James R. Shiflett worked at the Houston Advanced Research Center (still going today) in The Woodlands, Texas, where from the 1980s his job was administering the centre's NEC SX-2 — a vector supercomputer announced in 1983, roughly 1.3 GFLOPS peak, the kind of machine that filled a room and was booked by the hour.1 His work there was administrative. The interesting part happened after hours.
I'd do my administrative trick during the day, and at night I'd sit on my computer and think up new algorithms to add to Dazzle. James R. Shiflett, quoted in the Houston Chronicle4
He called it a sort of therapy. That is worth sitting with for a second: the most popular kaleidoscope program of the DOS era, a thing that ended up on hundreds of thousands of screens and in a feature film, was somebody's way of unwinding after work.
He was a devout Christian, and he signed off accordingly — more on that at the end.
Dazzle was written in Turbo C, with the image-generation cores dropped down into x86 assembly.1 It supported EGA, VGA and IBM's 8514/A, and shipped both as a standalone DOS executable and as a TSR that fired on a keyboard-idle timer.
How many patterns did it have? You will see “about thirty algorithms” repeated in a lot of places, this site's own earlier draft included. It is worth being straight about that: the primary source gives no number. Shiflett described his own engine this way:
The image engine has numerous primary image drawing algorithms, most of which have at least two styles of presentation, many of which have multiple internal drawing variations. J.R. Shiflett, DAZZLE 5.0 package description2
Numerous, not thirty — and the structure he describes (algorithms × presentation styles × internal variations) means any single count is a simplification anyway. The Wikipedia article gives no figure either.1 If you want the real answer, the binary is preserved and you can go and count.
The signature was kaleidoscopic mirrored symmetry — plot a shape once, reflect it into every octant — and particle simulations, moved between with fades, dissolves, scrolls, wraps and melts. The fireworks, he said, came from watching an explosion scene in Star Trek.1
This is the part worth admiring. Dazzle looked impossibly smooth on machines that had no business being smooth, and it did it by barely moving any pixels at all.
Palette rotation. In VGA's 256-colour modes a pixel is not a colour, it is an index into a 256-entry hardware palette. Rewrite that palette and the entire screen changes colour in microseconds, without touching a single pixel of video memory. Draw a static field once, then animate purely by cycling the palette — the motion lives in the DAC.
Hardware panning. Ask the CRT controller to start reading from a different offset and the whole image scrolls, again without redrawing.
Mode X. The unchained planar 256-colour modes Michael Abrash popularised in Dr. Dobb's in 1991, which bought page flipping and faster fills.
Cheap geometry, mirrored into symmetry, coloured by a palette that never stops walking. That recipe is the whole aesthetic, and it is the thing JellyDazzle is chasing thirty-six years later on hardware that could brute-force it a million times over — but brute force is not what made it beautiful.
The source of Dazzle was never released. Nothing here is a port. Nothing was disassembled, decompiled or reverse engineered. JellyDazzle was written from scratch, on completely different hardware, in a completely different decade, chasing the same feeling — and the technical lineage above was reconstructed from published interviews, magazine reviews and the documentation Shiflett shipped with the program.
If you want the real thing rather than the homage, you should go and run the real thing. It still works.
There does not appear to be a single photograph of James R. Shiflett anywhere on the public web. No interview footage. For a man whose work ran on more screens than most programmers will ever reach, the record is remarkably thin — a Wikipedia article, some magazine quotes, a shareware README, and forty screenshots on archive.org.
I could not establish whether he is still with us. I looked, and found nothing either way, which is its own kind of answer about how software history gets kept.
If you knew Dazzle, or knew Jim — if you have a photograph, a letter, a registration card, a memory — I would genuinely like to hear from you. This page will be updated, gladly, and with credit.
Everything above is either linked inline or listed here. Where I could not source a claim, I have said so on the page rather than quietly repeating it. Items 3–17 are the contemporary press and primary documents catalogued by the Wikipedia article; I have not personally read the paywalled or print-only ones, and have marked what rests on them.