Someone on GitHub has shared a project in which they’ve unveiled new, enhanced firmware for the PEN-F. A comment immediately appeared on the rumour site claiming it was all just AI nonsense. So that piqued my interest on several levels. What can AI do? And what else can you cram into the PEN-F? But before I’d had time to delve into it, I had a phone call yesterday. During the call, someone casually analysed the firmware of the OM-1 and OM-1II and broke it down right down to the assembler code. A BIN file like that is over 200 megabytes. Every few minutes I was presented with a new analysis, and by the time I’d even read through and understood one, the next one was already there.
There was a local AI at work. Not a generative one, but an analytical one. It could analyse the stuff faster than I could even process the results. And yes, of course there are loads of lines of text in there that clearly sound like AI. But that doesn’t matter at all, because this is solely about analysing source code that you couldn’t analyse by hand in weeks. Just comparing the two files and finding the differences. Identifying jump addresses, memory accesses.
If you’ve got the documentation to hand, and know where to look, and where the calls for the ARM processor are, then none of that’s a problem. But if you have to figure that out first, it certainly is.
Oh, and just in case anyone asks: the OM-1 runs on Linux too.
And parts of the binaries are encrypted.
So: having seen first-hand what a local AI is capable of during analysis, I reckon it’s pretty likely that within a few months we’ll be getting DIY firmware updates for pretty much all cameras. The Reddit user mentions that he wants to develop a test system to test the firmware outside the camera – and, based on what I’ve witnessed, that too isn’t an insurmountable problem with the right setup.
In fact, camera manufacturers will increasingly find themselves having to explain themselves when hobbyists equip old cameras with features that the community has been demanding for years, but which were always claimed to be impossible. Perhaps it would make sense for the manufacturers to jump on this bandwagon whilst they still can.
The drone community has already demonstrated what’s possible. The range of hacks available for drones is simply incredible. If these kinds of hacks start to take off for cameras as well, it’ll be difficult for manufacturers to charge people a fair bit of money for just a bit more software.
When will we see the first E-M1II with LiveND?
Oh, and for anyone who wants to get stuck in themselves: you can download all the Olympus firmware here.
Cover image: This isn’t AI. It’s the real thing.
KI-Slop, auch als Slop-Inhalt oder einfach Slop bekannt, sind digitale Inhalte, die mit generativer künstlicher Intelligenz erstellt werden, denen es an Aufwand, Qualität oder Bedeutung mangelt und die normalerweise in großen Mengen produziert werden, um sich einen Vorteil zu verschaffen, Geld zu verdienen oder Menschen zu täuschen.
Es handelt sich um eine Form synthetischer Medien, die normalerweise mit der Monetarisierung sozialer Medien und Online-Werbung in der Creator Economy verbunden ist. Der in den 2020er Jahren geprägte Begriff hat eine abwertende Konnotation, ähnlich wie Spam. „Slop“ wurde sowohl von Merriam-Webster als auch von der American Dialect Society zum Wort des Jahres 2025 gewählt.
(Quelle: https://en.wikipedia.org/wiki/AI_slop)
Ich hab schon in den 90ern mit Disassemblern gearbeitet und die waren damals schon nicht schlecht. Da geht heute natürlich noch viel mehr. Software ist ja gut strukturiert und formal, das lässt sich sauber zerlegen.
Was jetzt noch dazu kommt, sind funktionale Beschreibungen, die man sich früher selbst zusammen reimen musste.
Spannend wird es, wenn sich Hersteller weigern was zu reparieren, das durch überanspruchen der Hardware durch modifizierte Firmware beschädigt wurde.