Revisiting the Demoscene: From Amiga Dreams to AI-Powered Reality
 
Credits: Story by Peer. Copy-Editing by Claude Sonnet 4.5
 
Memories from my Amiga Years
 
Those were the days. Sitting in front of my Amiga 500 in the early 1990s, after getting some floppy disks from a friend, I would watch in fascination as the latest competition demos unfolded on screen. These were not just programmes; they were digital art, pushing hardware to its absolute limits in ways the engineers at Commodore never imagined. The demoscene represented something unique in computing history: a subculture dedicated purely to creative technical expression, with no commercial objectives. Demo groups spent countless hours crafting audiovisual experiences that served no purpose beyond demonstrating skill and pushing boundaries.
 
The Party Scene
 
The demoscene revolved around demoparties, gatherings where groups competed to create the most impressive audiovisual displays. The Gathering in Norway and Assembly in Finland (both beginning in 1992) became legendary venues, though smaller parties had existed since the late 1980s. These events brought together hundreds, then thousands of enthusiasts who would spend days coding, socialising, and watching competitions on massive screens. The atmosphere was electric. Groups would work frantically to finish entries, often completing them minutes before the deadline. When competitions began, the hall would fall silent, then erupt in cheers depending on what appeared on screen.

Competitions were organised around strict size restrictions. The 64k intro limited entries to 64 kilobytes, while the 4k intro allowed only 4 kilobytes. To be clear, these are kilobytes, not megabytes. A 64-kilobyte limit is remarkably lean, considering that programmers had to generate dazzling graphics and sound. Even leaner still, 4 kilobytes is roughly the size of a single empty email. Within that tiny sliver of data, demosceners use procedural generation to 'unfold' entire digital worlds. Historically, size-restricted intros emerged from the cracking scene, where small programs, often serving as audio-visual signatures, were squeezed into leftover space on floppy disks. By the early 1990s, however, demoscene parties had adopted size limits as formal competition categories. While the ‘full demo’ category allowed more freedom, it still demanded extreme optimisation. These restrictions forced demosceners to employ procedural generation, executable compression, and the kind of algorithmic creativity that still impresses today.
 
The Amiga 500 Advantage
 
The Amiga 500 dominated the demoscene due to its revolutionary architecture. At its heart sat a Motorola 68000 CPU running at 7.16 MHz on NTSC or 7.09 MHz on PAL, but the real magic came from its custom chipset. The Agnus chip managed memory and housed the Blitter (for rapid graphics operations) and the Copper (a coprocessor enabling cycle-accurate screen effects). Denise handled video output, supporting up to 32 colours from a 4,096-colour palette in standard modes, with special modes like Hold-And-Modify (HAM) displaying all 4,096 colours simultaneously. Paula provided four-channel, 8-bit PCM audio, meaning four independent sound samples could play simultaneously.

This was revolutionary. The chipset could handle DMA (Direct Memory Access), meaning graphics and sound could proceed without burdening the CPU. The Amiga 500 originally shipped with 512 KB of Chip RAM, but it was commonly expanded to 1 MB via the trapdoor memory expansion, which became the de facto minimum for serious demos and games. This integrated design meant smooth graphics and multichannel audio with minimal CPU overhead - a stark contrast to IBM PCs, which relied heavily on the CPU for basic VGA graphics and simple beeper sound. The Macintosh (Mac), whilst offering a polished GUI, lacked dedicated hardware for real-time animation and audio. For demosceners, the Amiga was perfect.

The full Amiga story can be found here:
 
A Personal Favourite
 
One demo that captivated me was the RSI Megademo by Red Sector Inc., released in September 1989. It took an eternity to load from floppy discs, but the wait was worthwhile. Consisting of two 3.5-inch floppy disks, its size was a whopping 2.88 megabyte. The demo showcased vector graphics, smooth scrollers, and copper effects, all synchronized to music by Romeo Knight. What made demos like this special was not just the technical achievement, but the creativity. Groups developed signature styles, and the best demos demonstrated personality through code. A huge collection of these demos is now available on YouTube, preserving this digital art form.

Here are links to some iconic demos from those days: The Amiga demoscene is still very active with multiple large gatherings every year. Here are some newer competition entries I picked at random:
 
From Dream to Reality
 
Since my first exposure to demos, I have dreamed of creating one myself. I made several attempts in Java - lacking advanced assembler knowledge - but never progressed beyond simple scrollers, star fields, or rotating cubes. Then came AI-powered coding tools. Using Gemini and DeepSeek, with Claude helping fix errors, I have finally achieved what I could not alone. The AI tools bridge the gap between creative vision and technical implementation. The tools have changed, but the wonder remains. Whether through hand-coded assembler on an Amiga or AI-assisted development in a modern browser, the demoscene's core principle endures, pushing boundaries and creating beauty from constraints. I finally understand what those demo groups felt when their creations appeared on the big screen.

