> Farting around with Amigas in 2026 means actively choosing to make things harder for the sake of making things harder. Making that choice and still outsourcing the bulk of the craft and creative process is like claiming to be a passionate hobby cook while serving professionally catered dinners and pretending they're your own concoctions.
People wanting to explore the use of generative AI for vintage computers is happening not just for graphics but for code too.
I think in the case of code though, it's still interesting because I don't believe there's been any success yet. I hear of people having success with Claude in contemporary settings but it seems to fare less well when working for older computing platforms. There's a reason for that of course and it's worth exploring.
However, it will cease to be interesting as soon as the first person manages to create something substantial. At the point, the scene should probably shun it for the reasons stated in the quote.
There's definitely been success in using generative AI for vintage Computers. Just the other day I got it to produce a bootable floppy for my Amiga 1200. It loads the network driver, uses BOOTP to get an ip address, connects to a server and then downloads code via UDP that it will then execute. I doubt you'll get it doing amazing graphical scenes like you see in the demo scene though.
What I find mind-boggling is the handwave over the rest. "Loads the network driver" - ok, which one? There's no standard network driver, only a specification for writing drivers (SANA-II). Was it a driver for SLIP/PPP over the serial port, or a PCMCIA Ethernet adaptor, or something else? Was it a copy of a driver someone's already written?
Also, it would be madness to try doing this in a bootblock, or insinuating that the bootblock did it. Demo bootblocks take over the hardware and start using their loading routines, eschewing the main AmigaOS, and that's the implication of saying something was done in the bootblock (you have under 1KB of space so the first thing you need is your own loader).
What's much more mundane and normal is a standard bootblock which returns control to AmigaDOS and lets it run the startup-sequence, whereupon you can use normal files, libraries, devices, including a full suite of other people's networking software, including BOOTP (AmiTCP comes with a client) and TFTP (see Olaf Barthel's tftpclient: https://github.com/obarthel/amiga-sana-ii-tftpclient). But it stopped being the "bootblock" that did it as soon as it started AmigaDOS.
That gives you a standard OFS bootblock that returns to AmigaDOS. Mine is a custom bootblock, same DOS\0 magic and checksum format so Kickstart accepts it, but it never enters AmigaDOS.
> What I find mind-boggling is the handwave over the rest
Fair, I should have been more specific. The network driver is the popular cnet.device which is compatible with my PCMCIA ethernet card. It's loaded from fixed floppy sectors.
> it would be madness to try doing this in a bootblock
Agreed, and I don't. It's a multi-stage boot that stays at exec level throughout, AmigaDOS is never started, no process is created, no startup-sequence runs, _DOSBase is explicitly NULL.
The disk is a standard 880K ADF with no filesystem at all, it's just raw binaries at fixed sector offsets. The only ROM libraries used are exec.library, intuition.library and graphics.library for a debug display. Everything else is self-contained on the disk.
So you're right that it isn't the bootblock doing the networking.
I used cursor with a mix of Gemini 3.1 and opus 4.6.
It referenced the Amiga ROM Kernel Reference Manual, appendix C to create a boot block in assembly. It's a raw sector-mapped image, the build process creates a blank adf, which then writes everything at it's fixed offsets and we go back with another tool to patch the bootblock with the right checksum so the kernel accepts it.
I copied that adf to the A1200 so I can then write it to a real floppy.
I no longer drink in pubs but in my neck of the woods, the pubs that specialised in cask ale often had lined glasses.
The problem was that many people insisted on the glass being filled to the brim, because they felt they were being short changed. So it solved one problem but created another.
Yes, there are a few tools that do this. Looking at /bin and the softlinks that are there, the various xz tools do it (unxz, lzcat, etc.). Also, vim. vimdiff and view are just softlinks to vim.
The only difference is that those tools have chosen easy to remember names rather than embedding the arguments as metadata in the filename.
As a generalisation of the idea though, the blog post is neat.
exiftool can embed options to executable name, not only main mode of work like grep/egrep/zgrep — it is main difference. Like running `exiftoo(-k)` is equivalent `exiftool -k`.
I think the split is between people who are in a hurry and those who are not. I'm not in a hurry and so choose not to spend money to get a quicker result.
Taking time to solve a problem myself is pleasurable and I make no apologies for that.
I've heard people say that these coding agents are just tools and don't replace the thinking. That's fine but the problem for me is that the act of coding is when I do my thinking!
I'm thinking about how to solve the problem and how to express it in the programming language such that it is easy to maintain. Getting someone/something else to do that doesn't help me.
But different strokes for different folks, I suppose.
I'm similar, but I do find some natural places where LLMs can be helpful.
Just today I was working on something that involves a decent amount of configuration. It's in Python unfortunately and I hate passing around dictionaries for configs, I usually like to parse the JSON or YAML or whatever into a config class so I have a natural way to validate and access without just throwing strings around.
As I was playing with the code for the actual work that needs to be done, I was thinking what configs I needed and what structure made sense. Once I knew what I needed I gave the JSON to an LLM with some instructions regarding helper functions and told it to give me the appropriate Python code. It's just a bunch of dataclasses with some from_dict or from_string methods on them, not interesting or difficult to write. Freed me up to keep working on the real problem.
"All of them are moving into the direction of "less human involved and agents do more", while what I really want is better tooling for me to work closer with AI and be better at reviewing/steering it, and be more involved."
I want less ambitious LLM powered tools than what's being offered. For example, I'd love a tool that can analyse whether comments have been kept up to date with the code they refer to. I don't want it to change anything I just want it to tell me of any problems. A linter basically. I imagine LLMs would be a good foundation for this.
Any terminal tool like Claude Code or Codex (I assume OpenCode too, but I haven't tried) can do it, by using as a prompt pretty much exactly what you wrote, and if it still wants to edit, just don't approve the tool calls.
One problem I've noticed is that both claude models and gpt-codex variants make absolutely deranged tool calls (like `cat <<'EOF' >> foo...EOF` pattern to create a file, or sed to read a couple lines), so it's sometimes hard to see what is it even trying to do.
"Any terminal tool like Claude Code or Codex (I assume OpenCode too, but I haven't tried) can do it, by using as a prompt pretty much exactly what you wrote, and if it still wants to edit, just don't approve the tool calls."
I'm sure it can. I'd still like a single use tool though.
But that's just my taste. I'm very simple. I don't even use an IDE.
edit: to expand on what I mean. I would love it if there was a tool that has conquered the problem and doesn't require me to chat with it. I'm all for LLMs helping and facilitating the coding process, but I'm so far disappointed in the experience. I want something more like the traditional process but using LLMs to solve problems that would be otherwise difficult to solve computationally.
I’m glad I’m not the only one who’s noticed these seemingly arbitrary calls to write files using the cat command instead of the native file edit capabilities of the agent.
reply