From 2010 to 2015, I ran Meet3D — one of many largest OpenSimulator grids on the planet, with 309 areas. Academic establishments, companies, and communities used it for coaching simulations, digital occasions, and reside collaboration. It was one of the vital energetic grids exterior Second Life, and constructing it taught me every thing I find out about distributed infrastructure at scale.
Once I shut it down in 2015, I moved on to different issues. I based Sedina in Uruguay — the primary firm to legally commercialize cannabis-derived merchandise in Latin America. I constructed a commodities buying and selling operation in Brazil. I moved to China. Ultimately, I landed in Thailand, the place Sedina now operates within the authorized hashish market. And someplace in the midst of all that, I began enthusiastic about AI governance.
I by no means actually left digital worlds. I simply took a detour by hemp, soybeans, and synthetic intelligence.
At this time, I run ThinkNEO, an enterprise AI management aircraft firm primarily based in Hong Kong. I’m additionally constructing FOFO), a 100% open-source home AI robotic. And I’ve returned to digital worlds — this time with one thing I couldn’t have in-built 2010: ThinkSim.
ThinkSim is the successor to Meet3D. However it’s also one thing fully completely different. It’s, so far as I do know, essentially the most technologically superior OpenSimulator grid working on the planet as we speak.
You possibly can go to the grid by way of hypergrid at thinksim.house:9000 or create an avatar and log in immediately.
Guests from every other OpenSim grid can are available in by way of hypergrid, entry all of the options, and take issues again house with them — together with 3D property generated in-world, which belong to the avatar and may be taken again to their house grid.
Land rental isn’t out there.
ThinkSim isn’t competing with different grids for residents — that’s by design. I’ve little interest in pulling individuals away from their communities. What ThinkSim is, at its core, is a expertise showcase and enterprise platform. Corporations that wish to use the house for shows, conferences, product demonstrations, or simulations can lease house on a undertaking foundation — suppose venue rental, not neighborhood.
The grid continues to be being constructed, and its major function is to showcase what ThinkNEO can do.
And, talking of expertise…
AI governance, built-in voice, 3D content material creation, and AI brokers
I didn’t return to OpenSim to construct one other social grid. I returned as a result of I spotted {that a} digital world — correctly architected — is the proper demonstration atmosphere for AI governance. Each interplay may be logged. Each agent may be audited. Each mannequin inference may be attributed and managed. So that’s what we constructed. ThinkSim isn’t a grid with AI as an add-on. It’s a grid the place AI governance is the inspiration.
For years, voice in OpenSimulator was a persistent headache. The choices had been fragile, costly, or deserted. I’m proud to say we solved it. ThinkVox is a local dual-mode voice answer constructed immediately into ThinkSim. It delivers full in-region proximity voice and grid-wide voice out of the field, with no dependency on exterior companies and no further value to residents. In case you have tried to run voice in OpenSim earlier than, you’ll perceive why this issues.
One of many issues I’m most enthusiastic about is native 3D content material creation immediately contained in the digital world. Residents can generate 3D meshes in-world utilizing Hunyuan 3D from Tencent. However we didn’t cease at technology — each asset passes by an automated Blender headless pipeline that fixes normals, generates 4 ranges of element, and unwraps UV coordinates. The completed mesh goes on to the consumer’s stock, prepared to make use of. No exterior instruments. No handbook post-processing. You immediate, and also you obtain a usable 3D object in your stock.
ThinkSim Authorities is an actual digital certificates system constructed into the grid. It points verifiable, cryptographically signed paperwork — certificates, credentials, tutorial data — immediately from inside the digital atmosphere. The certificates are usually not props or decorations. They’re actual, auditable, and chronic. We demonstrated this functionality reside with researchers from USP/ICMC — the College of Sao Paulo’s Institute of Arithmetic and Laptop Sciences.
OpenClaw is our in-world AI agent framework. OpenClaw brokers work alongside residents — they will deal with requests, help with duties, work together with the atmosphere, and execute governance workflows. Each inference name made by an OpenClaw agent routes by ThinkNEO. Each determination is logged. Each motion is auditable. This isn’t a chatbot hooked up to a digital world. These are ruled AI employees working inside a digital atmosphere.
