Stenographer Mode logo

Launch Preview

Stenographer Mode

Honest Token Study

Stenographer mode compresses well for technical work, not universally.

Demo compares baseline prose, caveman mode, and stenographer mode across 50+ samples in 7 categories. Includes failure cases where steno underperforms. Token counts via gpt-tokenizer.

Best for: code review, debugging, docs Weaker for: onboarding, stakeholder

Primary command: /steno

Baseline full prose Caveman semantic cut Steno shorthand map Compression strategy drop filler -> standardize shorthand -> preserve exact literals

Samples tested

0

across 7 categories

Caveman reduction

0%

vs baseline

Steno reduction

0%

vs baseline

Steno beats caveman

0%

of samples

Category Breakdown

Where steno helps and where it doesn't.

Release Channel

preview

Productized bundle with install and export flow.

Platforms

0

Current starter-pack coverage.

Bundle Status

ready

Installer, prompt, skill, demo, packs.

Install Surface

End-to-end product flow: bundle, install, activate, export.

1. Install

Run npx --yes github:AkashAi7/stenographer-mode install --scope user for a user-wide install, or switch to --scope project for .github/prompts.

2. Activate

Use /steno in prompt picker or trigger it with “steno mode” phrasing. For context conversion, use /steno-compress temporary or /steno-compress permanent.

3. Export

Install the CLI globally with npm install -g github:AkashAi7/stenographer-mode, or run install/export-pack.ps1 to generate a shareable distribution bundle.

How it works

Steno mode uses controlled shorthand, not raw truncation.

1

Trim

Remove filler, social framing, and repeated setup.

2

Map

Swap repeated concepts to stable abbreviations like cfg, deps, and ctx.

3

Link

Use symbols like ->, w/, and vs to compress relations.

4

Protect

Leave commands, code, paths, exact errors, versions, and identifiers untouched.

Mode Comparison

Baseline vs caveman vs stenographer

Infographic

Exact token profile

Exact benchmark via gpt-tokenizer. Artifact: benchmarks/latest.json.

Distribution

Caveman-style reach needs platform packs, not one universal file.

Ready now

VS Code skill, VS Code user prompt, local caveman comparator, installer scripts, static demo, product manifest, Claude pack, Cursor pack, ChatGPT pack.

How caveman scales

Same concept, multiple platform-native wrappers. Each platform gets its own instruction surface and trigger syntax.

Next product step

Create platform packs for Claude, Cursor, ChatGPT, and web docs so activation feels native everywhere.

Release Kit

Core artifacts for shipping, installing, and sharing.

Install

install/install.ps1

Bundle

bundle.json

Release

release.json

Prompt

bundles/vscode/steno.prompt.md