April 21, 2026
When pixels gaslight
Expansion Artifacts
Your pics can lie: Internet splits over 'compression' chaos
TLDR: The piece explains how shrinking files can subtly change them—even swapping numbers in scans—and how those flaws can also help detectives verify edits. Readers split between meme‑loving nostalgia and trust fears, with one tying it to AI writing prompts, asking if our feeds are now artifact soup.
The story lays out a simple, slightly scary truth: to make the internet fast, we shrink files—and sometimes the shrinking mangles reality. The crowd latched onto the wild bit where a Xerox scanner once “helpfully” swapped numbers on a floor plan. Cue panic posts about rent magically shrinking next, and “trust no JPG” becoming the meme of the moment. Fans of glitch art cheered, claiming compression “mistakes” gave us deep‑fried memes and datamoshing beauty; archivists and engineers clapped back: “funny until your medical record gets ‘quilted’.”
The drama split into two camps. One side shouted efficiency is worth it—your ears won’t miss what MP3s shave off and your eyes don’t care about perfect pixels. The other side went full reality check—if files can quietly morph over many saves, how do we trust photos, invoices, even courtroom evidence? A sleeper subplot: one commenter veered into AI land, linking Mike Caulfield’s post on prompts that force chatbots into styles, comparing it to a new kind of “compression of writing”—another layer of invisible edits shaping what we see.
Jokes flew fast—“my six‑pack got compressed to a one‑pack,” “PSNR of my dating pics after filters: oops”—but the punchline hit hard: is compression harmless polish, or a quiet gaslight built into our digital lives?
Key Points
- •Compression enables efficient storage/transmission by discarding data humans are unlikely to perceive, underpinning platforms like YouTube, Spotify, and Instagram.
- •Specific codecs exploit perceptual limits: MP3 (auditory masking), JPEG (visual structure over detail), and MPEG (temporal redundancy via key frames and motion).
- •A 2013 case showed Xerox WorkCentre scanners using JBIG2 could silently substitute numerals, altering documents like building plans.
- •Repeated encoding cycles cause cumulative degradation (“generation loss”), illustrated by a PSNR drop from infinite to 14.59 after 10,000 cycles.
- •Compression artifacts act as meta-information for digital forensics and have been adopted as creative aesthetics (e.g., datamoshing, glitch).