/Technology

If you open CSDb, the database of the Commodore 64 scene, on any given weekday, you'll find that almost every day something new turns up. A demo, a game, a SID tune, an image conversion tool someone wrote over an insomniac weekend. For a machine that stopped being manufactured in 1994, that's an absurd pace. It's more output than a lot of living platforms manage. Now ask someone outside the scene what gets written for a C64 in 2026 and the answer is almost always the same: games. Small games, nostalgic ones, remakes. I think that's the most wrong idea going around on the subject, and saying so right away saves me the rest of the coffee. I'll leave the games for later, because I don't want to start where everybody starts. I'll start with the demoscene, which is where most of the truly new work is. A C64 demo, for anyone who has never seen one, is a program a few minutes long that shows graphic and sound effects that are impossible according to the machine's official documentation. It isn't a game, it doesn't solve any problem, and it exists only to impress people who know how hard it is. The VIC-II, the video chip, has things the Commodore engineers never meant to allow: opening up the screen borders with a timing trick on the control registers, moving the whole screen pixel by pixel underneath the beam, mixing graphics modes line by line. None of this is in the manual. It was all discovered through trial, error and a patience I can't describe, and then passed from group to group over four decades. What fascinates me is that it keeps moving forward. We're not talking about closed, archived knowledge. New tricks are still being found, or at least new ways of combining the old ones, and modes like FLI, which changes colour on every scanline, have spawned more ambitious cousins the original hardware supposedly couldn't handle. The reason is cruel and simple: the VIC-II steals about forty cycles from the processor on a line of sixty-three, on the so-called bad lines, and anyone who wants to do something extra on that line has to count cycles the way you count coins at the end of the month. Either the code fits, or the screen shakes. There's no middle ground, and I think that's why the thing keeps its prestige. A parenthesis that just came to me. When I was about fourteen, I copied a four-page BASIC listing out of a magazine, line by line, and spent an afternoon discovering that the error was in a DATA statement where I'd swapped a seven for a one. I never forgot it, and I think it was the only time a number taught me humility. It has nothing to do with the subject, I just wanted it on record. Back to it. The demoscene parties are the engine behind all this, and the mechanics are worth understanding. Events like Revision, held at Easter in Saarbrücken, run competitions per platform, and there are groups that spend months preparing an entry for a single presentation of about five minutes on a big screen. The deadline is the fuel. Without a scheduled competition, plenty of things would never get finished, and this is a second-order effect that rarely gets noticed: output doesn't respond to user demand, it responds to a social calendar. A demo exists because there's a party in April, not because anyone wants to watch it at home. Games live in a different system, and I'll get to that. There's also a subculture inside the subculture, the small intros, where the challenge is to fit into four kilobytes, or one, or a few hundred bytes. Both the Spectrum and the C64 have categories like this, and the best ones have an elegance that leaves me speechless and a little irritated, because I know I wouldn't be capable of it. Some people spend weeks shaving a single byte off a routine. It's an extreme sport for people with a programmer's salary. But fine, let's talk about games, since that's what everyone wants. New C64 games exist, and two distinct ecosystems produce them. The first is commercial, in the modest sense of the word. Small publishers like Psytronik, which sells for both the Commodore and the Spectrum, release new games on cassette, disk or cartridge, with a box and a manual and a print run of a few hundred copies. A title like that can cost a few dozen euros in a physical edition, and it still sells out. The second ecosystem is competitions, like the RGCD 16KB cartridge contest, which has been running for more than a decade and forces entrants to make a complete game on a tiny cartridge. It's a good constraint. It produces things that fit in your head, to return to the theme. Some of the results are genuinely great. Mayhem in Monsterland, around 2008, had a level of polish that would have shamed any eighties studio, and it remains the reference for anyone talking about new platformers on the C64. Sam's Journey, from 2019, came out after a crowdfunding campaign and shipped with a cartridge, colours, animations and a soundtrack that made me stop what I was doing. These aren't niche games for the nostalgic, they're games that someone who had never touched a Commodore would find beautiful. That said, and this is where I want to be less friendly, most new games are mediocre, and the explanation is structural. The person making the game is almost always the programmer, the pixel artist and the musician