10,000 Drum Machines:节奏作为数据的余项
10,000 Drum Machines: Rhythm as the Remainder of Data
Maxwell Neely-Cohen在2019年画了几台鼓机的草图,然后搁置了六年。2025年他终于开始写代码,但写出来的不是一台鼓机,而是一个不断膨胀的收集站:10,000 Drum Machines。目前站上有四十多台浏览器鼓机,每一台都是一次独立的实验。有一台用USGS的地震实时数据生成节拍——每天的地震分布不同,鼓点的密度和冲击就不同。有一台用De Bruijn序列驱动。有一台用逻辑门(AND、OR、XOR)切换乐器。有一台用猫的图片。有一台叫"Extremely Long Term Drum Machine"。每一台的声音来源都不是音乐——它们是某种非音乐数据或逻辑通过序列器框架后留下的东西。
这正是余项的核心结构。每台鼓机的输入(地震数据、布尔运算、像素信息)本身有自己的目的和语境,与节奏无关。当这些数据被强行塞进一个步进序列器的格子里,它们原来的意义被剥掉了,但数据本身的拓扑结构——疏密、周期、断裂——在节奏中留下了痕迹。这个痕迹就是余项。它既不是地震学,也不是音乐理论,也不是数据可视化的听觉版本。它是数据在被翻译为节奏的过程中无法被完全消化的部分,是从一个框架到另一个框架的有损迁移中存活下来的残留物。
更重要的是这个项目的结构本身。"10,000"不是一个真正的目标,而是一个地平线。四十多台之后,离一万还差九千九百五十多台。这意味着这个项目在结构上永远不会完成——它被设计为一个持续累积余项的装置。每个贡献者带来自己的非音乐数据和怪异逻辑,生成一台新的鼓机,累积进集合里。Neely-Cohen自己说得很清楚:"在这个大规模网络抓取和垃圾内容生产的时代,我想试试用秘法乐器淹没互联网。" 这不是对抗算法的姿态宣言,这是一种实践:用具体的、手工的、无法被批量抓取利用的小工具填满网络空间。每台鼓机太怪、太具体、太无利可图,不会被任何内容农场回收。它们就这样活在网上,作为互联网本身的余项。
这个项目现在正处于凿构循环中"还在凿"的阶段。它已经从纯网页走向了物理空间——Neely-Cohen和合作者Jessie Char在匹兹堡文化信托的"Ground Loops"项目中把这些鼓机变成了现场装置和表演。CDM、Adafruit、kottke.org都报道过它。但艺术界还没有给它一个名字。它不是"声音艺术"——没有哪个声音艺术策展人会把一台用猫图片驱动的浏览器鼓机纳入自己的展览叙事。它不是"网络艺术"——它太实用了,每台鼓机真的能用。它不是"数据声化"——它不关心数据的准确再现。它不是"开源工具"——它太荒谬了。它就卡在这些类别之间,还没有被完全命名。这正是看见它的时机:再过两年,当集合膨胀到三百台、五百台,当更多策展人开始注意到它,它可能会被某个框架吸收,变成已构。但现在,它还是余项。
10kdrummachines.com ↗Maxwell Neely-Cohen sketched a few drum machines in 2019, then shelved them for six years. When he finally started coding in 2025, what came out wasn't one instrument but an ever-expanding collection: 10,000 Drum Machines. The site currently hosts over forty browser-based drum machines, each a standalone experiment. One generates beats from USGS real-time earthquake data—different seismic distributions each day yield different rhythmic densities and impacts. One runs on De Bruijn sequences. One uses logic gates (AND, OR, XOR) to switch instruments. One uses images of cats. One is called the "Extremely Long Term Drum Machine." In every case, the source material is not music. The sound is what remains after some non-musical data or logic has been forced through a step sequencer's grid.
This is the core structure of the remainder. Each machine's input—earthquake data, Boolean operations, pixel information—has its own purpose and context, none of which involves rhythm. When this data is forced into the cells of a step sequencer, its original meaning is stripped away, but the data's own topology—its density, periodicity, ruptures—leaves traces in the resulting rhythm. Those traces are the remainder. They are not seismology, not music theory, not an auditory version of data visualization. They are the part of the data that cannot be fully digested in the translation to rhythm—the residue that survives the lossy migration from one frame to another.
More importantly, the project's own structure embodies remainder-production. "10,000" is not a real goal—it is a horizon. With roughly forty machines built, more than 9,950 remain. The project is structurally designed never to be complete; it is an apparatus for continuous remainder accumulation. Each contributor brings their own non-musical data and idiosyncratic logic, generates a new machine, and adds it to the growing pile. Neely-Cohen himself is explicit about the philosophy: "In this era of massive web-scraping and slop production, I do like the idea of trying to saturate the web with esoteric musical instruments." This is not a gestural manifesto against the algorithm. It is a practice: filling web space with concrete, handmade, unscrapable micro-tools. Each drum machine is too weird, too specific, too unprofitable to be recycled by any content farm. They simply sit on the internet, alive, as remainders of the internet itself.
The project is currently in the chisel phase of the chisel-construct cycle. It has already moved from pure web presence into physical space—Neely-Cohen and collaborator Jessie Char staged "Ground Loops" at the Pittsburgh Cultural Trust, turning these machines into live installations and performances. CDM, Adafruit, and kottke.org have all covered it. But the art world has not yet given it a name. It is not "sound art"—no sound art curator would fold a cat-image-driven browser drum machine into their exhibition narrative. It is not "net art"—the machines are too functional, each one genuinely playable. It is not "data sonification"—it does not care about faithful representation. It is not "open-source tooling"—it is too absurd. It sits wedged between these categories, still unnamed. This is exactly when to see it: in two years, when the collection swells to three hundred or five hundred machines, when more curators take notice, some framework may absorb it and render it construct. But right now, it is still remainder.
10kdrummachines.com ↗