I’m an NVIDIA Inception member, and we’ve got built-in NVIDIA fashions natively into ThinkSim. Nemotron and different NVIDIA fashions can be found to residents and brokers immediately contained in the grid. We’ve got additionally constructed digital NVIDIA DGX machines inside ThinkSim — bodily representations of the infrastructure working behind the scenes. Whenever you stroll as much as a DGX machine in ThinkSim, you’re interacting with the precise governance layer that controls AI inference throughout your complete grid.
I began my profession in media. I co-hosted a radio program on Radio Roquete Pinto in Rio de Janeiro. I constructed one of many first reside streaming platforms in Brazil within the Nineteen Nineties, years earlier than YouTube existed. So it felt proper to convey music again into the digital world. Radio SUNO is an AI-generated radio station working repeatedly inside ThinkSim, producing music in actual time because the ambient soundtrack of the grid.
Each single inference name made inside ThinkSim — by residents, by brokers, by the grid itself — routes by ThinkNEO’s AI management aircraft. Runtime guardrails are enforced in actual time. An immutable audit path is maintained for each motion. Prices are attributed per consumer, per area, per operation. Nothing runs ungoverned. This isn’t a function. It’s the structure.
The demonstration we ran with researchers from the College of Sao Paulo’s Institute of Arithmetic and Laptop Sciences was, for me, one of the vital significant moments on this undertaking up to now. USP/ICMC is considered one of Brazil’s main AI analysis establishments. The researchers who participated had been Ricardo Marcacini, Brucce Neves dos Santos, and Willian Hans Goes Correa.
The demo ran an entire five-stage authorities AI governance workflow inside ThinkSim. Within the first stage, a citizen submitted a request in-world by way of an OpenClaw agent — precisely as they may stroll right into a authorities workplace and fill out a kind, besides the agent acquired the request and started processing it autonomously. Within the second stage, the AI agent processed the submitted doc and validated the requester’s identification, cross-referencing the knowledge in opposition to the grid’s knowledge layer. Within the third stage, ThinkSim Authorities issued an actual digital certificates — cryptographically signed and QR-verifiable. This was not a simulation of a certificates. It was an precise verifiable doc. Within the fourth stage, a SHA-256 audit path was generated and saved immutably — each motion within the workflow recorded in a tamper-proof chain. Within the fifth stage, ThinkNEO logged each inference determination made all through the method, creating an entire governance path from citizen request to licensed final result.
The USP researchers made unprompted references to future use instances for AI governance in Brazilian public establishments. That dialog is ongoing.
The AI engine that runs all of it
ThinkNEO is the AI management aircraft that powers ThinkSim’s governance layer. It’s also a standalone enterprise product that I promote to organizations that want to manipulate their AI deployments.
The core thought is straightforward: each inference name routes by ThinkNEO earlier than reaching any mannequin supplier. ThinkNEO enforces runtime guardrails, offers full observability, attributes prices per enterprise unit, and maintains an immutable audit path. It’s provider-agnostic — it really works with Claude, GPT, Gemini, Nemotron, or every other LLM — and it requires no code adjustments on the appliance aspect.
As of April 2026, ThinkNEO is an NVIDIA Inception member and an accredited Anthropic Companion Community member. Our MCP Server is listed on awesome-mcp-servers and Glama, with a free tier available with 500 calls per thirty days, 12 governance instruments, no bank card required.
Jeff Huber, the CEO of Chroma, responded personally to our FOFOCA undertaking on a Sunday: “Very cool! Any method we are able to enhance Chroma in your use case?” Pedro Minatel from Espressif’s Developer Relations workforce reached out inside 48 hours of our outreach and invited us to publish a technical article on the official Espressif Developer Portal.
Can individuals license the tech for their very own non-public grids? Sure. And that is really essentially the most thrilling half.
I’m at present growing a package deal of modules that may permit different OpenSim grids to combine ThinkSim’s expertise immediately. My aim has by no means been to promote OpenSim itself, however to supply the improvements constructed on high of it to the broader group. The mixing mannequin works like this: grids will probably be in a position
to acquire an API key by a digital residency certificates issued by ThinkSim Authorities — so the certification system turns into the gateway
to the expertise ecosystem.
ThinkVox, the voice answer, is already out there for licensing by different grids and is, the truth is, the one business service at present provided.
The imaginative and prescient is straightforward: ThinkSim is a spot the place individuals come to find new expertise and take it again to their house grids. Not a vacation spot to reside in, however a hub to be taught from.
Contact me at fabio@thinkneo.ai to be taught extra, or follow us on Facebook, on X, on GitHub, on Discord, or subscribe to our YouTube channel.
