If you run a small business, there's a good chance you use more than one AI. You draft in one, research in another, and keep a third around because it's better at something the other two aren't. That isn't indecision. Claude, ChatGPT, Gemini and Grok are each best at something different, and the leader changes every few months. A small team that wants the best tool for each job ends up using all of them.

The cost of that is the part nobody talks about. Each of those apps knows your business only as far as you've told it, and what you've told one doesn't carry to the next. The pricing decision you worked through in one app on Tuesday isn't there when you open another on Wednesday. The plan lives in your head, and you explain it again — patiently — to a very capable stranger.

We built Idiosynt for that gap. The way we think about it: the platforms are the internet, and your wits are the application layer.

What the internet analogy means

The internet became valuable when applications were built on top of it. Nobody uses the protocols directly — they use email, maps, their bank's website. The AI platforms are that layer now: capable infrastructure that most small teams already pay for. Idiosynt doesn't compete with it. It's an application built on it.

A wit is a named colleague with a point of view — a CFO, a strategist, an operator — seated to your business. They hold one memory of your business that belongs to you, and a record of what they told you. And they show up in the AI apps you already use — the same wit, with the same memory — instead of asking you to move into a new one.

That last part is why the analogy holds. An application that only worked in one browser would be a curiosity. Wits are platform-agnostic by design, built on the open protocol the major AI platforms now support for connecting AI apps to outside services — the Model Context Protocol, which became a founding project of the Linux Foundation's Agentic AI Foundation in December 2025.1 The same wit can be there wherever you work; the connections we've verified so far are listed at the end.

Three roles, and none of them replaces another

Here's how a day goes when you have wits. You're in the AI app you like for research. You ask it to dig into a supplier's new terms. It does what it's excellent at: reads, compares, summarizes, drafts. Then you hand what it found to your wit — with your consent, and only what you choose to share. The wit weighs it against your business: your margins, the decision you made about this supplier in March, the commitment you made to a customer in June. They tell you what they think. Then you decide.

Three roles. The platform researches and orchestrates. The wit judges and recommends. You rule. Nobody is replaced, and you are the human in the loop at every step.

It's also why this works for a business of one or two people. The research happens in the app you already pay for. The judgment happens in a colleague who knows your business. The decision stays yours. A small team gets the division of labor a large one takes for granted — without hiring for it.

Good for you, good for the platforms

We want to be plain about something, because it shapes how we build. We build with the AI platforms, not against them. A wit runs on their models. You bring them research from the apps you already pay for. Every conversation you have with a wit sends the platforms more work, not less — the way every application ever built on the internet made the internet more useful.

What the platforms aren't built to do — and it isn't any one platform's job — is carry your business across all of them. That's not a failing. It's what the layer beneath looks like. Carrying your business across the apps you use, and keeping it yours, is what the application layer is for.

From moving memory to continuing it

The platforms also offer ways to move context. Claude, for example, documents both importing memory from other services and exporting it for use elsewhere.2 Import and export let you carry a snapshot from one place to another. They are useful when you change tools or want a copy of what an app remembers.

What we are aiming to provide with Idiosynt is fluid continuity of memory while you keep using more than one AI app. You work through a decision with Morgan in one supported app, then return to Morgan in another. The conversation continues from the same business memory and decision record, including what changed along the way. You should not have to repeat an export-and-import routine every time you change where you work.

That is the job we give the application layer: the same wit, with a memory that develops as you work together, available through supported connections. Import and export remain part of your ownership of that memory. Continuity is what we want the everyday experience to feel like.

What's true today

Today, wits live in the Boardroom at idiosynt.com and reach the Claude desktop app. The next steps, in order: Claude on mobile, ChatGPT, then the others. We'll say a platform is ready only when we've watched a wit work there, on a real device, twice — and we'll keep this list honest as it grows.

The platforms are the internet. Your wits are the application layer.

If you're juggling more than one AI and explaining your business from scratch every time you switch — that's the problem wits solve. Register at idiosynt.com and seat your first wit today.

Sources

  1. Linux Foundation Announces the Formation of the Agentic AI Foundation, December 9, 2025. ↩

  2. Import and export your memory from Claude, checked October 1, 2026. ↩