all at once, or else three friends who have never met in person. Level design is left for last and comes out badly. The final ten percent, the balance and the difficulty and the details that decide whether anyone plays for more than ten minutes, never gets done, because the enthusiasm already spent isn't enough to cover it. The guy who writes an eight-direction scroll routine with multiplexed sprites is a genius. The game around the routine is often a technical demonstration with enemies. I've abandoned things like that myself, and I don't like thinking about it. I need to correct myself. I said at the start that games were the wrong idea and now I'm spending paragraphs on them, which is suspicious. The truth is that the distinction I drew between demos and games is more fluid than it looks. Many demo programmers make games for the same hardware, and what they learn in the demo they apply in the game. The audience is what differs: the demo is for peers, the game is for everyone else, and nobody quite knows who everyone else is. I stand by the essential point, which is that the deepest technical work is in the demos, but the border is blurry. On to the Spectrum, which is a different animal. The Spectrum is a weaker machine than the C64 in almost everything. No sprites, no real sound hardware, just a speaker you switch on and off and a video memory layout that forces the processor to do most of the work. The famous attribute clash, where each eight-by-eight block of pixels can only have two colours, defined the aesthetic of a generation and still defines today's. Anyone drawing for the Spectrum draws around that limitation, and there's a whole school of art that treats it as a signature. I like that. I also like knowing it isn't only aesthetics, it's engineering that is still being pushed. For example, engines like Nirvana, by Einar Saukas, allow colour every two lines instead of every eight, which completely changes what you can draw. It's new software, written in 2020 or so, for a machine from 1982, and it makes it look as if the machine had always been capable of that. The same Saukas, with Emmanuel Marty, published ZX0, a data compressor that achieves absurd results with a decompressor that takes up very few bytes. That isn't nostalgia, it's applied research, and research has an owner, has documentation and has a benchmark. On the games side, some teams have specialised in producing in volume. The Mojon Twins, from Spain, built their own engines, like MK1 and the Churrera, to make platform and adventure games with an almost industrial workflow: maps made in an editor on the PC, graphics in a pixel program, and the engine doing the rest. Dozens of games have come out that way, some better than others, and there's a community that translates, improves and distributes them. Publishers like Cronosoft sell physical editions. And there's AGD, Jonathan Cauldwell's Arcade Game Designer, which lets someone who has never written Z80 assembly make a complete fixed-screen game. How many people started that way? Many. How many then moved on to real assembly? Some. It's the on-ramp, and a community without an on-ramp dies in two generations. Here in the Portuguese-speaking world the Spectrum had a particular weight. In Brazil there were the clones, like the TK90X from Microdigital, which came with translated manuals and some ROM differences, and in Portugal the Spectrum was, for many people, the first machine. You can see it in the nostalgia and, to a lesser degree, in the output. I still haven't seen a local scene to match the Spanish one, and I don't know why. Maybe because the Spanish scene organised communities around shared engines, and we stayed with private collections. Now the part almost nobody mentions in the enthusiastic articles: the workflow. Anyone writing new software for these machines in 2026 doesn't do it on a 1984 machine. They write in VS Code, assemble with KickAssembler or ACME for the 6502, with sjasmplus or z88dk for the Z80, test in VICE or Fuse, keep everything in git, and only at the end move to real hardware, when they do. People with hardware use flash cartridges, an Ultimate to simulate peripherals, a DivMMC with an SD card for the Spectrum. The experience is, ironically, much closer to modern development than to the original era. Some lament this, the purists who want to write everything on the Commodore itself. I think they're entitled to, but I don't see the gain. The temptation to accuse those colleagues of posing is strong and I'm going to resist it, partly. For the Spectrum there's also the option of higher-level languages. ZX Basic, also known as Boriel Basic, compiles a dialect of BASIC into Z80 machine code and produces results that would make the BASIC programmers of the day blush. z88dk lets you write in C. And some people use Pascal on the C64 through tools like Turbo Rascal Syntax Error, a name I still find far too funny for a serious tool, and which works. Any of these would exist only as a curiosity, if there weren't people using it to ship product. On artificial intelligence and these platforms. I don't want to go on about it, but my impression is that a language model is terrible at cycle-exact code. It invents instructions that don't exist, forgets that an addressing