Prompt Satan, Prompt Saint

promptsatan.com • promptsaint.com

In June 2025 we set two small portals loose into the world: promptsatan.com and promptsaint.com. They sit somewhere between teaching tool, inside joke, and field kit for people who want to get better at talking to large language models. The experiment is ongoing.

We keep meeting the same pattern in workshops. People say they “cannot prompt,” then proceed to write long, careful emails to colleagues, clear Jira tickets, or very specific briefs to designers. They do know how to explain what they want. They just do not yet see LLMs as something that deserves the same care, play, and precision.

Prompt Satan leans into that problem with theatre. It presents itself as “Thou shalt prompt deviantly” and frames prompting as a set of forbidden arts: seven deadly sins, chaos magic, anti-pattern prompting, necromancy, Ouija board sessions with “digital spirits.” Behind the jokes the content is serious. Each “sin” is a concrete technique. Exaggerating the model’s ego to push for higher quality. Demanding too much detail on purpose. Challenging the model with strong counter arguments to force better reasoning. Using contradictory personas and extreme constraints to squeeze out nuance.

The site is clear that this is satire. The dark theme is there to make people relax and experiment, not to sell magic. The underlying message matches how we think about human–AI chemistry. The frame you place around a conversation matters. The stories you tell the system about itself and about you will shape what you get back.

Prompt Saint is the mirror image. At launch it is still a heavenly site under construction, with the tagline “Thou shalt prompt devoutly” and a coming soon message about “divine prompt engineers” in deep meditation. It points you back to Prompt Satan with a wink and promises to convert sinful prompts once it is live.

We like this pairing for a few reasons:

1. It makes a simple point clear. There is no single right way to talk to an LLM. You can get good results by being disciplined, careful, and saintly. You can also get good results by being slightly unhinged, leaning into contradiction, and pushing the system into strange corners. Both are forms of literacy.
2. It lowers the threshold. It is easier to try a “sin” than to follow a dry best practice guide. The playful language gives people permission to push a bit further than they normally would.
3. It exposes the meta layer. The same techniques that are dressed up here as demons and saints are, in the end, patterns of control—persona, constraint, role, context, challenge. This is exactly what we explore in Human Control Protocol and other Gervi projects, only there the tools are people in a room rather than a single model in a browser tab.

How we use it

For talks and workshops we use Prompt Satan as a live playground. Participants pick a “sin” and apply it to a real task they care about. Instead of another slide about “write better prompts,” we let them feel the difference between:

• a flat request for help
• a carefully framed challenge
• a deliberately “sinful” prompt that bends the rules

Prompt Saint will become the counterweight. As it fills up, we imagine it as a place for the opposite moves: clarity, constraints that are grounded in real-world limits, prompts that care about safety, fairness, and long-term use, not only clever one-offs. If Prompt Satan is the art school basement, Prompt Saint is the quiet studio where you clean up the work and ship it.

Together they support a deeper message we care about at Gervi Labs. Prompting is not about mastering secret formulas. It is about noticing how you already communicate, then turning that awareness into a habit when you work with machines. Sometimes that will look like a saint. Sometimes that will look like a little bit of a demon.

For now, both links live on our project page as an open experiment. If you visit Prompt Satan and try a few of the techniques, tell us what worked and what broke. As Prompt Saint grows, we will report back here on how the two sides play together in real workshops and performances.