I posted a link to HN a few days ago that got literally zero upvotes and zero comments (Applying NASA's satellite-imaging trick to ancient museum objects https://news.ycombinator.com/item?id=50023128) and it was hit with insane bot traffic. HN was the only place I had posted it.
Author here! I read about DStretch being used on rock art to discover hidden images in a thread here last month (https://news.ycombinator.com/item?id=49645437) and thought it would be interesting to apply it to museum artifacts. So I did, and it was! I ran it over ~4,600 high-res photographs from the British Museum, the Met, Berlin, and other collections, and then slowly went through them looking for anything out of the ordinary.
You can see my favourite results in the post. The color stretched versions are false color (they probably weren't bright neon) but the hues are likely generally in the right direction. The overwhelming majority of artifacts didn't show anything at all, but the ones that did were wild. It took a lot of time to sort through these, and I can only imagine there are a lot of objects in other museum collections that haven't undergone technical/fluorescent photography where you could find something if you looked.
Last week I used Opus 5.5 to decipher a number of Waffen-SS Truppenschlüssel messages from 1941 that had never been broken. It was a similar approach, I used it to find, transcribe, and process over 10K images from various archives of WW2 material. Within that material was text I could use to help break day ciphers, to use as cribs, and so on. Opus didn't find a new vulnerability in these 85 year old ciphers, but it allowed me to process data at scale in languages I don't speak, and to write a dozen solvers in the time it would have taken me to write one or two by hand. Took what would have taken a research grant, three assistants and a year of time and turned it into a weekend project.
It's incredible when outlined in the context of a known problem domain, and the estimated effort in legacy terms, how far we've come in such a short space of time.
If the introduction HP-65 was the first time an individual was capable of fully replicating the logical and cryptanalytic functionality of the WWII Turing-Welchman Bombe, then the roll-out of Agentic AI could be said to represent the first time an individual was capable of replicating the functionality and deliverables of the WRENs of Bletchley Park. Mindboggling.
Even more impressive is the Astra-6 report recently. Given access to the Crypto Cellar Research archive of unbroken intercepts, the AI wrote its own custom Python and C++ software to simulate an Enigma machine and a mechanical Enigma Bombe.
Opus did this as well, I had it working on the (now-solved) Enigma messages but I killed the run after a couple of days of CPU time. When the Astra-6 report came out I compared the method to what I had been running, and I had the right crib about 4 hours down the road in my queue and it would have been solved then. (Or, and there's a lesson to learn here, if I had made my solver a little more tolerant towards garbled letters, it would have solved it in the first couple of hours).
Very interesting, have you written this up anywhere? Curious if these add up to new historical knowledge or if they are more just conveying information we already knew. At a certain point as we gather more finds like this I suspect it will tip from the former to the latter.
I haven't written it up yet (maybe this weekend) but I did submit my solves to Cryptocellar, and I'm credited with them on the 1941 Messages List. The contents were varied but ranged from things like "700 field howitzer shells received" to "[so-and-so] admitted to hospital at [location]" to artillery targeting information. I don't suspect there's a lot of new historical value in these, but no doubt those more qualified to judge will look at all these new messages soon enough.
I'll put something together this weekend. I had actually started writing it up very early for several days worth of messages, but someone else broke them and submitted them while I was building a little site with a web decoder and writing my results up. 85 years of sitting there unbroken, and then Opus 5.5 comes out and multiple people are solving it within hours of each other.
In any case I stopped writing it up at that point and tried to focus on the messages. But I have the mostly-done writeup and I'll post it soon.
... Come on now. This is obviously because I quoted the LLM written parts from the article. Remove the quoted parts and try again. Do people not understand what LLM detectors do?
And, to be fair, short texts are one of the areas where even Pangram struggles, with a longer text it would be able to identify it as a mix of human and AI written text. They clearly disclose that short text detection is less reliable.
The fact that anyone as any faith in an AI language detector at all blows my mind. It seems that every AI detector that gets productized gets as bad as the AI CEOs themselves and starts making claims their products cannot accomplish. They try to sell it as a black and white detection product while covering up the error bars are huge.
There are a number of ways they're able to tell the difference between arrowheads and sharpened stones put on a stick, including high-resolution CT to look at the stress microcracks of the arrowhead. Bow-fired arrows and thrown spearheads have different launch stress profiles as well as impact profiles. There are a lot of overlapping types of analysis that happen to establish what technology did or did not exist in a given area at a given time.
I set up my own Ableton MCP rig last year to do this. It works extremely well. I have had Claude organize, rename, colour, and prepare 100+ track songs for mixing. Unlike the OP I have not used it for any kind of creative use, although I did experiment with giving it access to my vst library and trying commands like "put a vocal compressor on the lead vox, with a high ratio and a fast attack", with absolutely terrible results.
The organizational and assistive ability though was/is outstanding, and using it in my own projects honestly made me sad that we're seeing AI integration everywhere that automates the creative work and leaves you with the tedious work. I don't expect we'll see AI updates to DAWs from vendors anytime soon with features like "speak into your mic and tell your DAW to group all the snares, color code them different shades of red, and rename them", but that's the part of the process I wish we were trying to get AI to tackle.
