Crypto Gloom

Meet3D Founder Returns to AI-Powered OpenSim Grid – Hypergrid Business

Meet3D Founder Returns to AI-Powered OpenSim Grid – Hypergrid Business
Snapshot of the ThinkSim grid. (Image credit: Fabio Bastos.)

From 2010 to 2015, I ran Meet3D, one of the world’s largest OpenSimulator grids with 309 regions. Educational institutions, businesses, and communities have used it for educational simulations, virtual events, and real-time collaboration. It was one of the most active grids outside of Second Life, and building it taught me everything I know about large-scale distributed infrastructure.

I closed this company in 2015 and moved on to other things. I founded Sedina in Uruguay. It is the first company to legally commercialize cannabis-derived products in Latin America. I built a commodity trading business in Brazil. I moved to China. I eventually ended up in Thailand, where Sedina now operates in the legal cannabis market. And somewhere in between, I started thinking about AI governance.

I have never actually left the virtual world. We just took a detour through hemp, soy, and artificial intelligence.

Snapshot of the ThinkSim grid. (Image credit: Fabio Bastos.)

Currently, I run ThinkNEO, an enterprise AI control plane company based in Hong Kong. We are also making FOFO), a 100% open source domestically produced AI robot. And I’m back in the virtual world. This time we’re back with ThinkSim, something we couldn’t make in 2010.

ThinkSim is the successor to Meet3D. But that’s also something completely different. As far as I know, this is the most technologically advanced OpenSimulator grid operating in the world today.

You can visit the Grid via HyperGrid. thinksim.space:9000 Or create an avatar and log in directly.

Visitors from other OpenSim grids can come in through the hypergrid, access all features, and take items home, including 3D assets created within the world that belong to an avatar and can be taken back to the home grid.

Land rental is not possible.

ThinkSim does not compete with other grids for residents. This is by design. I am not interested in alienating people from their communities. At its core, ThinkSim is a technology showcase and enterprise platform. Companies wishing to use the space for presentations, conferences, product demonstrations or simulations can rent space on a project basis. Think about renting a place outside of your neighborhood.

The grid is still being built, and its main role is to demonstrate what ThinkNEO can do.

And speaking of technology…

AI governance, integrated voice, 3D content production, AI agent

I haven’t returned to OpenSim to build another social grid. We came back because we realized that a properly configured virtual world is the perfect demonstration environment for AI governance. All interactions can be recorded. All agents can be audited. All model inferences can be attributed and controlled. So that’s what we created. ThinkSim is not a grid with AI added. This is the grid on which AI governance is based.

For many years, OpenSimulator’s voice has been a constant headache. Options are weak, expensive, or abandoned. I’m proud that we solved the problem. ThinkVox is a native dual-mode voice solution built directly into ThinkSim. Provides complete local proximity voice and grid-wide voice instantly, without dependency on external services and without additional costs to residents. If you’ve ever tried running voice in OpenSim before, you’ll understand why this is important.

Snapshot of the ThinkSim grid. (Image credit: Fabio Bastos.)

One of the things I’m most excited about is creating native 3D content directly within virtual worlds. Residents can use Tencent’s Hunyuan 3D to create 3D meshes within the world. But we didn’t stop at creation. All assets go through an automatic Blender headless pipeline that modifies normals, generates four levels of detail, and unwraps UV coordinates. Completed meshes are moved directly to your inventory, ready to use. There are no external tools. There is no manual post-processing. When prompted, your inventory will display available 3D objects.

ThinkSim Government is a real digital certificate system built on the grid. Issue verifiable, encrypted signed documents (certificates, credentials, academic records) directly within your virtual environment. Certificates are not props or decorations. It is real, auditable and sustainable. We demonstrated this feature in real time with researchers from USP/ICMC, the Institute for Mathematics and Computer Science at the University of São Paulo.

Snapshot of the ThinkSim grid. (Image credit: Fabio Bastos.)

OpenClaw is our in-world AI agent framework. OpenClaw agents can work with residents to process requests, support tasks, interact with the environment, and execute governance workflows. All inference calls made by the OpenClaw agent are routed through ThinkNEO. All decisions are recorded. All operations are auditable. This is not a chatbot connected to a virtual world. These are managed AI workers operating within a virtual environment.

