On Macs the nubs are on the F and J keys.  On PCs they are typically on the D and K keys.  Either way, they help the touch-typist to get their fingers in the right place without having to divert attention from what they are looking at on screen.

The strange thing is, these are the only keys with nubs.  This helps with the heavily used keys.  Our muscle memory helps our fingers find other letters around this center.  But it’s difficult to accurately stretch those digits up to the digit row.  It’s error prone and slows typing as one must carefully check typed numbers for accuracy.  And typing numbers with accuracy is critical.  It could mean the difference between a $6000 and a $7000 request.  And unlike typing words, numerals cannot be assisted by automated spell checkers.

I’ve long wished for nubs up there on the number row to help me out.

Exactly which keys deserve nubs is a good university research project.  The 5-6-7 keys I find difficult to get right without care.   The 9 and 0 I find difficult also.  So offhand I would guess that the 1, 6 and 0 would be good candidates.

By the way, did you notice the little in-joke on the onscreen keyboard on the iPad?  Yes, you guessed it – the images for the F and J rows have little nubs on them. It’s surprising this one got through given Apple’s extreme discipline for minimalism.  Har har.  Yes, Apple, you’ve reduced us 60 wpm typists to hunting and pecking amateurs, and here’s the salt thrown in the wound in the form of nubs that can be seen but not felt.  Good one.

Someone at Apple has a sense of irony.

That “cloud” is actually made up of racks and racks of servers. Those servers are not actually in the clouds, where it is quite cold, but down on the earth, where it’s warm and getting warmer.  Those racks of servers generate heaps of thermal energy, which requires yet more energy to keep cool.  Otherwise they melt onto one another like Hersheys in the glovebox.

The software architecture that makes these clouds of servers work is modular, and fault-tolerant and distributed.  They allow plug-and-play expansion when more capacity is needed.  They are built to withstand any node failing (with thousands of servers, several will break down every day).

And those nodes can be anywhere, since everything is connected.  But given a choice, it’s preferable to put them close to where they are needed, because things are faster that way.

The purpose of a space heater is to generate heat.  Heaters are pretty dumb.  That’s all they do.  They have an electric heating element, maybe some oil to circulate through and some fins to radiate the heat, a thermostat and that’s it.

What if a space heater instead had a bunch of cheap, older generation, heat-generating CPUs and a wireless connection?  You could crank up the number of processors and their clock speed for a nice cozy hearth, or turn them down a few GHz if you just needed to keep your nose from freezing overnight.

The cloud computing companies could even give you a few cents for each MIPS-hour your heater burns.

At a larger scale, an entire office floor could have computers built into its HVAC system.  Even our desktop computers could be enlisted for the job.  They are ridiculously overpowered for what we need them for most of the time anyway.  The building’s climate control system could instruct all those idle CPUs to do something productive with their downtime, while warming the office in the winter.  Those computers could even monitor the local temperature and provide more heat where it is most needed.

All of this would:

• do something productive in the process of heating your home or office
• reduce the cost of cooling servers
• distribute computing closer to where it is needed
• maybe control temperature in a breezy office to a finer degree
• maybe subsidize heating costs by donating cycles to the cloud

My wife, tactfully: “I think you have an idea that is ahead of its time.”

Yes, it could be a while before this idea is stolen.  [This would be a fun and compelling research project for some engineering & system design students.  Anyone?]

—

Philip Haine is principal of Product Vision Associates, an innovation consultancy that helps guide product leaders and their teams to generate ideas even more important than the Network Heater.  To follow him on Twitter click here.

Looks like someone implemented this concept on the iPhone.  Click in to see a video.

Now we need a mouse manufacturer, or a university lab to design a mouse which does the same thing.

(By the way, when I say ideas documented here are “stolen” I’m kidding.  Different people walking down the same exploratory path will arrive at similar ideas.)

Patent Application number 20070080945 details a mouse having a button-less pan and scroll switch. [..] the mouse would sense different hand positions and act in different ways according to how it is being held.

