I just posted a list of errata for a paper of mine at an upcoming conference. My favorite?

“The published paper says the XXXXXXX patent [1] was issued June 30, 2000, but that should be June 20, 2000. (United States patents are issued on Tuesdays, not Fridays.)”

The cal(1) command comes in handy at the strangest times!

The other day I realized that asking LLMs for advice is a lot like casting the yarrow stalks and divining from the I Ching. You can extract meaning from many strange places.

I asked a few chatbots about my “interesting” insight and they all agreed with me.

Yesterday I visited a class on Recommender Systems at UCSD. It was the first lecture of the semester.

I was fascinated to hear the professor tell his class that many people enjoy taking it specifically because recommender-system jobs seem immune to elimination by LLMs—so far, at least. Recommender-system students can get jobs! His class has about a thousand students among its various sections.

Good to know! (Me, I’m retired, and it’s not my field anyway.)

After the professor’s hour of administrivia, I gave a brief guest talk, to practice my ten-minute presentation of a paper I’m giving at a recommender-system conference next week.

@ricci you have a PhD, isn't every room that you wait in a doctor's waiting room?

Everything on my feed right now: APPLE’S NEW OS ADDS THREE NEW FEATURES I’VE BEEN DEMANDING FOR YEARS (BUT I WOULD HAVE DONE THEM MUCH BETTER)

I do think it’s sad that Eckart & Young got their own retroactive DOI but Eliot did not.

I like to treat driving like performance art.

Pro Tip To Authors™: Listing your references can provide unique opportunities for artistic expression!

This morning I played Jerry Orbach’s recording of “Try to Remember” (“/ The kind of September”) from The Fantasticks, as I do every September 1, and danced with my wife. She had no idea what was going on, at first anyway, just like every year.

That is all.

Django is a giant scary dog, but sometimes _very_ lazy! #DogsOfMastodon

I often think that if I were King Of The World™, things would be VERY different—and what that assessment lacks in precision, it more than makes up for in accuracy.

And he reads and likes my (short) tweets!

I just noticed that Jeff Dean follows me.

This morning I started asking my little AI friends (a.k.a. Google Gemini 3) in detail about my recent mishaps with ACM’s TAPS—asking, for example, are the “interesting” caching problems deep inside TAPS well-known and well-understood? Google Gemini agreed I was a genius. Thus emboldened, I continued by showing it some of my amusing edits to classic Warner Brothers cartoons starring my colleague “Wile E. Coyote (Genius)” and it responded in kind by drawing one of its own—which I had not asked for, thereby raising some “interesting” copyright concerns….

I’m having such problems with the ACM…

For the last week and a half I’ve been fighting the ACM’s TAPS system for a document submission to a conference. Today’s the last day, my HTML output is still wrong (one figure is still coming out blank), but I’ve finally discovered that TAPS uses some sort of caching scheme internally, presumably for efficiency, and that’s why changing the input isn’t changing the output. Fun!

I tried sending some research notes of mine through an LLM for comment, not remembering they contained extensive complaints about the same damn idiotic LLM. Its reaction? (And I quote…)

“Getting an LLM to accurately maintain face-texture orientation while tracking a double-speed 240-degree tumbling orbit in a custom coordinate space is notoriously difficult.”

Yeah, yeah, “notoriously” difficult. Can I cite that?

I’ve been reading over a draft of an upcoming CS paper of mine at breakfast today, and I notice that Roger Ebert once (in effect) compared me personally to the Wizard of Oz (the character, not the movie). This will look *great* on my résumé!

One month from now will be the sixtieth anniversary of the first program I ever wrote, in my first programming job and having just read the first few chapters of McCracken’s _A Guide to FORTRAN IV Programming_. Just before I submitted my card deck to the batch queue, my co-workers asked if I thought it would work on the first try. I said, sure, why not? [NARRATOR: IT DID NOT, IN FACT, WORK ON THE FIRST TRY.] I still remember my exact mistake:

X1 = (-B + SQRT(DISC)) / 2.0 * A
X2 = (-B - SQRT(DISC)) / 2.0 * A

This would be a fine question for flunking plenty of today’s job candidates too. I luckily had already been hired.

I recently wrote a retrospective paper for an upcoming conference, describing an old system I built some while ago. I’m exceedingly proud of the following introductory sentence:

“During the design and implementation of [this system], I was largely ignorant of related prior work and proceeded by engineering intuition to ‘invent’ numerous techniques, many of which (it turns out) already existed.”

Feel free to reuse this sentence in your own work!