What ME is building, and how far along each one is.

A general purpose x86-64 operating system written from scratch, where agents are first class system components rather than programs bolted on top. Execution, permissions, memory, task routing, recovery, and human oversight are meant to be OS level concepts.

• Standard Linux and Unix file formats, filesystem layout, and executable formats wherever practical
• No ME specific wrapper or package format required of applications
• Desktop organized around agents, tasks, memory, and live system state
• Reliability and recovery treated as core features, not add ons

A ceiling mounted volumetric display. The unit would live overhead so the shared workspace below stays physically clear, which suits desks, meetings, and collaborative engineering work.

• Fixed external chassis with replaceable internal modules, so emitters, sensors, and compute could be serviced independently
• Capability improvements shift into software: calibration, scene generation, control logic, and agent driven updates
• Control software is display method agnostic, so the physical mechanism can change without a rewrite
• Intended control surfaces include agents, bracers, a holo pencil, hand tracking, and ordinary engineering software

Carl is the agentic system ME products are built to talk to. The design goal is one command path shared by every surface, so an instruction from a person, a wearable, or a piece of engineering software travels the same route and is checked the same way.

• Model agnostic by design. ME expects to rely mainly on strong external models rather than train a frontier model from scratch
• Products integrate against a Carl interface boundary, never against a specific model SDK
• Human override is a first class requirement. If Carl is suspected compromised, people revoke its authority through independent systems

Wearable forearm units for ME staff. Intended roles include hand and arm tracking for holographic interaction, identity, local agent access, notifications, sensing, and context aware controls.

• Hologram control: hand and arm tracking accurate enough to select and move objects in a display volume
• Carl access from wherever the wearer is, through the same command path every other surface uses
• Identity and authentication, so a shared machine can tell who is standing at it
• Inertial and optical tracking first. Surface electromyography for finer control is a research question, not a plan
• Modular, so different departments can add specialized capabilities
• Calibration at first use and continuous adaptation afterwards, because bodies and wear positions differ
• Accessibility: another input channel for people for whom a keyboard and mouse are not the best fit

A modular protective suit for lab and prototyping work. The base suit carries identity and connects to an Employee Bracer; standardized locking rails let modules be added for the work actually being done.

• Modular protection rather than one suit for every job
• Base suit integrates with Employee Bracers for identity and authentication
• Standardized locking rails, so modules attach the same way every time
• First module: an upper chest sensor pod measuring temperature, humidity, basic air quality, ambient light, and motion and orientation from an inertial unit
• Modular visor. V1 is anti fog with glare control and no electronics
• A later visor could show a heads up display: safety state and the current task first, measurements and agent output only after that earns its place
• The bracer is the intended input: a future neural band style channel for hands free control
• First workflow: navigating a checklist without using your hands
• Important actions require confirmation, and dismissing a safety warning takes a deliberate physical tap on the bracer

A precision powered screwdriver, first of an eventual multi tool family. Torque accuracy comes first; interchangeable bits and further tool heads come later.

• Torque accuracy is the first priority, before any other feature
• Interchangeable bits later, other tool heads later still
• Carl and the task context supply which fastener is expected, and the driver checks that against what it is doing
• Automatic stop at the target torque, with a haptic confirmation the hand can feel
• An abnormally fast torque rise stops the driver immediately and warns of a likely jam, cross thread or wrong fastener, with a concrete next action rather than an error code
• The result eventually attaches to the task record, so what was tightened, to what, and when, is not a memory

A common platform for the tools every ME branch keeps rebuilding: simulation, measurement, data capture, experiment tracking, and the scientific computing under all of them.

• One toolchain shared across computing, energy, robotics, materials, and biotech work
• Results and instrument data carry provenance rather than living in scattered files
• Agent access through the same interface boundary products use
• A closed loop: question, simulation, experiment, sensing, analysis, notebook, next experiment