I am an NVIDIA Inception member and have natively integrated NVIDIA models into ThinkSim. Nemotron and other NVIDIA models can be used directly by residents and agents inside the grid. We also built a virtual NVIDIA DGX machine inside ThinkSim. It is a physical representation of the infrastructure running behind the scenes. When you access a DGX machine in ThinkSim, you interact with the actual governance layer that controls AI inference across the entire grid.

I started my career in media. I co-hosted a radio program at Radio Roquete Pinto in Rio de Janeiro. I built one of the first live streaming platforms in Brazil in the 1990s, years before YouTube existed. So it felt right to bring music back into the virtual world. Radio SUNO is an AI-generated radio station that runs continuously within ThinkSim, generating music in real time with an ambient soundtrack from the grid.

Every single inference call made inside ThinkSim by residents, agents, or the grid itself is routed through ThinkNEO’s AI control plane. Runtime guardrails are applied in real time. An immutable audit trail is maintained for all operations. Costs are calculated per user, region, and task. Nothing is out of control. This is not a feature. It is architecture.

Snapshot of the ThinkSim grid. (Image credit: Fabio Bastos.)

The demonstration with researchers from the Institute of Mathematics and Computer Science at the University of São Paulo was one of the most significant moments of the project so far. USP/ICMC is one of Brazil’s leading AI research institutions. The researchers involved were Ricardo Marcacini, Bruce Neves dos Santos, and Willian Hans Goes Correa.

In the demo, we ran a complete five-step government AI governance workflow within ThinkSim. In the first step, citizens submitted requests from within the world through OpenClaw agents. It’s exactly like walking into a government office and filling out a form, except the agent takes the request and starts processing it autonomously. In the second step, the AI ​​agent processed the submitted document and verified the requester’s identity, cross-referencing the information with the grid’s data layers. In the third step, ThinkSim Government issued physical digital certificates that were cryptographically signed and enabled for QR verification. This is not a certificate simulation. It was actually a verifiable document. In the fourth step, a SHA-256 audit trail was generated and stored immutably. This means that every action in the workflow is recorded in a tamper-resistant chain. In the fifth step, ThinkNEO recorded all reasoning decisions made throughout the process, creating a complete governance trace from citizen request to certified outcome.

USP researchers were quick to comment on future use cases for AI governance in Brazilian public institutions. Those talks are currently ongoing.

Snapshot of the ThinkSim grid. (Image credit: Fabio Bastos.)

AI engine that runs everything

ThinkNEO is the AI ​​control plane that supports ThinkSim’s governance layer. It is also a standalone enterprise product sold to organizations that need to manage their AI deployments.

The core idea is simple. All inference calls are routed through ThinkNEO before reaching the model provider. ThinkNEO enforces runtime guardrails, provides full observability, determines cost per business unit, and maintains an immutable audit trail. It is provider agnostic, works with Claude, GPT, Gemini, Nemotron or any other LLM, and requires no code changes on the application side.

As of April 2026, ThinkNEO is an NVIDIA Inception member and an approved Anthropic Partner Network member. MCP Server is registered with awesome-mcp-servers and Glama and offers a free tier with 500 calls per month, 12 governance tools, and no credit card required.

Chroma CEO Jeff Huber personally responded to the FOFOCA project on Sunday: “That’s pretty cool. Is there any way we can improve Chroma to fit your use case?” Pedro Minatel from the Espressif developer relations team contacted us within 48 hours and invited us to publish a technical article on the official Espressif developer portal.

Can people license the technology for their own private power grid? yes. And this is actually the most interesting part.

I am currently developing a modular package that will allow direct integration of ThinkSim’s technology into other OpenSim grids. My goal was not to sell OpenSim itself, but to bring the innovations built on top of OpenSim to the broader community. The integrated model works as follows:
Obtain your API key via a digital residency certificate issued by ThinkSim Government. So the authentication system becomes the gateway.
To the tech ecosystem.

The voice solution, ThinkVox, is already available for licensing in other grids and is in fact the only commercial service currently available.

The vision is simple. ThinkSim is where people come to discover new technologies and bring them home to the grid. It is not a destination to live, but a hub to learn.

contact me FabioTo learn more, contact us @thinkneo.ai, follow us on Facebook, X, GitHub, Discord, or subscribe to our YouTube channel.

Fabio Bastos
Latest posts by Fabio Bastos (See all)