Why it’s interesting
mem_brane gives an autonomous AI agent a visible inner life across four transparent LED screens. One layer carries its current thoughts, while other layers hold older fragments and newly generated dream imagery. Because the displays overlap in physical space, visitors can watch memories and ideas accumulate rather than seeing a single flat video. The work presents a designed artistic system, not a claim that the software is conscious or that its generated memories are factual records.
How to try it
Open Bjørn Staal’s first-party project page and start with the close-up installation video. Notice how text, images, and motion occupy different depths, then compare the wider footage to see how the four screens form one sculptural object. The original installation listened to ambient speech in the exhibition space and later folded transcribed fragments into its memory stream. Viewing the web documentation itself does not require microphone access, an account, or local installation.
Requirements and sources
Staal’s project page documents the installation and its presentation at Pioneer Works in New York in September 2025. A language model produces internal thoughts, a second agent expands visual prompts, Flux generates images, and Seedance turns them into video. Speech transcription supplies delayed traces of nearby conversation in the physical exhibition. The author’s page and the recorded Everything NYC talk support the creator, date, venue, model roles, and current archival access described here. The entry recommends the documented artwork; it does not promise remote control of the original installation.