.png)
There is a view in the retro computing world that using AI is cheating. That if a tool or a port or a piece of documentation was made with help from a machine, it does not really count. I understand where it comes from. This hobby is built on craft, on the hours spent with a disassembler and a cup of cold tea, and nobody wants that devalued.
But I want to say plainly where TatungBytes stands, because I would rather be honest about it than have anyone find out later and feel misled. We use AI. We have used it for the biggest things this site has done. And we intend to keep using it, because our mission is not to prove how hard we can work. It is to preserve, explore and promote the Tatung Einstein, and AI has turned out to be the most useful new tool for that job since MAME.
For decades the only way to put a file on an Einstein disc image was a program from the nineties that needs an old copy of Windows to run. We now have a set of scripts that do it on any modern machine, and they exist because an AI could hold the whole disc format in its head, propose an experiment, read the result, and try again, hundreds of times, without getting bored. The course that teaches Z80 game programming on the Einstein, the only one of its kind anywhere, was written the same way, with every hardware fact checked by running code on the machine rather than copied from a manual that might be wrong. Stephen Dobbs built Albert, the first dedicated Einstein emulator, over a weekend with Claude Code, and said so on the tin. And this week Simone Voltolini brought five games to the Einstein 256, a machine that has barely had five new games in forty years, with ChatGPT and Codex helping him port them from the original source. He credited the tools on the disc itself. Good.
Cheating implies a contest, and there is no contest. Nobody is keeping score. The Einstein does not care whether the person who wrote its new disc tools typed every line or directed a machine that did. The people who own one and want to run their software care that the tools exist and that they work. Every port, every tool, every manual we recover is one more reason for the machine to be remembered, and the world has a limited supply of people willing to spend their evenings on a computer that sold in modest numbers in 1984. If a tool lets one of those people do the work of ten, that is not a loss for the hobby. That is the hobby surviving.
Two things, and they matter more than who or what typed the code. First, it has to be right. An AI will confidently tell you that the Einstein's disc directory is laid out one way when it is laid out another, because it read it in a book that was wrong. So nothing goes into our reference or our course unless it has been proven on the machine, and the experiments that prove it are written up where anyone can check them. Second, it has to be said. If AI helped, the page says so. We would rather you knew than wondered.
If you have an Einstein project you gave up on because it was too big for one person, it might not be too big any more. A port that stalled, a disc format you never cracked, a pile of documentation nobody has the hours to transcribe. Try it with help. Verify what comes out. Tell us how it went. We will put it on the site with your name on it, and we will say how it was made, because that is nothing to be ashamed of.
The machine is forty two years old. It needs every friend it can get, and it is not in a position to be fussy about how they work.
