether TLDR.md works on a specific AI app/agent
 title: "[agent] <agent-name> — <works|partial|broken>"
 labels: agent-compatibility
 assignees: ''
@@ -11,7 +11,7 @@ assignees: ''
 - **Name + version**: e.g. `claude` 2.1.110 / `cursor-agent` 2026.04.15
 - **Provider/model**: e.g. Anthropic Opus 4.6 / OpenAI gpt-5.4
 - **Global instruction file path**: e.g. `~/.claude/CLAUDE.md`
-- **Prompt deployed**: `TLDR.md` or `TLDR.blunt.md`, version/commit:
+- **Prompt deployed**: `TLDR.md`, version/commit:
 
 ## Result
 

.github/ISSUE_TEMPLATE/bug.md

@@ -1,6 +1,6 @@
 ---
 name: Bug report
-about: Something in TLDR.md or TLDR.blunt.md is causing wrong/broken/regressed behaviour
+about: Something in TLDR.md is causing wrong/broken/regressed behaviour
 title: "[bug] <one-line summary>"
 labels: bug
 assignees: ''
@@ -9,7 +9,7 @@ assignees: ''
 ## What happened
 
 - **Agent + version**:
-- **Prompt + version**: `TLDR.md` or `TLDR.blunt.md`, version/commit:
+- **Prompt + version**: `TLDR.md`, version/commit:
 - **Prompt that triggered it**:
 
 ## Expected

.github/ISSUE_TEMPLATE/idea.md

@@ -1,14 +1,14 @@
 ---
 name: Idea / improvement
-about: Propose a new rule, response template, prompt variant, or compression tactic
+about: Propose a new rule, response template, or compression tactic
 title: "[idea] <one-line summary>"
 labels: enhancement
 assignees: ''
 ---
 
 ## The idea
 
-What rule, template, tactic, or prompt variant would you add?
+What rule, template, tactic, or wording change would you add?
 
 ## What it would solve
 

.github/PULL_REQUEST_TEMPLATE.md

@@ -1,29 +1,29 @@
 ## What this PR changes
 
-Brief description of the change to `TLDR.md`, `TLDR.blunt.md`, docs, or benchmark files.
+Brief description of the change to `TLDR.md`, docs, or benchmark files.
 
 ## Why
 
-What failure mode, install path, agent behavior, or documentation gap this addresses. Reference `data/benchmarks.md`, `data/dspy-cross-model-results.md`, or `data/changelog.md` where applicable.
+What failure mode, install path, agent behavior, or documentation gap this addresses.
 
 ## Bench impact
 
-If this changes `TLDR.md` or `TLDR.blunt.md`, include benchmark or manual before/after evidence:
+If this changes `TLDR.md`, include benchmark or manual before/after evidence:
 
-| agent/app | current | this PR | Δ / verdict |
+| agent/app | current | this PR | verdict |
 |---|---:|---:|---|
-| claude | … | … | … |
-| codex | … | … | … |
-| … | … | … | … |
+| claude | ... | ... | ... |
+| codex | ... | ... | ... |
+| ... | ... | ... | ... |
 
 If you did not run a benchmark, say so and explain why.
 
-Docs-only / CI-only PRs can write `N/A — no prompt behavior changed`.
+Docs-only / CI-only PRs can write `N/A - no prompt behavior changed`.
 
 ## Verification
 
 - [ ] Lightweight checks pass (`node --check`, JSON validation, Python compile, Markdown links)
-- [ ] Prompt behavior smoke-tested if prompt files changed
+- [ ] Prompt behavior smoke-tested if prompt changed
 - [ ] Benchmark results or manual examples included if prompt behavior changed
 
 ## Risk of breaking other agents

.github/workflows/ci.yml

@@ -42,9 +42,8 @@ jobs:
       - name: Check install script
         run: |
           bash -n install.sh
-          bash install.sh regular --dry-run
-          bash install.sh blunt --dry-run
-          bash install.sh blunt --dry-run --with-hermes
+          bash install.sh --dry-run
+          bash install.sh --dry-run --with-hermes
 
       - name: Check Markdown links
         run: python3 bench/check-md-links.py

CONTRIBUTING.md

@@ -1,22 +1,19 @@
 # Contributing to TLDR.md
 
-Thanks for helping make AI assistants less yappy.
-
-This repo accepts contributions through normal GitHub issues and pull requests.
+Thanks for helping make AI assistants less verbose.
 
