Experience THINX™
See procedural intelligence at work.
Watch the narrated walkthrough, then try the interactive procedure illustration yourself. The evidence behind the technology is set out below.
01
Watch the technology
A narrated walkthrough of what THINX™ does and what has already been tested, synchronised with engineering visualisations.
Bench assembly
Model 01- 01 BASE PLATE · SOCKETED
- 02 LONG ROD · LOCATING
- 03 SQUARE BLOCK · BORED
- 04 ROUND CAP · PUSH FIT
MODEL 01 · ORTHOGRAPHIC PROJECTION · SIMPLIFIED SCHEMATIC — NOT A MANUFACTURING DRAWING
- 01Base plate
- 02Long rod
- 03Square block
- 04Round cap
Stage 1 · Guidance
Place the base plate on the bench.
A fictional socketed base, locating rod, bored block and push-fit cap. Follow the four-stage procedure to assemble the model.
Narrated walkthrough · American English
0 / 238sPrerecorded narration
Welcome. If this is your first time seeing INSTRUCTIBLZ, let's start with a simple question: what am I looking at? These are engineering drawings of a small model, seen from the front, the side and the top. It's a made-up object, chosen so you can follow along easily. And just so you know, everything you'll see here is a labelled simulation, not the real THINX engine.
Full transcript
- 0:00 — Welcome. If this is your first time seeing INSTRUCTIBLZ, let's start with a simple question: what am I looking at? These are engineering drawings of a small model, seen from the front, the side and the top. It's a made-up object, chosen so you can follow along easily. And just so you know, everything you'll see here is a labelled simulation, not the real THINX engine.
- 0:28 — Here's the model pulled apart. There are four parts: a base, a long rod, a square block, and a round cap. Our job is to put them together, in the right order.
- 0:44 — So, how does this help someone perform a real task? Step one. Place the base on the bench. The system checks what has happened, confirms it, and only then moves you on.
- 0:59 — Step two. Push the long rod into the hole in the middle of the base. Each confirmed step updates the drawing, and adds a line to the record.
- 1:11 — Now, imagine a learner gets stuck and asks: how does the block fit? Instead of hunting through a manual, they get a short answer about this exact step. Line up the hole in the block with the rod... and slide it down.
- 1:29 — But what if someone reaches for the cap too early? Here, on the left, is what the procedure expects. On the right, is what was actually observed. The difference is spotted, progress is held, and the guidance explains exactly how to put it right.
- 1:48 — The block goes on. The missing step is now done, and the procedure moves forward.
- 1:56 — Finally, the cap. All four stages are confirmed, and there's a complete record of what was done, including the correction along the way.
- 2:07 — So, what makes THINX fundamentally different? First, it pays attention to what is really happening. A manual, or a training video, plays the same way whether you're right or wrong. A chatbot answers questions, but it can't see your work. THINX compares each action with the approved procedure. And when the view is blocked or unclear, it withholds its decision. It doesn't guess. Nothing is marked complete until there's enough evidence.
- 2:42 — Second, the core engine itself doesn't change from one task to another. Only the procedure it follows does. In Óscar's development testing, the same engine guided a five-step bookshelf assembly and, in separate sessions, a four-step GRUBBAN assembly, with no rewriting of the engine in between.
- 3:06 — And here's why those results matter. In a controlled demonstration, THINX guided all five of five steps. All six of six GRUBBAN sessions were completed. Mistakes were identified in about two seconds, quickly enough to correct them before moving on. And seventy automated tests are passing. These are reported development results, not independently certified commercial performance. But they show the idea working in practice.
- 3:37 — That's INSTRUCTIBLZ, powered by THINX: your approved procedures, turned into guidance that follows along, answers questions, and helps people get it right. Remember, this walkthrough is a simulation. Try the procedure yourself in interactive mode.
02
Try it yourself
Explore the same four-stage assembly at your own pace: follow the guidance, ask scripted questions and see how a mismatch is handled.
Bench assembly
Model 01- 01Base plate
- 02Long rod
- 03Square block
- 04Round cap
Stage 1 · Guidance
Place the base plate on the bench.
A fictional socketed base, locating rod, bored block and push-fit cap. Follow the four-stage procedure to assemble the model.
03 · The engineering behind THINX™
From concept to working intelligence.
