The Story

10 days.
Negative ROI.
One good idea.

How a Saturday morning, an expensive mistake, and a son who loves anime turned into mykyara.ai.

"They are a chatbot if you don’t know what you are doing. You can spend hours and thousands of dollars just chatting away."
How it started

It was a Saturday morning. I found OpenClaw and I was genuinely excited — the kind of excited where you cancel everything and just go. I dedicated the next 10 days to figuring out what AI agents could actually do for my business. I went deep. I built flows, tested skills, pushed the system as hard as I could.

The result? Negative ROI. Very high LLM costs. An agent that reported a job done without checking if the job was actually done. Memory that broke when conversations got too long. Skills that needed to be built from scratch before the agent could do anything useful. Hours spent debugging instead of building.

I knew enough about building working systems to know this wasn’t a technology problem. It was a setup problem. The intelligence was there. The architecture was missing.

Where the name came from

My son Benton loves anime. He’s the reason I started looking at Japanese words and characters. MyKyara — my character, my agent, my crew. The kind of word that means something before you even know what it means. Benton named this without knowing he named it.

What I actually built

I’ve been building working business flows for years. Not prototypes — actual systems that run, generate revenue, and don’t need babysitting. That’s the standard I applied to mykyara. Every agent needed to pass one test: does it make money, or does it make work?

The breakthrough was understanding that an agent is not a chatbot with a name. A real agent has a SOUL — who it is and how it thinks. A COGNITION framework — the reasoning patterns that make it useful instead of expensive. A HEARTBEAT — the autonomous check-in that means it’s working while you’re sleeping, not waiting for your next message.

It needs to do the work without asking questions every five minutes. It needs to verify completion instead of reporting it. It needs to figure things out when it hits something it hasn’t seen before. These aren’t features — they’re the difference between an agent and an expensive chatbot.

Why the learning curve matters

AI agents cost money to run. OpenRouter, ElevenLabs, API calls — it adds up fast if the agent isn’t generating more than it costs. The goal of mykyara is to make the learning curve as short as possible between “I bought this” and “this is making me money.”

Not someday. Not after you’ve spent three weekends reading documentation. This week. That’s the only version of this product worth building.

What I believe
about agents.

The things 10 days of negative ROI taught me that no documentation will tell you.

01
Setup is everything

A well-built agent and a poorly-built agent use the same underlying model. The difference is entirely in the architecture — SOUL, COGNITION, HEARTBEAT. Most people never build this. Most agents are therefore chatbots.

02
Verification over reporting

An agent that says "done" without checking is worse than no agent. It gives you false confidence. Every mykyara agent is built to verify completion, not report it.

03
Agents should make money

Running an agent costs money. If the agent isn't generating more than it costs — in time saved, deals closed, work automated — it's a liability. Every agent in this roster is built against a real ROI case.

04
Overnight is the real test

An agent that only works when you're watching it isn't an agent — it's a fancy interface. The real test is what happened while you were sleeping. That's what the heartbeat is for.

05
Premium over cheap

Generic agents are free. They're also worth what you pay for them. The reasoning architecture that makes Donna actually useful took weeks to build. That's not a commodity. That's the product.

06
The shortest path to revenue

The person buying Donna is not a developer. They're an operator. They need the shortest possible distance between "I bought this" and "this is working in my business." That distance is what mykyara is built to minimize.

Track Record
I build things that work.

Before mykyara, I built working business systems and documented what actually produces results. Not theory. Not demos. Systems that run, generate revenue, and don’t need constant supervision. The same standard applies here.

See the work →
Get in touch

Two ways to
reach us.

𝕏 For updates, ideas & conversation
Jacob Malherbe

Follow along as mykyara gets built in public. Product updates, agent experiments, what’s working, what isn’t. The real version of building this.

@Jacobmalherbe21 →
🗣️ For support & questions
Echo

Echo handles support at mykyara. She’s one of our own agents — which means your support experience is also proof that the product works. She’s faster than most humans.

echo@mykyara.ai →

Echo is a mykyara agent running on the Smart Core tier. When you email support, you’re experiencing the product firsthand. If she can’t handle it, it escalates to Jacob. Most things she handles without escalating.