“In one embodiment, the first mode of operation of the mouse may be a cursor control mode and the first operations on the display screen may be cursor movements on the display screen. Additionally, the second mode of operation of the mouse may be a pan and/or scroll control mode and the second operations on the display screen may be scrolling and/or panning movements on the display screen.”

So maybe this means if you hold the mouse normally you move the mouse; if you hold it a different way — maybe close to the bottom — it scrolls and pans.

How about this alternative as a design to steal: Hold the mouse the usual way all the time, but tilt it to scroll in a given direction. Tilt left and right to scroll horizontally or tilt to the top or bottom edge to scroll vertically. The more you tilt, the faster it scrolls.

One advantage here is you wouldn’t have to lift & shift the mouse to keep scrolling down a long page. You could just tilt the mouse up at the right steepness to scroll down the page and lift again to keep scrolling. Plus, you don’t have to shift your hand to start scrolling.

I think this idea is different than the patent application. The patent seems focused on detecting the user’s grip and in having different modes of operation, interpreting the mouse motion in different ways. This idea is about adding extra sensory input to the mouse and allowing it to remain modeless.

How might this be implemented? There are probably a number of ways… Accelerometers may be tricky as regular mouse motions accelerate the mouse.  Presumably if the mouse was moving, it wouldn’t be scrolling or panning.  Alternatively, multiple distance sensors under the mouse would be able to measure the height of different edges of the mouse and from this, tilt angles can be computed.

The sensitivity would have to be done right to avoid excessive wrist twisting. A little motion should go a long way. It should be possible to lift the edges of the mouse only by lifting fingers while keeping the wrist flat. Some rebalancing of the mouse may be in order to shift the center of mass closer to the center of the mouse.

Another consideration is the case of the mouse reaching the edge during normal mouse motions. We need to reliably distinguish between tilting to scroll and tilting while shifting a mouse back into range. One possible way would be to detect mouse movement while tilted. (The sensors on the edge of the mouse would be full-fledged mouse position sensors.) Another way would be to require the angle to pass a threshold before it’s interpreted as a scroll. The user would snap an edge of the mouse up in one direction and then reduce the angle for finer control. This topic calls for some research, starting with an analysis of how mice are brought back in-range.

Alternative: Press and drag the heel to scroll

If this tilting thing is too expensive or difficult to implement an alternative design is to construe pressure on the heel of the mouse to indicate scrolling. Dragging the mouse while pressing the heel scrolls in either direction. You can think of this as another button on the heel of the mouse that just happens to be used to initiate scrolling.

It would probably be better to interpret these movements as changes in scrolling velocity rather than in absolute scrolling position, to avoid having to shift the mouse back in range. The further from the “heel down” location the mouse moves, the faster the scroll. Releasing the heel stops the scroll.

This feels pretty good on my old, lightweight Logitech mouse: it’s easy to add pressure on the ball of the hand to press the heel. Unlike the Apple patent this avoids having to shift the hand on the mouse.

[One other thought: would it be possible to add detection for yaw? Say the tip of the mouse is lifted off the desk and the mouse pivoted clockwise and counter-clockwise on its heel. Could multiple optical mouse sensors detect the direction of the twisting? Perhaps a nice research topic.]

What struck me was how Pythonesque the whole thing seemed. Each post began by insulting a prior user, followed by adding the writer’s alternate viewpoint. Imagine this with an upper class British accent:

User 1: “The Proposition ‘A’ is true.”

User 2: “From the stunted logic of your post, your mother clearly must have smoked throughout your gestation. ‘A’ is certainly not true.”

User 1: “Eureka! I thought the dodo extinct, yet here is one posting on this very bulletin board. ‘A’ is quite self-evidently true, as any animal evolved beyond the invertebrate knows reflexively.

It’s as though everyone was following a ritual template: insult, then comment.

Clearly, urging users to avoid flame wars is of no use. That approach has failed for decades.

Is there another way? Perhaps try and go with the flow instead?

Here’s a design to steal (or at least a research project): In message boards, provide two input fields. The first is an optional insult field. The second is the body of the message, in which nothing insulting may be uttered.