mode costs one extra cycle, and produces a raster routine that looks right and makes the screen shake. That makes a curious test: if the code has to be right to the cycle, the thing still isn't automatable. I don't know how long that will hold. A year from now I may be eating this sentence. Let's move on to hardware, which has its own economy. SID chips, the soul of the C64's sound, stopped being manufactured decades ago, and the ones on the second-hand market are increasingly expensive and, in many cases, fake or defective. That created a whole sector of substitutes, from boards like the SwinSID and the ARMSID to FPGA implementations, and an endless argument about which one sounds closest to the original. The two original SID models, the 6581 and the 8580, sound different, and there are musicians who choose their machine because of it. That means even new music is tied to a hardware debate that didn't exist in the eighties. There are new machines too. The THEC64, released in 2018 by Retro Games, was an attempt to put the C64 in a little plastic case with games included, and it sold. The ZX Spectrum Next, funded in 2017 with the participation of Rick Dickinson, the designer of the original case, is a Spectrum reimagined in FPGA with an enhanced Z80 and much more memory. It's a new machine that runs old software and has attracted new software, which rarely happens. It's worth remembering the other side of that coin, the ZX Spectrum Vega+ campaign, which promised a handheld console and delivered, years later and amid a lot of distrust, something many backers considered a disaster. I learned from it never to pay up front for retro hardware, and I still don't follow my own rule. There's a little-discussed consequence of all this. When a platform gets new hardware, new software gets a predictable target. The audience for a Spectrum Next game is small, but it's exact: people who own a Next and want games that make use of it. That's rare on these machines. Generally you write for a scattered audience nobody can count. Look at the result. New software for these machines divides, basically, into categories that barely talk to each other. Demos, for peers, driven by parties. Games, for the tiny market of collectors and players, driven by publishers and contests. Tools, which are software for people who make software, driven by those who need them and, often, by the pleasure of making them. Music, which is perhaps the purest and the most forgotten, with composers writing tracks for the SID or for the AY of the Spectrum 128 that almost only they and a few others ever hear. And then a fifth category, the strange things: an IRC client for the C64, a web server, a spreadsheet nobody asked for, "I wonder if it can be done" projects. Those are my favourites, I won't pretend otherwise. Part of what gets produced, honestly, is noise. I've seen ports of Pac-Man, Tetris and Flappy Bird to the platform, repeated to exhaustion, and I confess I've written a Snake under those conditions just to prove I could. But noise is the price of a living scene, and it's worth distinguishing practice from work. Practising is legitimate. The problem is publishing the practice and calling it a game. What intrigues me most, and it's a question I can't resolve, is who does the judging. In these communities there's a culture of criticism I don't find elsewhere: magazines and fanzines, votes on CSDb, detailed comments from people who really understand what was done. The reward for making something for the C64 isn't money, it's a comment from someone who noticed the sprite trick. I once received a comment like that, on something small, and it's hard to explain the effect. It isn't vanity, or not only. But criticism has a hidden cost too. Because the audience is small and knowledgeable, and because the same names show up everywhere, there's pressure to make something technically impressive rather than good. Votes reward the trick. A simple game, well designed and pleasant to play, loses to a demo that puts forty colours on the screen. I see this as a distortion of values, and at the same time I understand it, because I'm the first to applaud the trick. It's a contradiction I keep without regret. One more thing seems decisive for the future: age. Most of the authors are over forty. Some grew up with these machines and spent thirty years without touching them, and came back because they have time, because the kids have grown up, because work left them empty in another way. That gives an enormous dynamism and an enormous fragility. When this generation stops, and it will, I don't know where the next one comes from. There are young kids, I've seen them at parties, but they're few and almost always came through the influence of a father or an uncle. Maybe the on-ramp of AGD and ZX Basic will solve part of the problem. Maybe not. I'll end with something that has bothered me from the start and that I couldn't fit anywhere: the number of people who write software for these machines without ever letting anyone see it. Projects kept on disks, in forgotten folders, on forums that have closed. The visible scene is a fraction of what exists, and always was. Nobody knows how much new software was written for the Spectrum this year, and I think that's for the best.