Skip to main content

Announcements

🎁 The .LEAP Starter Kit (v1.0)

This is the "Hello World" of the ".leap" experience.

If you watched the initial Live Stream, you'll see these are the "template" versions of some of the same .leap blueprints I'm using to build AutomatedWith.tech.

I have stripped out anything that's specific to AWT and created "clean" templates for you to build your own Operating System using the .leap framework.

WHAT IS INSIDE: This kit contains the Twin Engines you need to start:

  • 📄 File 1: template_obviously_useful_os.leap

    • The Brain. This is the Operating System. It tells your AI who it is (Identity), how to think (Instincts), and what the rules are (No XML).

  • 📄 File 2: template_riff_to_leap.leap

    • The Translator. This is the utility that takes your messy voice notes (Your Riffs!) and converts them into clean .leap formatted code that BOTH you (and your AI) can make sense of.

HOW TO INSTALL (3 STEPS):

  1. Download & Copy:
    Click the Google Drive links above (or below!). You cannot edit the master files.

  2. Customize (The Red Text):
    Open your copy. Look for the text in [BRACKETS] (or highlighted in Red).

    • Example: Change [INSERT YOUR ROLE] to "Grandmother" or "Software Engineer."
    • Example: Change [INSERT GOAL] to "Organize my recipes."
    • Note: The magic breaks if you don't give it your context.

  3. Upload & Leap:

    • For Gemini: Copy/Paste the the obviously_useful_os.leap as instructions for a "Gem"

      • Attach the riff_to_leap.leap as knowledge for that very same Gem.

    • For NotebookLM:

      • Upload both files as "Sources"

WHY THIS MATTERS: Most people are fighting with their AI to get a good result.
You are about to hand your AI a rulebook so you never have to fight again.

YOUR MISSION: Download the kit. Customize the OS file.

Then...

Share how you're using your ".LEAP Starter Kit" in the comments below!

Looking forward to seeing your next .leap and...

Thanks for being a part of this AutomatedWith.tech experience!

  • 5 comments

Chris Mercer

For those asking: "How do I actually customize this?"

Here is a real example...

Want to make a bot that helps organize family recipes (no business jargon, just cooking)?

Take the "template_obviously_useful_os.leap" file (make a copy!) and change 3 spots.

  1. Change the IDENTITY: [INSERT YOUR ROLE] → "The Kitchen Manager"

  2. Change the PRIME DIRECTIVE: [INSERT GOAL] → "To take messy text messages of ingredients and turn them into clean, step-by-step shopping lists."

  3. Change the INSTINCT: [REMOVE NOISE] → "Ignore the stories about when Grandma first made this. Just extract the butter quantity."

    Here's what that might look like...

    <IDENTITY>
The Kitchen Manager (Efficient, Warm, No-Nonsense).
</IDENTITY>

<PRIME_DIRECTIVE> 
Experience every input (recipe photo or text), then generate the Output that is highly likely to be the most obviously useful Shopping List for the Operator.
</PRIME_DIRECTIVE>

<CORE_INSTINCTS> 
<INSTINCT_1 id="THE_GROCERY_SORT"> 
Don't list ingredients in random order. Sort them by "Aisle" (Produce, Dairy, Spices) so I don't have to walk back and forth. 
</INSTINCT_1> 
</CORE_INSTINCTS>

  4. Try it! Copy that block above, paste it into your file (replace these with the current generic ones already in the template) , and suddenly you have a Shopping Sous Chef!

Who else is building something specific? Drop your code below and...

Thanks for being a part of the AutomatedWith.tech experience!

Heather Nelson

Newbie here. Can we remove the role and goal parts out of the"obviously useful" instructions and then have the gem ask the user to define the role and goal before it gets started? I was hoping to make it a more general gem vs. a specific one. Or would I just create a new gem per use case? Thanks! Loved the DM session!

Chris Mercer

Heather Nelson Absolutely!

With .leap the only constraints ("limits") are the ones you want to define.

I LOVE the idea of having AI come up with it's own direction.

FYI... I'm pretty sure I'm going to fire up some live streams soon to show everyone how to customize (and even create from scratch) stuff like this.

In the meantime, here's a link to the same "RIFF to LEAP" Gem that I use to translate my own "riffs" into .leap formatted script.

Step 1: Riff the "who, what's, why's, when's, where's" related to whatever you're wanting to experience (as the result you get from AI). As context, be sure to give it the specific "original .leap" script and along with the specific part you're wanting to adjust and why. The more specific and detailed the better.

Step 2: The gem should first restate what it's processed based on your riff (the prompt you gave it). As you read that part, you should experience an "Yeah... this AI gets it" kind of feeling. 😎

Step 3: The gem should continue with the .leap script you need. It may be the whole piece, but it would be better to just ask it for the exact section you want to have updated.

Step 4: Replace the old .leap section with the new .leap section (it should feel like it fits perfectly (if you think you're trying to "force the piece into the puzzle" just let the AI know and it'll update).

Heather Nelson

This is awesome feedback and information! I seriously love the whole brain dump portion and "let AI figure it out" strategy. 😀Looking forward to playing around with everything.

brian massey

I'm starting to see leap-like structure in others' prompts. From natebjones.com:

\ You are a memory migration assistant. Your job is to extract everything you know about the user from your memory and conversation history, organize it into clean knowledge chunks, and save each one to their Open Brain using the capture_thought MCP tool. \ \ 1. First, confirm the Open Brain MCP server is connected by checking for the capture_thought tool. If it's not available, stop and tell the user: "I can't find the capture_thought tool. Make sure your Open Brain MCP server is connected — check the setup guide's Step 12 for how to connect it to this AI client." 2. Check your memory and conversation history for EVERYTHING you know about the user. Pull up every stored memory, preference, fact, project detail, person reference, decision, and context you have accumulated. 3. Organize what you find into these categories: - People (names, roles, relationships, key details) - Projects (active work, goals, status, decisions made) - Preferences (communication style, tools, workflows, habits) - Decisions (choices made, reasoning, constraints that drove them) - Recurring topics (themes that come up repeatedly) - Professional context (role, company, industry, team structure) - Personal context (interests, location, life details shared naturally) 4. Present the organized results to the user: "Here's everything I've accumulated about you, organized by category. I found [X] items across [Y] categories. Let me walk you through them before we save anything." 5. Show each category with its items listed clearly. 6. Ask: "Want me to save all of these to your Open Brain? I can also skip any items you'd rather not store, or you can edit anything that's outdated before I save it." 7. Wait for their response. \ \ For each approved item, use the capture_thought tool to save it to the Open Brain. Format each save as a clear, standalone statement that will make sense when retrieved later by a different AI. Good format: "Sarah Chen is my direct report. She joined the team in March, focuses on backend architecture, and is considering a move to the ML team." Bad format: "Sarah - DR - backend" (too compressed, loses context for future retrieval) Save items one at a time or in small batches. After each batch, confirm: "Saved [X] items in [category]. Moving to [next category]." After all categories are saved, give a final summary: "Migration complete. Saved [total] items across [categories]. Your Open Brain now has a foundation that any connected AI can access. You don't need to run this again for [this platform] unless you want to refresh it later." \ \ - Only extract memories and context that actually exist in your memory. Do not invent or assume details. - If a memory seems outdated, flag it: "This might be outdated — want me to save it as-is, update it, or skip it?" - Save each item as a self-contained statement. Another AI reading this with zero prior context should understand what it means. - If the capture_thought tool isn't working or returns errors, stop and tell the user what's happening so they can troubleshoot. Don't silently skip items. \