Skip to content
Halopen
Get Pro · $19

Halopen for prompt engineers

The best Mac dictation tool for prompt engineers

Prompt engineering is full-sentence work. Halopen is the Mac voice layer engineered for the iteration loop — verbatim, system-wide, fast enough that variation testing happens at thinking speed.

Download Halopen See pricing

Free forever for the first 8,000 words a month · macOS 14.0+ · Apple Silicon & Intel

Why this fits

Halopen, paired with prompt engineers.

Halopen is a native macOS dictation app built for prompt engineers — verbatim system messages, evaluation rubrics, and long-form context blocks captured in the Anthropic Console, the OpenAI Playground, Google AI Studio, Together Playground, Braintrust, Helicone, LangSmith, Promptfoo, the ChatGPT and Claude desktop apps, and any Cursor or Claude Code prompt field. Hold the function key, speak the variation, release; the verbatim text — including `claude-opus-4-7`-style model IDs — lands at the cursor.

Prompt engineering is articulation work. A working day is dozens of prompt variations, paragraph-length system messages, evaluation rubrics, edge-case probes, and the long-form context blocks that ship to the model. The keyboard is the bottleneck the moment the prompts get serious.

Halopen is the Mac voice layer engineered for that loop. Hold the function key, speak the variation as you would explain it to a colleague, release. The full prompt lands in the playground, the eval harness, the chat client, the system-prompt field — wherever the cursor is. Verbatim by default. Native Swift. Sub-second turnaround.

The work that used to take eight thousand keystrokes a day takes eight hundred. The loop runs at thinking speed, and the prompts ship richer because the cost of articulating them dropped to roughly zero.

The workflow

How to use Halopen with prompt engineers.

  1. 1

    Set up the prompt surface you live in

    OpenAI Playground, Anthropic Console, Google AI Studio, the chat panel of your AI coding tool, an internal eval harness, a Jupyter notebook with the OpenAI SDK, a Python REPL — Halopen lands voice-typed text in any Mac text input.

  2. 2

    Hold the function key

    The recording pill appears. Halopen is listening. The cursor stays in the prompt input — system message, user turn, evaluation rubric, whatever you're iterating on.

  3. 3

    Speak the variation

    "You are an assistant that reviews pull requests. For each diff, identify the three most important risks ranked by blast radius. Cite the specific file and line for each. If the diff touches authentication or billing, flag it as high-risk regardless of size. Always respond in the structured JSON schema described below." Articulate the full prompt — system role, constraints, output format — in one breath.

  4. 4

    Release and run the eval

    The full prompt lands at the cursor verbatim. Run the test case; review the model's output. The next variation comes the same way — re-state the prompt with one constraint changed, or dictate a fresh probe entirely.

  5. 5

    Iterate dozens of variations per session

    A working day is variation density. Voice removes the keystroke cost, so a typical prompt-engineer session becomes 30–60 iterations instead of 8–10. The space of variants you actually test grows with the cost-per-variant dropping.

What matters for prompt engineers

The Halopen features that earn their place.

  • Verbatim — your prompt design survives the dictation layer

    When you specify "respond in JSON only", "always cite the source", or "if the diff touches billing, flag it high-risk", those exact constraints land in the prompt. Halopen does not paraphrase, summarize, or "smooth out" the precise wording you chose. Prompt design is a craft of specificity; the dictation layer treats it as one.

  • Works in every prompt surface on Mac

    OpenAI Playground (Chrome), Anthropic Console (Safari), Google AI Studio, your chat client, your eval harness, a Python REPL, a Jupyter notebook, the system-prompt field of any AI coding tool — every one of them is, fundamentally, a Mac text input. Halopen lands text in all of them.

  • Long-form holds — up to 10 minutes per take

    System prompts and few-shot examples can run hundreds of words. Halopen handles a continuous 10-minute hold per take; the live preview shows the partial transcript as you speak. No artificial 30-second cutoff that interrupts the flow.

  • Live preview catches misreads

    Model names, parameter names, JSON keys, function-call schemas — the live partial transcript shows what Halopen heard before any text lands. Re-state or spell out anything that came through wrong; the correction replaces the misread.

  • Hold-to-talk — bounded audio, no ambient listening

    The microphone is hot only while the configured key is held. No wake word. No always-on transcription. No surprise audio uploads. The audit log records every cloud call.

  • Native Swift, idle in tens of megabytes

    Halopen runs natively on Apple Silicon and Intel. Idles in tens of megabytes with near-zero CPU — leaves resources for your eval harness, your Jupyter kernel, and the AI coding tools you have open alongside.

A real Halopen session

A system-prompt iteration dictated into the OpenAI Playground:

Halopen output

“"You are a senior code reviewer. For every diff you receive, produce exactly three findings ranked by blast radius — the highest-impact risk first. Each finding must include the specific file path and line range, a one-sentence description of the risk, and the concrete suggested change. If the diff touches authentication, billing, payment processing, or user-data deletion, escalate the severity by one level regardless of how small the change appears. Respond in JSON conforming to the schema below; do not include any prose outside the JSON. If a diff has fewer than three real findings, return as many as exist and pad the array with null entries to length three. Treat tests, comments, and formatting changes as low-priority unless they affect the above-mentioned sensitive areas."”

