LAS VEGAS, Jan. 06, 2026 (GLOBE NEWSWIRE) — Building upon its Visual Engine product – a generative architecture that visually interprets user workflows and reimagines human-computer interfaces – Brelyon announced today at CES the unveiling of an adaptive intelligence feature that enables automation directly from visual content. (More info here)
Visual Engine goes beyond current AI-driven recommendations, which are typically constrained by user-defined parameters or application-specific interfaces, and consequently still depend on manual inputs. Brelyon’s approach removes that bottleneck: as users interact with their computers as they normally would, Visual Engine infers the details from rendered pixels, without explicit scripting or configuration.
Brelyon Visual Engine uses NIMs capabilities along with real-time shader programming to extract data across different applications and interfaces.
At its core, Visual Engine observes changes in screen content with everyday interactions like scrolling, clicking, and dragging, and it learns the underlying correlations between those actions and user tasks. “Working with large vision models, Visual Engine applies deep learning at the visual layer to decipher those correlations and extract a sequence of actions and tasks from the changes in the image, adding a layer of intelligent automation on top of existing software and across platforms,” says Professor Alireza Aghasi, Head of Algorithms at Brelyon.
“Drawing from principles in algorithmic genetics, Visual Engine maps user behaviors much like sequencing genes—analyzing interactions to form what we call an Action Genome. By identifying and extending these recurring behavioral ‘motifs,’ the system constructs a statistical foundation to predict and automate complex workflows. In essence, it learns operational sequences the way genetic models uncover functional patterns,” explains Prof. Sheshmani of Massachusetts Institute of Technology Institute for Artificial Intelligence and Fundamental Interactions (MIT IAIFI)
Barmak Heshmat, Brelyon CEO, explains potential applications: “Last year we announced our visual engine with NVIDIA; this year we are really building a platform that goes beyond just visual rendering and unlocks a fundamental approach to progressively automating via vision. It’s like automating automation via vision; instead of spelling out to LLMs word by word what you want every time, Brelyon Visual Engine leverages vision to progressively automate your interface.”
Because Visual Engine operates on visual content, it is independent of the application layer, underlying codebase, or software frameworks. This omni-compatible design enables augmentation of legacy systems in vision-based visualization settings. Screen-agnostic, it integrates with any LCD, OLED, or Brelyon Ultra Reality® display for further immersive experiences. The product launches in Q2 2026, targeting enterprise visualization portals—especially simulation training, teleoperation, and operation centers with repetitive GUI tasks.
Click here for more information about Visual Engine.
ABOUT BRELYON
Headquartered in San Mateo, California, Brelyon is a pioneer of hardware and software technologies for headset-free immersive experiences that transcend the boundaries of physical space. Helmed by a founding team of scientists and industry executives from MIT, UCF, DARPA and IMAX, Brelyon has been selling no-headset display technology with monocular depth for defense and commercial applications since 2022. For more information about Brelyon and its virtual display solutions, please visit www.brelyon.com.
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/a48077cd-3248-4502-95a8-9ecd07d243dc.
