# Origin
This research folder grew out of a short conversation with an llm
in May 2026. I had spent a day or two porting 2048 to Nushell as a
coding exercise and asked, half in passing, whether the game's pull
was actually written about anywhere.
That question opened the door to Threes!, the 1024 lineage, the
clone wars, the addiction-mechanics literature, Universal Paperclips,
prestige loops, Cookie Clicker, and the question of whether 2048
just strips Threes! down to its raw addictive core.
## What I found, roughly in the order I found it
**Threes! had a 14-month design process.** Vollmer and Wohlwend
iterated against dozens of variants. They prototyped the pure-
doubling mechanic that 2048 would later use, and rejected it on the
grounds that it made the game shallower and more compulsive. They
were trying to ship a game you'd play for years, not binge for a
weekend.
**Apple named Threes! its iPhone Game of the Year for 2014.** The
award and the eclipse happened in roughly the same window. Threes!
was being garlanded by the platform that hosted it while its
audience was migrating, week by week, to a free web clone of a
clone.
**The clone of a clone.** Cirulli wrote 2048 over a weekend in
March 2014, riffing on 1024!, a free Veewo Studio mobile game that
had already stripped the 1+2=3 rule out of Threes! and made
everything pure doubling. So the lineage isn't Threes! -> 2048.
It's Threes! -> 1024! -> 2048. And each link in the chain went more
viral than the one before it -- a $2.99 paid app, then a free
mobile app, then a free open-source page on the web -- precisely
because each link stripped a bit more friction.
## The recursion
I only knew about Threes! because I had pulled on the string of its
clone. The
Nushell port I had spent the weekend building was downstream of
Cirulli's open-source repo, which was downstream of 1024!, which
was downstream of fourteen months of careful design I had never
heard of. Three rungs of strip-and-spread sat between me and the
people who actually invented the thing I was reimplementing.
Sudoku and the crossword had filed themselves in my head as
*eternal* -- genre furniture with no author. I had quietly filed
2048 the same way. Discovering Threes! was discovering that 2048
*had* an author, and that the author had been wronged in a
particular way that I, building cheerfully on the clone, was a
small ongoing instance of.
The chapters under this folder -- the history, the design ethics,
the psychology, the incrementals genre -- all came out of that one
conversation and the reading it sent me to. I used the llm the way
I'd use any other research tool: it surfaced threads, named books
and papers I hadn't heard of, and asked the next obvious question
faster than I would have. The reading, the cross-checking against
sources, and the synthesis on these pages are mine. The discovery was that I had been a perfect example of the dynamic
I was about to write about.