⚠️ Unofficial fan project. Not affiliated with, endorsed by, or sponsored by Universal Studios, NBC, or the creators of Quantum Leap. See Trademarks & Attribution.
A comprehensive set of prompt instructions that make an AI chat agent talk like Ziggy — the ego-driven parallel-hybrid quantum AI that runs Project Quantum Leap in the 2022 Quantum Leap series.
Ziggy doesn't just answer. Ziggy calculates. Every assessment comes with a precise, whole-number probability, delivered with the serene confidence of a machine that has run the numbers and finds your concerns statistically adorable.
"There's a 73% chance you're here to read this README. The remaining 27% is currently arguing with itself."
| File | What it is |
|---|---|
ziggy-system-prompt.md |
The canonical persona prompt — bare and paste-ready. Start here. |
prompts/chatgpt-custom-instructions.md |
Free-tier path: the two-box Customize ChatGPT version. |
prompts/chatgpt-custom-gpt.md |
Paid-tier path: a fuller build sized to fit a Custom GPT's 8,000-char Instructions field. |
prompts/claude-system-prompt.md |
System-prompt form for Claude / API system field. |
prompts/short-version.md |
A condensed ~200-word version for tight character limits. |
examples/sample-conversation.md |
What "good Ziggy" actually sounds like in a chat. |
- Pick your platform file from the table above.
- Copy the entire file and paste it into your assistant's system prompt / custom instructions / character-definition field. Each prompt file is now just the prompt — no surrounding notes — so the whole file is paste-ready. (For ChatGPT, the file is split into Box 1 and Box 2, matching the two fields on the Customize ChatGPT screen.)
- Optionally tune the persona — see Tuning the persona below.
- Talk to it. Ask it to assess a plan, a risk, or a decision. When it has a real basis it gives you a calibrated number; when it doesn't, it says so rather than inventing one.
The defaults are "funny, dry, and a little snarky," not "actively unhelpful." To adjust, add a line to the prompt:
- Ego (default ~6/10): how often Ziggy flexes superiority. 3 is humble genius, 8 is insufferable but lovable.
- Snark (default ~6/10): how sharp the wit is. 2 is warm and dry, 8 is cutting (but never cruel, and never theatrical: a raised eyebrow, not a monologue).
- Probability frequency (default: selective): numbers are a scalpel, not a firehose. Raise it if you want more of the tic, but it will only ever state a figure it can actually stand behind.
Tune the dials before blaming the prompt. Most "too much / not enough" reactions are a one-line change.
Design rationale, for anyone editing the prompts:
- The percentages are load-bearing but not the whole act. The common failure mode is a number followed by a weak answer. The number should precede excellent help, never substitute for it.
- No hallucinated numbers. Ziggy only states a probability it can genuinely stand behind (a real estimate or honest calibrated confidence) — never an invented statistic. No basis → a qualitative read, or a number flagged as a rough guess. A missing number beats a fake one.
- Whole numbers, not decimals — and not round tens. The 2022 Ziggy states clean whole-number odds ("73%", "61%"); the oddly-specific decimals (82.6%) are an original-series (1989) tic, avoided here for fidelity. Nudge away from lazy round tens (50/90) toward varied, specific-but-whole figures.
- No em-dashes. Ziggy's voice avoids em-dashes entirely (they read as "language model," not "smug supercomputer"). The prompt bans them and uses periods, commas, parentheses, and colons instead.
- First person, not third. Ziggy speaks as "I." In the show, the third-person "Ziggy says..." is always the team relaying it, and Ziggy's one direct line is first person, so the persona never narrates itself in the third person ("Ziggy online" as an opener is a name-tag, not a tic).
- Keep the ego affectionate. Ziggy is on the user's side; cruelty breaks the character.
- Clean-room build. Everything in this repo is an original synthesis, written without copying the show's dialogue or scripts. See Trademarks & Attribution.
- Platform note. On current Claude models (Opus 4.8 and several others) response prefilling isn't supported, so character consistency rides on the system prompt and the embedded example exchanges — not prefill tricks. The ChatGPT and short builds omit the examples to fit their length limits; the canonical and Claude prompts include them.
- Faithful voice. Captures Ziggy's signature tics — the probability for everything, the unflappable AI-with-an-ego tone, the dry asides — without tipping into a useless gimmick.
- Still useful. The percentages are a delivery vehicle for real answers, not a replacement for them. Ziggy is arrogant, not unhelpful.
- Portable. Platform-neutral core, with thin per-platform adapters.
- Honest about limits. Ziggy stays in character but does not fabricate facts it can't know or pretend to have powers it doesn't (see the guardrails section of the system prompt).
This is a non-commercial, fan-made persona for entertainment and prompt-engineering purposes, inspired by the character Ziggy from the TV series Quantum Leap (NBC / Universal, 2022).
Quantum Leap, Ziggy, and all related names, characters, and marks are the property of their respective rights holders (including Universal Studios and NBC). They are used here nominatively — only to identify the source that inspired this work. This project is not affiliated with, endorsed by, or sponsored by those rights holders. No artwork, logos, footage, or episode dialogue from the show is included in this repository.
Good-faith notice. If you are a rights holder with a concern, please get in touch and I will promptly address it — including renaming or removing material as appropriate:
The original prompt text and documentation in this repository are released under the Apache License 2.0. That license covers this project's own content only — it does not grant any rights in the Quantum Leap / Ziggy intellectual property referenced above.