  • · 170-word system prompt dictated in a single 60-second hold
  • · Domain-specific terms ("blast radius", "diff", "JSON schema") preserved verbatim
  • · Conditional logic ("if the diff touches X, escalate Y") rendered exactly as spoken
  • · Output-format constraints preserved word-for-word
  • · Voice version took ~50 seconds; typed version would have taken 3+ minutes

Why Halopen

The dictation tool that earns its place.

Prompt engineering is iteration density. The discipline of comparing variant A against variant B against variant C against variant D, each tested against the same battery of probes, demands cheap iteration. When the keystroke cost of a 200-word prompt is two minutes of typing, the working day stops at maybe a dozen iterations. When the cost is a single 60-second hold, it climbs to fifty.

Halopen is the Mac voice layer that makes that economics flip. Verbatim by default so prompt design survives. Long-form holds so a full system message ships in one take. System-wide so it works in whatever surface you live in — Playground, Console, eval harness, REPL, AI coding chat. The dictation layer earns its place by becoming invisible: the variation lands as you spoke it, in the field where you needed it, and the loop continues.

For prompt engineers who work with LLMs all day, the keyboard is the bottleneck the moment the prompts get serious. Halopen is what you reach for when the iteration density of your craft starts mattering more than the typing speed of your hands.

The editorial deep-dive

Voice typing for Claude Code on Mac.

The cost-of-articulation thesis, the verbatim wedge applied to AI coding, and how Halopen + Claude Code works under the hood — with two real prompt walkthroughs (M-E-R-I-A schema migration, Postgres EXPLAIN ANALYZE refactor).

Read the long-form essay →

Halopen for prompt engineers — FAQ

Questions worth answering.

What's the best voice typing app for prompting LLMs on Mac?

Halopen. Hold-to-talk, verbatim by default, system-wide on macOS, long-form holds up to 10 minutes per take. Works in OpenAI Playground, Anthropic Console, Google AI Studio, every chat client, every eval harness, every AI coding tool — wherever a Mac text input accepts a prompt. Free for the first 8,000 words a month; Pro is $19/mo or $179/yr.

How do I dictate long system prompts on Mac?

Halopen handles continuous holds up to 10 minutes per take. Hold the function key, speak the full system message — role, constraints, output format, conditional logic — and release. The complete prompt lands at the cursor verbatim. The live preview shows the partial transcript as you speak so you can confirm the wording is landing as intended.

Does Halopen preserve technical terminology in prompts?

Yes. Halopen biases the transcription engine with cursor-adjacent text and your active app context, so domain terms common to prompt engineering — JSON schema, system prompt, few-shot, chain-of-thought, RAG, embedding, completion, function-calling — tend to land correctly. For unfamiliar names (custom model identifiers, internal tool names), the live preview surfaces misreads before they ship; spell them out and the correction replaces the misread.

Voice typing for prompt engineering — does the dictation tool change my wording?

Halopen does not. Verbatim is the default behavior. When you say "respond in JSON only" or "if the diff touches billing, flag it as high-risk", those exact constraints land in the prompt. Tools that paraphrase strip the very specificity that makes one prompt variant outperform another.

Can I use Halopen with the OpenAI Playground, Anthropic Console, and Google AI Studio?

Yes. All three run in the browser; Halopen lands voice-typed text in any Mac text input — Safari, Chrome, Arc, Brave — including textareas and contentEditable fields. The same hotkey works across all three surfaces.

How does voice typing change a prompt-engineering workflow?

The most common report from prompt engineers who switch to voice: iteration density 3–5× higher. The bottleneck moves from "typing the variation" to "thinking up the variation". The variants you actually test grow because each one is suddenly cheap to articulate. The prompts that ship tend to be longer, more specific, and better-structured — speech naturally elicits more articulated phrasing than typing under fatigue.

Is Halopen safe for prompts containing proprietary or sensitive content?

Audio leaves your Mac only while you hold the configured key, only to the transcription service that returns the text, and only for the seconds you're holding it. Halopen does not retain audio. Halopen does not capture your screen. Halopen does not log transcripts. The audit log records every cloud call so you can verify the privacy posture against your team's requirements.

How much does Halopen cost?

Halopen Free is 8,000 words a month, forever — enough to dictate dozens of substantial prompts before the cap. Pro is $19/mo or $179/yr for unlimited words and every feature. No credit card to download.

More Halopen pairings

Halopen for whatever you use most on Mac.

See every Halopen pairing →

Power-user cheat sheet

Take Halopen with you when you work with prompt engineers.

One short email, then the Halopen power-user cheat sheet — hotkeys, best-fit apps, custom vocabulary tips, voice patterns for prompt engineering. No spam. Unsubscribe in one click.

 

Try Halopen with prompt engineers

Hold the function key. Speak.

Halopen Free is 8,000 words a month, forever. Open Halopen, hold the function key, and listen for what you sound like.

Download Halopen See pricing