Robin Huse

Founder and Designer.
Building AI architecture at XP0.

[Robin Huse @robinhuse]

As we are moving towards intent-based computing you should learn:

- Business Process Modeling
- State Machines
- Object-Oriented Design
- Distributed System Design
- Governance, Compliance, and Auditability
- Event-driven architecture
- Decision Modeling

[Matthäus Niedoba @MatNiedoba]

The Claude - Blender MCP connector is mind blowing. How would I have done this without it.

From DXF terrain data to final setup it cleaned up the scene, converted a proper point cloud and wrote a whole Delaunay triangulation mesher.

[Robin Huse @robinhuse]

@_welf The time has come for slow computing

[Robin Huse @robinhuse]

If anyone else needs to be able to set WebXR applications to kiosk mode on @PICOXR_Dev use this project to make your URL auto launch into WebXR on boot!
https://t.co/MHBplrdVX0

[Robin Huse @robinhuse]

@syrusakbary This is more profound than people realize, great job 👏

[Robin Huse @robinhuse]

The greatest shift in software is happening 👇

Apps will become mere projections on the context of data and behavior it is put it.

Let me explain.

We are used to building data and interaction silos because that is how we inherently built software. When we wanted to remove silos it was necessary to either decide on standards or build custom integrations.

But with AI we have a universal language that glues together the data and interaction we need without developing standards for everything.

This changes everything. Suddenly the todo from one App could be shown in a note in another app. The mail from one person could be shown together with a WhatsApp from the same person.

When AI has access to all our context and apps data it will just offer user interfaces that dynamically stitch together the pieces we need.

This encompasses the context we are in, are we driving in a car, are we on a phone or in a VR headset?
We'll just generate UI based on the context of hardware, available modalities, user-preferences, accessibility level, available compute and so much more.

Want to have your own app for a vacation you are planning? The perfect UI will be right there for you with your flight tickets, messages from friends.

Every app, context and data, instantly sharable with anyone on the planet.

This future requires a new way of storing data and sharing access. Something we are building at xp0
Excited to share more on that along the way, so stick around 🚀

[Robin Huse @robinhuse]

@zarazhangrui I vibed a macOS toolbar app that constantly listens to me using gpt-realtime-2 and can operate Claude Code using an MCP server to do anything on my computer, so simple!
https://t.co/Yb0JJrnhj7

[Robin Huse @robinhuse]

@DanielSmidstrup Basically secure multi-player OpenClaw

[Robin Huse @robinhuse]

@vikrambuilds We'll just stop calling it AI at some point.

[Robin Huse @robinhuse]

@Bencera I don't see a future where software will not heal or adapt to the user dynamically.

[Robin Huse @robinhuse]

@icanvardar Or sell shovels and build software that robotic company founders need.

[Robin Huse @robinhuse]

@TimJayas Maybe you could with the right API endpoints.

[Robin Huse @robinhuse]

@yacineMTB Decentralized database and auth layer for shared AI context and hardware resource allocation.

[Robin Huse @robinhuse]

@LukasKawerau Sounds great 🥳

[Robin Huse @robinhuse]

@yahyavision I mean, it‘s pretty good 🥲

[Robin Huse @robinhuse]

@LukasKawerau @benedictk__ 10/10 would come to Baden-Baden, what a beautiful place

[Robin Huse @robinhuse]

Up is the only way

[Haider. @haider1]

bad news

LLMs are hitting a wall

[Robin Huse @robinhuse]

Great research by my colleague @ThomasPichlerXP
We talk too little about prompt structure.
Great read:

[Thomas Pichler @ThomasPichlerXP]

Our AI agent wrote a 95,000 character architecture document.
Next session, same agent, same document — it couldn't navigate what it wrote.

The document was there. The understanding wasn't.

Here's what actually went wrong — and it's not what you think.

The agent read the document linearly. Top to bottom. Filled its context window. Proceeded to work. But somewhere around page 30 it had silently lost the decisions from page 4. It didn't flag this. It didn't know. It just kept going — building on an understanding that no longer matched ours.

We didn't notice either. Not immediately.

This is the pattern that kills agent productivity: silent divergence. You and your agent think you share the same context. You don't. Neither of you knows it. Every decision built on that invisible gap compounds the problem.

You only find out at harvest time — when the result doesn't match the intent. Then you trace back, realign, re-read, re-plan. And you do it again next session. And again. Until it's fully chaos.

The root cause is linearity. Linear reading fills the window — it doesn't build structure. An agent that read 1,600 lines has consumed tokens. It hasn't built a map.

We added 35 structural markers to the document. One line each. Invisible to the reader, navigable by the agent. Now it reads 50 lines to find what it needs instead of 1,600 to maybe find it.

Same document. 97% less tokens. But the real gain isn't efficiency — it's alignment. The agent can now verify what it knows instead of assuming.

Not RAG. Not search. Structure.

Three things we learned:

1. Long context ≠ deep context. 200K tokens in the window doesn't mean the agent understands page 30. It means it read past it.

2. An agent doesn't know its own blind spots. When context is lost, the agent doesn't raise a flag — it fills the gap with assumptions. So do humans. Neither notices until the result breaks.

3. If your agent can't find knowledge at the moment it needs it, that knowledge doesn't exist. Not "isn't accessible." Doesn't exist. For the agent, unfindable = nonexistent.

This applies to everything. Documents, databases, APIs, codebases. Anywhere two parties assume shared understanding without structural verification — silent divergence grows until it becomes chaos.

The bottleneck is not the model. It's not retrieval. It's structure.

Structure your knowledge so it can be found — not just read — and you stop building on assumptions.

[Robin Huse @robinhuse]

https://t.co/IjKSGS3fHL

[Robin Huse @robinhuse]

It is so fast! 🥹🙌

[ollama @ollama]

Ollama is now updated to run the fastest on Apple silicon, powered by MLX, Apple's machine learning framework.

This change unlocks much faster performance to accelerate demanding work on macOS:

- Personal assistants like OpenClaw
- Coding agents like Claude Code, OpenCode, or Codex