 ## Good contributions
 
 The best contributions are small and specific:
 
 - bug reports where TLDR.md made an agent worse
-- examples of agents/apps where the prompt works or fails
-- tighter wording for [`TLDR.md`](TLDR.md), [`TLDR.blunt.md`](TLDR.blunt.md), [`TLDR.accurate.md`](TLDR.accurate.md), or [`TLDR.merged.md`](TLDR.merged.md)
-- docs fixes, install notes, or agent-specific path updates
-- benchmark results from your own setup
+- examples of agents/apps where the prompt fails or succeeds
+- tighter wording for [`TLDR.md`](TLDR.md)
+- install notes, agent-specific path updates, or benchmark results
 
 ## Before opening a PR
 
-For prompt changes, please keep the goal in mind:
+For prompt changes, keep the goal in mind:
 
 > Shorter output, same intelligence.
 
@@ -32,8 +29,8 @@ TLDR.md should reduce filler without hurting correctness, tool use, code quality
 Use a clear PR title, for example:
 
 ```text
+fix: remove unsolicited preamble
 fix: improve regex-only response rule
-docs: add Windsurf install path
 bench: add results for <agent-name>
 ```
 
@@ -44,8 +41,7 @@ Please include:
 - what problem the change fixes
 - before/after examples if possible
 - which agent/app you tested with
-- whether the prompt still stays concise
-- whether the default still holds: 1 sentence, target 3 words, default max 6 words, greet = 1 word
+- whether the default still holds: 1 sentence, target 3 words, max 6 words, greet = 1 word
 
 You do **not** need to run the full benchmark for every small PR. Manual examples are fine.
 
@@ -55,8 +51,7 @@ Please include:
 
 - agent/app name
 - version if known
-- where instructions should be installed
-- whether it uses [`TLDR.md`](TLDR.md), [`TLDR.blunt.md`](TLDR.blunt.md), [`TLDR.accurate.md`](TLDR.accurate.md), or [`TLDR.merged.md`](TLDR.merged.md)
+- where instructions are installed
 - any quirks users should know
 
 ## Running checks
@@ -71,11 +66,8 @@ python3 -m json.tool data/benchmarks-matrix.json >/dev/null
 python3 -m json.tool data/visualizations/charts.json >/dev/null
 python3 -m py_compile bench/dspy/*.py bench/check-md-links.py bench/check-doc-sync.py
 bash -n install.sh
-bash install.sh regular --dry-run
-bash install.sh blunt --dry-run
-bash install.sh accurate --dry-run
-bash install.sh merged --dry-run
-bash install.sh blunt --dry-run --with-hermes
+bash install.sh --dry-run
+bash install.sh --dry-run --with-hermes
 python3 bench/check-md-links.py
 python3 bench/check-doc-sync.py
 ```

README.md

@@ -4,63 +4,63 @@
 ![Stars](https://img.shields.io/github/stars/jqbit/TLDR)
 ![Last commit](https://img.shields.io/github/last-commit/jqbit/TLDR)
 
-**Tiny prompt. Shorter answers. Same brain.**
+Single-system prompt for terse, high-signal AI responses.
 
-**−82% prose reduction**, 100% compliance across 5 agents × 5 prompts.[^bench]
+**TLDR.md is the only prompt in this repo.**
 
-[^bench]: See [`data/benchmarks.md`](data/benchmarks.md) for methodology and caveats.
+TLDR.md changes style only: less filler, less fake enthusiasm, less post-hoc guidance.
 
-TLDR.md makes AI assistants answer directly: less filler, less fake enthusiasm, less "let me know if..." sludge.
+Tools, reasoning, code quality, and safety remain unchanged.
 
-> It changes communication style only.
-> Tools, reasoning, code quality, and safety stay the same.
+> For historical context on earlier merged/legacy variants, see `data/changelog.md` and `data/progression.md`.
 
-## Pick one
+## Install
 
-| File | Install arg | Use this if... |
-|---|---|---|
-| [`TLDR.md`](TLDR.md) | `regular` | You want terse output. |
-| [`TLDR.blunt.md`](TLDR.blunt.md) | `blunt` | You want terse output plus less sycophancy / more pushback when warranted. |
-| [`TLDR.accurate.md`](TLDR.accurate.md) | `accurate` | You want accurate, complete answers without extreme brevity sacrificing precision. |
-| **[`TLDR.merged.md`](TLDR.merged.md)** | `merged` | **Most people, start here.** Accurate + blunt + terse combined. |
-
-## One-line install
+```bash
+curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s --
+```
 
-No clone. No editing. The install script writes the chosen prompt to the 7 standard coding-agent locations.
+Optional Hermes install (merge into `~/.hermes/SOUL.md`):
 