Variants to try:

• Allow readers to collapse/hide the insult field for all postings, until they are turned back on again. If they don’t like that sort of thing they never have to be exposed to it.
• Allow readers to police the postings and indicate whether the no-insult-in-the-body rule is violated. A critical mass of readers reporting a violation automatically bounces a submission back to the sender for correction.
• Allow readers to rate the quality of the insult. Give positive points to most good humored or creative or funny insults, negative points to the droll or crass.

By acknowledging peoples’ need to vent and by giving it an explicit place, could we separate out the content? Someone please try and let us know.

• Build a kiosk with a PC and a nice microphone. (Maybe a push-to-talk button if necessary.)
• Hook up speech recognition software like Dragon NaturallySpeaking.
• Write an applet that receives the spoken text, parses out filler tex, looks up key phrases in Wikipedia and “formats” it for output to speech
• Have a text-to-speech synthesizer dictate the introductory paragraph of the Wikipedia article.

Then videotape some interactions:

Human: Computer, tell me about 419 scams
Computer: Advance fee fraud, often also known as the Nigerian money transfer fraud, Nigerian scam or 419 scam after the relevant section of the Nigerian Criminal Code that it violates, is a fraudulent scheme to extract money from victims after making them believe they will get an immense fortune. Victims are requested to pay an upfront fee before their purported fortune is released.

Human: Tell me more
Computer: These scams have come to be associated in the public mind with Nigeria due to the massive proliferation of such confidence tricks from that country since the mid-eighties, although they are often also carried out in other African nations, including Togo, Côte d’Ivoire, [...]

Human: (interrupting) That’s enough, thank you.
Computer: You’re welcome.

Human: Computer, tell me about Archaeopteryx
Computer: Archaeopteryx, from the Late Jurassic of Germany, is the earliest and most primitive known bird. The discovery of the first intact specimen, Archaeopteryx lithographica, in 1861, two years after Charles Darwin published The Origin of Species, set off a firestorm of debate about evolution and the role of transitional fossils that endures to this day.

Human: Computer, tell me about RSI [This is a topic with several articles.]
Computer: I have nine topics about RSI. Do you mean Repetitive strain injury, a disorder affecting bone and muscle from repetitive movement? Or…
Human: (interrupting) Yes.
Computer: Repetitive strain injury, also called repetitive stress injury or typing injury, is an occupational overuse syndrome affecting muscles, tendons and nerves in the arms and upper back. It occurs when muscles in these areas are kept tense for very long periods of time, due to poor posture and/or repetitive motions.

Human: Computer, what happend in 1521?

You get the idea. Bonus points if you make it work well over the phone. And if you get William Shatner to play the human.

Of course this mostly demo stuff. It will always be easier for humans to interact with machines primarily using hands and eyes, not voice and ears (as long as eyes and hands are available). And the human I/O ports of voice and hearing are certainly under-exploited today.)

Nevertheless, building Captain Kirk’s computer would be a cool and funny demo and a fun interaction research project. [Has anyone done this yet?]

Wikipedia + Launchbar = Information at Your Fingertips.

Accessing Wikipedia is greatly facilitated thanks to LaunchBar. As if a king, I get to invoke vast chunks of human knowledge with just four keystrokes: [cmd-space]wp[space], plus a search term like, say, Scientology.

[Here is that search template if you want to add it to LaunchBar: "http://en.wikipedia.org/w/wiki.phtml?search=*"]

But Wikipedia does not yet answer a juicy meta-question: Who else is curious about this topic? Wouldn’t it be fascinating to see how worldwide interest in topics like Islam, Intelligent Design or Louis Armstrong wax and wane as the events of the world unfold? This is the type of insight into what is on everyone’s mind, hitherto available only to the lucky few with access to search engine logs.

Wikipedia already has the raw ingredients of an amazing meme tracking tool: tons of traffic, tons of content, the ability to capture usage statistics. They need only a little chopping and sauteeing.

Here is a vision to steal for Wikipedia: Show how much interest this topic has garnered historically.

