You are considered odd (possibly pitiable) if you don't have an MG in your head. The MG is about the size of a small fingernail implanted directly into the base stem of the brain. From there, microscopic strands grow out into your brain, where they end in small biochips attached to individual neurons. They can suck down data (so the MG knows what the neurons are doing) or upload data (modifying the neural signals). Most MGs are heavily restricted in what they can and can not do, but basic visual and audial data, and a thought interface, are considered the minimum.
Japan: Japan has gone whole hog (and even has a more advanced MG that they use), and many Japanese citizens live entirely in virtual reality, with their bodies cared for at outrageous rates. Note that Brazil and Europe commonly have 5-sense MGs, but these are not nearly as far gone as Japan.
Within your field of vision, flat windows exist, positioned in 3D reality, usually at the distance, angle and size most comfortable for you. A white-gloved hand floats in space somewhere near you. You control it by thinking, as easily as if it was one of your two meat-based hands. It can manipulate windows, gesture, snap its fingers, etc. There are a fair variety of programs which have "gesture spells" associated with them - if the hand makes particular gestures, the programs perform particular actions.
However, in addition to the hand, you can directly think windows open or closed, activate programs by thinking of them, and so on.
Some people use a virtual keyboard, which they can type by thought.
Most people also give their PAIA (Personal AI Assistant) an avatar, although this is certainly not necessary. But if you do, the avatar will usually be somewhere in your field of vision. Whether you do or don't, the PAIA will be able to manipulate everything you allow it to within your field of vision.
Dictation is more complicated. Words occur on many levels of the mind, and only the "near speech" level can be read by the most advanced MGs. This level, in turn, is difficult to activate without subvocalizing... and many people prefer not to have the MG reading their surface thoughts even then. So most people have a small implant in their skull which simply takes dictation from subvocalizing.
Japan: Some experimental Japanese MGs actually interface with thought on a variety of levels, providing such services as enhanced memory, context-based data enhancement, or even "psychic diaries". The full implications as it exists in the Japanese culture, however, generally give the rest of the world shudders.
Brazil: Software programmers in Brazil have developed what they believe to be an extremely good start on "uploading" one's brain into a purely software form, as well as a software backup of your brain. However, it is still deep in testing.
The PAIA is usually built into your MG these days. A PAIA is a pure logic system, but carries many human-seeming traits: the ability to establish specific goals based on broad statements of purpose; to intelligently analyze a situation based on those goals; and to make recommendations. Without a goal provided by an outside source, however, a PAIA feels no inner urge to do something, does not get bored, does not experience the passage of time.
Most PAIAs have a bland but attractive voice, although some people like to add accents or buy trademarked voices for use with their PAIA. People also often buy an avatar for the PAIA, to represent what the PAIA is doing at any given time, although the PAIA does not really need this. Some people get very attached to their PAIA setup.
PAIA's are insanely useful. The most common services are as follows:
Your MG has Global & Personal Positioning (GPP). It knows exactly where you are, down to the centimeter, and what direction you are facing. It is programmed to understand your field of vision, and your preferences.
Your MG is also connected to Webnet, tied into the global information system. All around you, data flows unimpede by the physical world, and the MG is aware of all this, and can bring it into your field of vision.
A specialized part of the invisible world is MuseWeb. The essence of it is this: an artist (from musicians to writers to painters) puts up a website on MuseWeb for free. The website can have art samples (or even free art), a biography, anything. And via the website, the artist can sell things, from abums to T-shirts to original performances. MuseWeb takes a small percentage (pennies on dollars) of the sales to maintain itself (and make a tidy profit), and the artist gets the middle man issues taken care of.
Any given artist can be found from anywhere on the globe, but they are "higher" in the search results as they become more popular in a particular region. And the artists "popularity score" times the "region size" for that score, determines how high she ranks in universal listings. Popularity is measured by the classic benchmarks - sales, downloads, and links to the site.
With the invisible world, you can listen to local musicians as you walk along (usually a random sample, or a sample based on your PAIA's understanding of your preferences). You might see a particular painter's "virtual museum" when you walk by his residence. And so on.
A utility technician for Austin, walking along the street, can have an overlay in his field of vision, of power lines, water lines, sewer paths, etc. When she opens up a power control box, it provides her with a technical manual for its operation and repair. If she's walking in one of the larger sewer pipes, it provides her with green lines showing the path of the pipes beyond Line of Sight.
Similarly, a mechanic operating on your car will be in communication with the car, as it explains what is wrong and provides technical manuals on the damaged elements. A soldier needing to fix his handgun, or a technician tearing open a Ink Padd, will have a similar experience.