In closed testingBuilt in the open

Ask in one line. Inspect every layer.

MyPromptVerse is a workspace for AI work you do more than once. You ask in plain words. Underneath, every run keeps the operator, playbook and knowledge it used, and is checked against the playbook or marked Not checked.

Follow the build See the library

20 records in the Library today, free under CC BY 4.0.

An example run, taken apart. The operator and the playbook are real records from the Library. Choose a layer to look inside.

01 The problem

Three weeks later: “What did we use last time?”

Chat is good for answers and bad at keeping the method. The context you explained, the approach that worked and the output worth keeping all sink into the history. MyPromptVerse keeps them as things you can reuse and inspect.

The context you explainedkept as an operator
The approach that workedkept as a playbook
The output worth keepingkept as knowledge

You · 3 weeks agoWhat did we use for the audit last time?Kept as a playbook

You · yesterdayPasting the refund policy again, sorryKept as knowledge

You · 2 days agoRemember: never invent a customer quoteKept as an operator

You · this morningTurn these notes into actions. Owners and dates.Kept as a prompt

You · last monthSame structure as the Q2 deck, pleaseKept as a playbook

You · 4 days agoYou're the analyst again. Cite what you opened.Kept as an operator

You · last weekHere's our brand voice doc, one more timeKept as knowledge

You · 2 weeks agoFill in client, date and tone, then draft itKept as a prompt

You · 5 days agoWhich version did the client approve?Kept as knowledge

02 The Studio

Prompts become operators. Operators run playbooks. The work compounds.

Four things you build once and use on every job.

What goes where

Operator³A reusable role: who the model is, what it is for, and the rules it keeps on every job.“Ask for success criteria and audience when the brief leaves them out, and list the assumption instead of guessing.”From Prompt Engineer Playbook¹¹A reusable method: the inputs it needs, how it works, and the exact shape of what comes back.“Summary · Ranked issues · Proof gaps · Rewrite · Priority checklist”From Landing Page Audit Prompt⁵A ready-to-run task with fill-in variables, or a snippet that sets tone and structure.“Never invent an owner or a date. Write [UNASSIGNED] or [MISSING DATE] instead.”From Meeting Notes to Action Plan Knowledge¹The material a run should work from: rubrics, policies, house rules, past findings.“Priorities P1 to P4, refund rules, escalation contacts: fill in yours once, use it on every ticket.”From Support Policy Template

03 What comes out

The deliverable, with its method still attached.

Example runs, labelled as examples: the job, the context it used, and what became reusable.

Example runReviewed

Landing page audit

Seven fixes grouped by impact and effort, each tied to the evidence on the page.

JobAudit a landing page
ContextScreenshot + brand context
RunPrompt snapshot · recorded

P1Clarify the hero promise
P1Separate preview and sign-in paths
P2Cut the mobile explanation depth

04 On the bench

Built in the open, in closed testing.

What works today, what we are building, and what is not built yet. No roadmap theatre.

Shipped ³

Oct 2026Publish Studio records to a public Library
Oct 2026Honest run checks: replies are checked against the playbook, or marked Not checked
Oct 2026Prompt Library, Prompt Builder and Prompt Improver

Building ²

Team runs: several specialist steps inside one budgeted run
Retrieval for large knowledge sources

Not yet ²

Scheduled automations
Shared workspaces for teams

From the notes

All notes

Newsletter

Get the workflow notes behind the build.

Practical prompts, operator ideas, playbook patterns and product notes, plus new Library records as they are published.