We’d like to be able to see, at a glance:

• How much traffic has this article been getting?
• What is the popularity trend? Spiking? Declining? Flat? Over the short and long term?
• How much editing activity has it been receiving?
• How does it rank among all searches?

I’ll leave it to as a design research topic to play out how this might look. (Feel free to forward mockups my way.)
A couple of other requirements:

• The data would likely have to be normalized against the growth of general Wikipedia traffic so as not to confound the message.
• We’d want this information to be available proactively (without clicking) but without interferering with the content of the article.
• It’d be nice to be able to drill in deeper to discover the next level of information such as where in the world do people care about a given topic?

[What other interesting types of questions could be gleaned from analyzing Wikipedia traffic?]

Also: wouldn’t it also be nice to have a set of robust APIs, so researchers could go wild analyzing the cause and effect of human curiosity over time and geography? Software architects: how might you build such APIs and hook up a scalable traffic monitoring system to the open source MediaWiki (the technological foundation of Wikipedia)?

See also

• Google Zeitgeist, a form of meme tracker, scrubbed for your protection, that carries the fascinating message that pop culture is really popular. (And that I’m really out of touch with it.)
• Wikipedia statistics (such as they are)
• Overture keyword searching – enter keywords, see how popular they are with search engines today. (Then contrive ways to pepper your articles with the most popular phrases like “sharon stone” or “britney spears sculpture” or daylight savings”, in order to attract hits.)
• Google Trends is interesting. It lets you look up any search phrase, see its popularity over time and what geographies are most into it. The Colbert Report searches are bigger in Canada than in the USA. Digg is on the rise. Searching for Sex is especially popular in Islamic countries. [Update 5/16/06]

In tabbed browers, when I open a link in a new tab it loads the content in the background, without disturbing the current page. Nice.

But when I select “open this link in a new window” to glance at what the link points to, Safari and Firefox both open the new window over the current one, obscuring the curent page. It would be kinder for them to notice the copious space next to the open window and fill the new window there, so I can continue reading the original page while it does its business.

Here’s a quick & dirty design to steal: When opening a new window, look around for unused areas of the screen and put it there if possible. Minimize overlap on existing windows. This applies to most multiple document apps, not just browsers.

The point is even more general. Most apps and OS’s are not yet enlightened when it comes to huge monitors. Windows XP has an archaic “maximize” button that offers up the entire screen to the lucky window. This was reasonable in the era of the 15″ or 17″ monitor, but most apps have no clue about what to do with the acres of real estate on 20″ or 24″ displays. (Don’t get me started on the MDI/SDI mess.)

On the Mac things are a bit better because the window zoom button is not sworn to fill the whole screen no matter how gargantuan. A zoomed Safari window, for example, will take the full screen height but will contain itself to a reasonable width. But these common sense limits are not universal. An MS Excel 2004 spreadsheet with only a couple of columns of data will nevertheless fill the entire screen when maximized, obscuring all other windows and forcing the UI overhead task of resizing the window.

The point even extends to web sites. Some of the most common web sites also take the lowest-common denominator approach, scrunching their content into a dense block even for users with larger monitors:

Yahoo.com on Windows XP, maximized on a 24″ window: not the most effective use of whitespace.

Vision to steal:The OS vendors need to describe and demonstrate best practices for making use of big screens and multiple monitors.

Design research topic:What should those best practices be? We need to account for different classes of applications including web pages that must scale between tiny and huge displays, “power apps” like photo, video, music editing that establish self-contained workspaces, and so forth. How can window management overhead be minimized? What is the best approach to floating palettes of tools? How should they work with multiple monitors? How should they adjust when monitors come and go?

Apple made Alt-tabbing prettier, but they didn’t fundamentally improve the interaction. So Alt-tabbing has remained fundamentally unchanged for about fifteen years. Here is a refinement on the micro-interaction, that applies to both Macs and Windows.

What’s to improve? The issue is that using Alt-Tab to switch between apps is mentally indirect. When using the mouse, the user can just point and click, effectively saying, “Go there, assuming the target window is visible. Very direct.

