CTP Madness Part Eleventy Gazillion

I needed to install Visual Basic on my development Tablet but the VS2005 RC installer blew up when I tried to do that, so I ended up uninstalling Whidbey, but before I could install the RC, I also had to uninstall SQL Server 2005. And then, when everything gets finally re-installed (after, like 8 hours of the machine of uninstall, install fails, uninstall something else, reinstall, etc.) my defect tracking software isn’t working anymore, since apparently it was connecting by way of my now-uninstalled Yukon .


Gawd, I can’t wait for this stuff to ship.

Googlieve: v. tr. To have faith in a Google return

To accept as true the results of a prominent Google return: I googlieve the wavelength of red light to be 650nm. See also: googlief.

Googlief has a complex relationship with belief, as one might accord a googlief of something like wavelengths a higher belief than one’s vague recollections, but on the other hand, one might accord low credibility to a googlief that “the most famous man who ever lived” is former FEMA director Michael Brown (a googlief I doubt will survive the month).

(Historical note: 10/10/05 12:24 Hawaii Standard Time, Google says there are no documents with the word “googlief” or “googlieve” in it.)

Julie Lerman’s “Persisting Ink on the Web” article is a must-read

"Persisting Ink on the Web" walks through each of the following tasks:

  • Transferring Ink to another Ink-Enabled Control on the Same Page
  • Transferring Ink to an Ink Control on Another Page in a New Browser Window
  • Moving Ink to Another Page in the Same Browser Window
  • Transferring Ink to Another Page as a GIF
  • Storing Ink in an XML File on the Web Server to Be Used at a Later Time
  • Storing and Retrieving Ink from a SQL Server Database
  • Sending Ink to a Web Service
  • Surviving a Postback

> >

Don’t Forget: www.acehaid.org

Things I’ve Been Noticing About Spam

Looks like either they’ve broken the CAPTCHA methods used by das Blog, or people are hand-posting comment spam.

I think that email spam may be using keywords derived from my blog as part of their randomly generated subject lines. That’s a clever way to get around any Bayesian filter.

I assume that when not randomly generated, if you see the same or very similar spam subject lines, they’re effective at producing click-throughs. If so, that’s depressing.

casey calls BS on ‘go code something’ as a way to employment

casey chesnut, whose propensity for lower-casing his name ticks off Word auto-correction, says that it is a myth that the best way to be hired as a programmer is to code something cool. casey’s right. He’s proved by his Web postings that he can write: neural nets to defeat CAPTCHA, machine vision applications, speech-based interfaces, mobile code, etc. At the very least, it’s clear that when given a problem he’s not going to be sitting around, stumped and discouraged.


Cool doesn’t work: once upon a time, I wrote a column called “Expert’s Toolbox” for a magazine called “AI Expert.” I published stuff like the first genetic algorithm in C++, the first fuzzy logic engine in C++, and the second C++ neural network (missed being first by 3 months! Darn it!). At the time, this was considered cool stuff. I can’t quite say it never made me a dime because my neural network code was used in a mumble mumble developed by mumble mumble and apparently used until at least the late-90s. But essentially, the cool stuff was worthless. Then, in the mid-90s, I programmed what was probably the first profitable non-pornographic Website (a registration system for the Software Development conference – one of the first 10,000 sites on the Web), wrote the first technical article on Java, the first article on servlets, and developed the first XML-driven Web site. Did I make any money off of that stuff? A little, but not much.


You want to know what made me money? Any of this cool stuff? Oh no. Reservation systems. Several years ago, when XML expertise was still relatively rare, I was hired to do what we’d now call the “Web Service” portion of an airline / hotel / car reservation engine targetting medium-sized corporations. Once I got a reputation in that field, the phone rang off the hook.


  • Every business in the world must have a computerized inventory-management system.
  • Inventory control is domain-specific. If your inventory is seats, you can’t use a system designed to sell books. If your inventory is pencil leads, you can’t use a system designed to sell slots in a marina.
  • The best way for an individual or small company to make money in software is to develop inventory control software for a niche. You’ll make money selling it, but you’ll make a ton more customizing and supporting it.


The problem with that is that from a coding standpoint, it’s insanely boring. Believe me, once you’ve solved all the variations of “What if the employee wants to add a lay-over in the middle of the trip and bring a companion along using frequent flyer points and upgrade using their credit card?” you really, really miss watching genetic algorithms co-evolving.

Query refinement generation via neural nets

Scoble laments that when asked “new york hotels” search engines do not know the difference between “hotels named new york ” and “hotels in new york ” (or, to some extent, “new hotels in york ”). Scoble wants the initial search-engine return to include questions intended to refine the search. Danny Sullivan agrees, saying that Ask Jeeves had such a thing, but the cost of humans creating relevant questions was difficult.


Here’s the solution, I think. You just throw a bunch of question templates “Are you looking for the history of X?” “Are you looking for reviews of X?” etc., hook them up with random Bayesian connections, and update them as necessary.