Working with engineer Óscar Palomar, New Icons Design has taken THINX™ beyond its original concept and into a functioning procedural intelligence system.
The challenge was not simply to display instructions through smart glasses. It was to develop intelligence capable of following the progress of a task, recognising when conditions have been met, responding to uncertainty and helping a user complete the correct procedure.
The development work has now demonstrated these capabilities through controlled testing, using smart glasses, structured procedures and a reusable intelligence engine.
5/5
Procedural steps guided
All five steps completed through the M300-connected THINX test.
6/6
New procedure sessions completed
Six GRUBBAN test sessions completed using the existing intelligence engine.
~2s
Error identification
A deliberately incorrect component orientation was recognised during the recorded M300 test.
70
Automated software tests passing
The procedural core passed its reported automated test suite.
Intelligence that follows the work
Instructions that respond to progress.
Óscar developed a system that combines a structured procedure with observations of the task being performed.
In controlled testing, the Vuzix M300 glasses provided visual observations to the THINX engine, running on a connected computer. THINX compared those observations against the requirements of the current step and returned the appropriate instructions to the glasses.
When the necessary conditions were recognised, the system automatically advanced to the next instruction. The user did not need to stop and manually navigate through the procedure.
Knowing when to proceed, and when to wait
Confidence matters as much as recognition.
THINX was developed to distinguish between an observed error and a situation where there simply isn’t enough information.
Óscar tested the system under changes in lighting, glare, blur, camera movement and partially obscured views. Rather than treating every unclear observation as a mistake, THINX could pause, wait for a clearer view and continue when the required evidence became available.
The tests also explored detection of incorrect components and orientations, with guidance explaining how to correct the problem and resume the procedure.
One engine, different procedures
A new procedure does not require a new intelligence engine.
One of Óscar’s most important achievements was demonstrating that THINX could guide more than one procedure.
Following the initial bookshelf test, he developed a converter that transformed an entry from the IKEA-Manual research dataset into a THINX Procedure Package. This package represented the components, their relationships, the required assembly sequence and possible procedural errors.
After adding task-specific wording from the original manual, the package was tested through the same underlying THINX engine. All six GRUBBAN sessions reached completion, including scenarios with deliberate mistakes. The engine itself did not need to be rewritten for the new product.
ONE THINX ENGINE
Unchanged between procedures
PROCEDURE PACKAGE A
Bookshelf — five steps
PROCEDURE PACKAGE B
GRUBBAN — four steps, from the IKEA-Manual research dataset
An architecture built for flexibility
The intelligence is separate from the device.
Óscar designed THINX so the procedural engine and the device presenting guidance remain separate. The system receives structured observations and returns instructions, confirmations, corrective messages and procedure status.
The M300 demonstration tested this through smart glasses. Separate GRUBBAN simulations exercised the same core through its communication interface. This architecture provides the foundation for INSTRUCTIBLZ experiences across different devices and working environments.
The device-independent interface has been tested through smart glasses and event-link simulation; complete production applications on every device have not all been demonstrated.
THINX CORE
The commercial significance
Why this matters to business.
Every organisation depends on procedures. They underpin training, quality, maintenance, safety and everyday operations. Traditionally, those procedures are communicated through documents, videos, classroom instruction and the experience of established employees.
THINX™ points towards a different model: approved procedural knowledge delivered as intelligent assistance at the moment it is needed. The same underlying technology could support different tasks, departments and working environments, with procedure-specific configurations rather than a separate software engine for every application.
For businesses, that creates the opportunity to rethink how people are trained, supported and guided through complex work.
The significance of THINX™ is not simply that it can deliver instructions. It is that those instructions can respond to the state of the work itself.
Discuss a PilotAbout the development tests
The October 2026 M300 tests used a controlled visual test stage displayed on a laptop, observed through the glasses’ camera and processed by the real THINX engine. The GRUBBAN sessions used structured simulations connected to the same engine through its event interface.
The interactive browser illustrations on this website are explanatory demonstrations rather than live connections to the Python engine. The development results are drawn from Óscar Palomar’s report dated 30 October 2026.
Your procedures. Your challenges. Intelligent guidance.
Whatever your sector, tell us where people need better training, clearer instructions or practical support. Let’s explore how INSTRUCTIBLZ™ could be adapted to your procedures.