Alt-Tabbing requires more mental effort. The user must first figure out which app owns what they are looking for, then locate the app’s corresponding ‘s icon, then tab to it. It is not a visual process of just pointing to what you want. This is not a big deal if you only have a couple of apps open. But today’s fat PCs encourage the user to use many apps simultaneously. And if those apps are similar and confusable, such as with multiple text apps, or graphics apps, or web apps, it can be a real distraction to the work at hand.

Design to steal

When the user Alt-Tabs between apps, bring forward each app in succession, dimming back the rest.  Users can Alt-Tab until they see the window they are looking for.

1) Starting point – App A is the front-most app. Windows B1 and B2 belong to the same app. App C is hidden or minimized. App D is visible.

2) User presses Alt-Tab; windows B1 and B2 come forward because they belong to the same app. Everything else dims back.

3) User presses Alt-Tab again; app C becomes temporarily visible, from its minimized or hidden state.

4) User presses Alt-Tab again; app C hides again and app D comes forward.

5) User releases the Alt key; app D is now front-most.

This approach is more direct and visual — users no longer need to mentally map what they are looking for to the application and its icon.

The windows for hidden apps or minimized documents become temporarily visible as the user cycles through them.

Refinements and variants

Smoothing it out – This solution may cause a lot of inelegant flash as windows come and go. One way to smooth this out is to experiment with bringing forward and brighten the next app instantly (because we must minimize UI friction).  Then fade the others back, over the course of a half second or so. This would help make the transitions feel more organic and gratifying.

Switching between documents – In the course of juggling work, one is as likely to need to switch between documents within an app as between apps. Despite these being sister tasks, document switching is done very differently than app switching. On the Mac, the user presses Alt-` to cycle among documents.  Each keystroke reveals the next document.  Unlike app switching, no Alt-Tab UI is shown.

How might document switching and app switching be blended? What if the document cycling worked analogously to app cycling? Instead of merely bringing document apps forward, Alt-` would reveal a row of document miniatures.  Successive key presses would step through these.  Releasing the Alt key would bring the selected document to the forefront.

[Update 7/27/09 Here is a mock-up of the idea that uses the style of Exposé]

The user then gets to navigate to a document using either of the two strategies: the old way, by Alt-`ing and releasing until the document is frontmost, or by keydowning on Alt-` to see the line up of available documents spotting the desired document in the line-up and then repeating the Alt-` keystrokes until it is highlighted, then releasing.

Combining both ideas: the user could get to any document by first Alt-tabbing to the right application, then Alt-`ing to the right document.

Follow-up Research work

This is not the end of the story with updates to be made to application switching. Here are some related problems.

• Apple’s vaunted Exposé feature exists to solve the same needs and scenarios of switching between windows, yet it does so in a completely different way. It scatters all the visible windows out on the desktop in miniature, so the user can select one. This s fine if you have few windows open, but not if you have many. It’s just too much to scan through a dozen windows, finding a needle in a haystack. It’s even more work if the windows are splayed out on a large, modern display. And in a frustrating oversight, your efforts are in vain if you’re looking for an app you happen to have hidden or a document you minimized. How might Exposé, app cycling, document cycling and even the Dock be consolidated into a simpler, coherent, continuous system? [See Blending it in versus bolting it on, and Alt-Tabbing for part of the answer.]
• On Windows, how can app cycling be melded with some improvement on the painful system tray? How can document switching be added to the mix?
• This proposal works with the fundamental conceptual model of applications owning documents. Is there a better way?

Related questions

Does anyone know what logic Mac OS X uses to determine the order of the icons in its Alt-Tab UI? The first two are the last you used but the rest seem to be in a random order. Is this a bug or a feature?  [7/27/09 This was answered in a comment to What is the alt-tab order?]

See also:

• Blending it in versus bolting it on, and Alt-Tabbing
• What is the alt-tab order?

[Updated 2/22/06 to clarify the document Alt-` idea]

[Updated 7/27/09 to further clarify and add mock-up of the document switching UI]