Here is my Old School Amiga Style Stormy Weather Mega Demo, vibe-coded in Python and run on a gloriously outdated PC. Note, that the below is a video, so it does not help if you press the space bar to get to the next scene :0). The music (elimination.mod) is coming from The Mod Archive. It was composed by Lizardking (Gustaf Grefberg).
 
Warning: This video contains flashing images which may affect photosensitive viewers.
 
 
It's even better, if you watch in full screen!
 
Repository
 
  • stormy-weather-mega-demo_0.2.8 (08/02/2026) [code]
To download the source code from GitHub, click the "Code" link above. This will take you to the correct section of the GitHub repository. In the window that opens, click the downward-facing arrow in the top-right corner to download the zip folder containing the source code and instructions for running it. The folder will be saved to your default download location.
 
 
Supplementary Material
 
 
 
Back to Top
 
 
When Agency Turns Against Itself: A Paradox of Intelligence and Self-Destruction
 
Credits: Based on a discussion between Peer and Claude Haiku 4.5
 
Have you ever wondered or discussed what "agency" actually means? I recently discovered a fascinating open access book called An Evolutionary Story of Agency: How Life Evolved to Act on its Own. This book examines how biological agency has evolved throughout life's history, exploring its origins, selective advantages, and increasing complexity across different species from bacteria to animals.

The paradox that really got me thinking is this: if biological agency evolved to help organisms thrive within their ecological systems, why do we - perhaps the most agential species ever - use that capacity to dismantle the very environment we depend on? To explore this head-scratcher, I fed the book to an LLM and we had a conversation about what the authors' framework of biological agency might tell us about human behaviour in our current predicament. What follows is what we worked through together.
 
Biological Agency
 
The authors of the book state that "biological agency is an adaptive interactive process performed by an individuated self-producing and reproducing system (an organism) that modifies its environment in ways that support its maintenance, either as an individual or as part of a specific lineage."

Key Characteristics of this definition
  • Adaptive interaction: Actions respond to environmental conditions and are flexible
  • Individuated system: The agent has clear organismic boundaries
  • Self-producing and reproducing: Self-maintenance and reproduction are essential
  • Environmental modification: The agent actively changes its surroundings
  • Functional outcome: Modifications support the agent's survival or lineage continuity
  • Circular relationship: The agent is both initiator and beneficiary of its own actions
  • Autonomy requirement: The agent sustains itself through its actions; failure means cessation
The authors emphasise that agency operates through action-detection loops: reversible processes continuously modulated by sensory feedback. This distinguishes it from constitutive processes (like growth or development), which are typically irreversible and form part of an organism's fundamental organisation rather than its interactive behaviour.
 
The Human Problem: Agency Gone Wrong
 
The book's definition focuses on agency as a biological capacity - the ability to act on one's environment in ways that support one's own maintenance and reproduction. By that definition, humans certainly have agency. The question is whether our contemporary environmental destructiveness represents agency well-exercised or agency misdirected.

The authors would likely argue it's the latter. Their framework suggests agency evolved to support an organism's survival within ecological systems. When human actions systematically undermine the conditions necessary for our own survival - degrading soil, water systems, climate stability, biodiversity - we are engaging in something paradoxical: we're using our agency against our own long-term interests.

A few observations:
  • Short-term vs. long-term maintenance: The definition specifies agency supports maintenance "either as an individual or as part of a specific lineage." Many human environmental destruction stems from prioritising immediate individual or generational benefit over lineage continuity. This is arguably a failure of agency, not its absence.
  • The complexity problem: Humans have developed such complex technological and economic systems that the feedback loops between our actions and their consequences have become obscured. We act without clear sensory detection of the damage - a breakdown of the action-detection loops the authors identify as essential to agency functioning well.
  • Collective agency confusion: Modern human activity often involves diffused responsibility across institutions, making it unclear who the "agent" actually is. This differs from simpler organisms where individual agency and outcome are clearly linked.
Conclusion
 
Rather than humans lacking agency, it is more accurate to say we are exercising it destructively or short-sightedly, which raises uncomfortable questions about whether intelligence necessarily improves our use of the capacity for purposeful action.
 
 
Back to Top