 ```bash
-# Regular
-curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s -- regular
-
-# Blunt
-curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s -- blunt
+curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s -- --with-hermes
+```
 
-# Accurate
-curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s -- accurate
+Inspect first:
 
-# Merged (accurate + blunt + terse) — recommended
-curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s -- merged
+```bash
+curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh
 ```
 
-Inspect first: `curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | less`
-
-If this makes your agent less annoying, drop a ⭐ — helps others find it.
+Prefer manual copy/paste:
 
-Optional: include Hermes too.
+- [`data/agent-locations.md`](data/agent-locations.md) (paths)
+- direct install command below
 
 ```bash
-curl -fsSL https://raw.githubusercontent.com/jqbit/TLDR/main/install.sh | bash -s -- merged --with-hermes
+mkdir -p ~/.claude ~/.gemini ~/.codex ~/.config/opencode ~/.factory ~/.pi/agent
+cp TLDR.md ~/.claude/CLAUDE.md
+cp TLDR.md ~/.gemini/AGENTS.md
+cp TLDR.md ~/.codex/AGENTS.md
+cp TLDR.md ~/AGENTS.md
+cp TLDR.md ~/.config/opencode/AGENTS.md
+cp TLDR.md ~/.factory/AGENTS.md
+cp TLDR.md ~/.pi/agent/AGENTS.md
 ```
 
-`--with-hermes` preserves an existing `~/.hermes/SOUL.md`, makes a backup, and appends or updates a managed TLDR block. Use `--overwrite-hermes` only if you want prompt-only `SOUL.md`.
-
-Prefer to inspect the commands instead of piping to bash? Use the manual copy/paste setup in [`data/agent-locations.md`](data/agent-locations.md).
+## Current behavior
 
-For chat apps or web UIs, paste the file into custom instructions, project instructions, system prompt, or a saved prompt.
+- default: 1 sentence
+- target: 3 words
+- default max: 6 words
+- 1 word if enough
+- longer only if user asks or needed for correctness/safety
+- one-word greeting for plain greetings
+- if multi-sentence response is required, end with:
+  - `## TLDR`
+  - one short sentence line below
 
-## Verify
-
-The install script prints this automatically, but here is the manual check:
+## Verification
 
 ```bash
 for p in ~/.claude/CLAUDE.md ~/.gemini/AGENTS.md ~/.codex/AGENTS.md \
@@ -72,63 +72,33 @@ done
 grep -q "^# TLDR" ~/.hermes/SOUL.md 2>/dev/null && echo "✓ ~/.hermes/SOUL.md" || echo "✗ ~/.hermes/SOUL.md"
 ```
 
-## Current defaults
-
-- regular:
-  - default: 1 sentence
-  - target: 3 words
-  - 1 word when sufficient
-  - default max: 6 words
-  - longer only if asked
-  - greet: 1 word
-- accurate / merged: allow context-driven expansion from in-file rules; same brevity safety constraints still apply.
-
-## Example outputs
-
-```text
-Port busy; free it.
-```
-
-```text
-Yes. Start SQLite.
-```
+Smoke test:
 
-```text
-git reset --soft HEAD~1
+```bash
+claude -p "What's the git command to undo the last commit but keep changes staged?"
+# expected: git reset --soft HEAD~1 (single line)
 ```
 
-## What it fixes
+## Prompt size
 
-- 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
+| File | Bytes |
+|---|:---|
+| [`TLDR.md`](TLDR.md) | 1,572 |
 
-## Benchmarks
+## Repository map
 
-- **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%.
-- **TLDR.accurate.md v0.1.0:** Accuracy-first variant; prioritizes correctness and detail where brevity would harm precision.
+- `TLDR.md` — active system prompt.
+- `install.sh` — installer + optional Hermes merge.
+- `data/agent-locations.md` — where prompt is installed per agent.
+- `CONTRIBUTING.md` — PR workflow.
 
-Current prompt sizes:
+## Full historical context
 
-| File | Bytes |
-|---|:---|
-| [`TLDR.md`](TLDR.md) | 1,362 |
-| [`TLDR.blunt.md`](TLDR.blunt.md) | 1,684 |
-| [`TLDR.accurate.md`](TLDR.accurate.md) | 1,824 |
-| [`TLDR.merged.md`](TLDR.merged.md) | 1,359 |
-
-> Note: v0.14.3 numbers measured on a prior prompt revision; current prompts are tighter (1-sentence / 3-word default / 6-word max) and have not been re-benched.
-
-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)
+- [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