Startup Founded by Ex-Nvidia Engineers Builds a Humanoid Robot Intern for Offices

Flexion Robotics, a startup founded by former Nvidia engineers, has unveiled a humanoid robot that can replace an office intern. The company developed an unconventional training method that lets the robot handle a wide range of tasks, from working on a computer to following spoken instructions. Wired calls the result "frighteningly competent."

AI-processed from Wired; edited by Hamidun News
Startup Founded by Ex-Nvidia Engineers Builds a Humanoid Robot Intern for Offices
Source: Wired. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceWired ↗
Editor-in-Chief
Key metric #1finding and moving objects around the office without prior mapping
Working on a computerkeyboard, mouse, navigating interfaces
Key metric #3following spoken instructions for unfamiliar tasks

Flexion Robotics, a startup founded by former Nvidia engineers, has unveiled a humanoid robot that can replace an office intern. Wired, which tested the system, called it "frighteningly competent", a phrase the industry does not use lightly: most current humanoids struggle even with basic tasks in an unstructured office environment.

Nvidia alumni

The Flexion team comes from Nvidia, whose GPUs became the foundation of the entire modern AI boom. The engineers spent years working on neural network architectures and parallel computing systems. They decided to apply that expertise differently: instead of building hardware for training models, they set out to create a physical agent that learns on its own to act in the real world.

The choice of focus is no accident. The office is one of the hardest environments for robots: an endless variety of objects, unpredictable interactions with people, and tasks that constantly change in form and context. That is why industrial manipulators, which work well on an assembly line, are practically helpless in an open-plan office.

An unconventional training method

Flexion's main competitive advantage is its approach to training robots. Most systems either require rigid programming for each specific action or suffer from the "transfer problem": skills learned in the lab translate poorly to real-world conditions. According to Wired, Flexion found a "clever way" around this limitation.

The company keeps the details of the method under wraps, but the result speaks for itself: the humanoid handles a wide range of tasks typical of an office intern:

  • finding and moving objects around the office without prior mapping
  • working on a computer: keyboard, mouse, navigating interfaces
  • following spoken instructions for unfamiliar tasks
  • adapting to unusual objects and changing conditions
  • interacting with employees in a shared space without protective barriers
"Frighteningly competent intern" is the headline

Wired chose, and the outlet rarely resorts to such superlatives in its reporting on humanoids.

The office as the next frontier

The humanoid robotics market is in an investment boom: billions of dollars have poured into it over the past two years. Figure AI, Physical Intelligence and Agility Robotics are picking up the pace, but most players focus on warehouses and manufacturing, where the environment is more predictable and controlled.

Flexion is betting on the office, a segment long considered too chaotic for automation. Intern work looks low-skilled, but it eats up a significant amount of time in any organization: delivering documents, printing materials for a meeting, entering data from paper forms into a system.

If Flexion's training method really does scale without reprogramming for each new task, the economics of deployment change radically: one robot could be quickly retrained instead of ordering a specialized system.

What this means

Office robots may move from the "someday" category to "within this decade." For businesses, this is not a threat from the distant future but a competitive question for the next three to five years. Flexion is one of those startups worth watching closely.

Story

Key Questions Answered

What are the key details on Nvidia alumni?

The Flexion team comes from Nvidia, whose GPUs became the foundation of the entire modern AI boom. The engineers spent years working on neural network architectures and parallel computing systems. They decided to apply that expertise differently: instead of building hardware for training models, they set out to create a physical agent that learns on its own to act in the real world. The choice of focus is no accident. The office is one of the hardest environments for robots: an endless variety of objects, unpredictable interactions with people, and tasks that constantly cha

What are the key details on An unconventional training method?

Flexion's main competitive advantage is its approach to training robots. Most systems either require rigid programming for each specific action or suffer from the "transfer problem": skills learned in the lab translate poorly to real-world conditions. According to Wired, Flexion found a "clever way" around this limitation. The company keeps the details of the method under wraps, but the result speaks for itself: the humanoid handles a wide range of tasks typical of an office intern: finding and moving objects around the office without prior mapping working on a computer: ke

What are the key details on The office as the next frontier?

The humanoid robotics market is in an investment boom: billions of dollars have poured into it over the past two years. Figure AI, Physical Intelligence and Agility Robotics are picking up the pace, but most players focus on warehouses and manufacturing, where the environment is more predictable and controlled. Flexion is betting on the office, a segment long considered too chaotic for automation. Intern work looks low-skilled, but it eats up a significant amount of time in any organization: delivering documents, printing materials for a meeting, entering data from paper fo

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…