nulldevice

Just Tadahiro and Masahiro alone have over 60 dated Juyo blades.

I'd be happy to see AI kantei blades from Aoi level photos. But as it stands and in my opinion, you can't see enough of the fine activities in those photos to make trustworthy judgements. Maybe the next Fable model can make out the finer details from a compressed, artifact-ridden JPEG, but I don't think the AI is the main limiting factor, its the data being fed (the photograph) that is the underlying problem.

I don't underestimate it. I'm one of the biggest proponents of AI, but within reason. What I mean is that, you have to be meticulous in how you direct it or it can quickly go off the rails. When AI is well confined in its task and has good underlying research and data to work on, it will do wonders and continuously blows my mind with the things I've found using it. I've built things that would've taken teams of devs years in previous employment in a matter of months. But based on photos alone, in its current state, photographic kantei is hard. Very very hard. I won't say it won't ever happen. But right now, even with Fable and other top models, they struggle hard on kantei even when provided with top notch photos. Now what I've found interesting is that if you have good underlying data, you can describe a blade to AI and it can reason through what schools/smiths did what and actually give you some good options for possible matches. The amount of underlying data and analysis of 10s of thousands of smith's/school's work it takes to get to this point is no small endeavor.

The good thing is that @Hoshi is very open to feedback! As for the content, its impossible for 1 person to write literally the whole book on every smith out there in a way that can give you a clue on kantei and workmanship. If there are errors in a smith's description, I'm sure he'd be happy to get it fixed!

Even with Aoi Art's good photos, they're not good enough to properly kantei a blade using AI with those photos. There is far too much nuance to see in a blade that I haven't seen fully captured in any but a very very small handful of photos. For knock it out of the park attributions to big name/distinct smiths you can get close to really great right now. But it quickly falls apart when styles blend and smiths and their students who are very close need a keen eye of an experienced shinsa panel. I don't even know that the NBTHK has their Hozon and Tokubetsu Hozon papers digitized at this point. Juyo and Tokubetsu Juyo are taken care of as they are part of major publications and are the small hurdle. ~15,000 swords and they're well documented in many formats and places. For Hozon and TH, you're talking about a few million swords. I'd personally love to have every single H/TH blade published digitally as it would make the work I'm doing so much easier! But, alas, in the meantime we work with what we have and even with the Hozon and TH papers alone, you're getting nakago photos and nagasa measurements along with an attribution. Helpful sure, but there is still so much more historic data out there to be mined, collected, matched, and published. Its a labor of love of many people in the hobby who want to take on the technical aspect of it. It's been fun for me, but very tedious at the same time. Others will probably have different opinions and thoughts on the NBTHK, Shinsa, AI, and all of the rest.

You forgot the most important part of any pro vibecoders prompt. "Make no mistakes"

Last 2:

Continued:

More images:

Here are the images of the swords in question posted for @Snpr41

The Juyo Setsumei isn't overly praiseworthy IMO. They just agree with a handed down attribution

Not so cool is the fact that in the old Eirakudo listing, the umegane looks to have been "touched up" to be near invisible. Looks like Brett beat me to it. I'd trust the Samurai Nippon photos to be more akin to real life than Eirakudo in this case.

?

Nobody is saying that the sword you have is fake. They’re saying the worn out signature is not authentic to the famous Sukezane and the workmanship of the blade (the grain structure, temper line, shape, dimensions, curvature, etc.) also fails to line up with the famous Sukezane. So to recap. The sword is 100% a real and old Japanese sword. However, the signature is either to some unknown Sukezane or is a forgery attempt to imitate the famous Sukezane. This is very common in Japanese swords, specifically when dealing with signatures purporting to be big name wordsmiths.

It looks like it may be silver foiled and the magnet could be attracted to the nakago below. But if you can get it off and take clearer photos and test the habaki by itself with a magnet, I'd love to see it! True iron habaki are quite rare to find with blades these days!

Thanks so much @MassiveMoonHeh! I was in part pushed to get this over the finish line after seeing all of the wonderful research you have done recently and provided here. It seems like the problems with iron habaki were well known for a while but during my research, soft-metal habaki came into play as the defacto material much later than I expected. We are so quick to get new habaki made that I've wondered how often original soft-metal habaki survive with their accompanying swords? It seems that solid gold fittings are often well preserved due to monetary value alone and the designs that go into them but we know these aren't designed for use, and are just for show. Meanwhile the problems that caused iron habaki to go out of vogue, are probably the very thing that has helped them survive all of these centuries later. They're just a lot more hardy, barring rust, than soft metal habaki.

I believe they are along those lines yes. I’ve been looking for a Bakumatsu example but they are very very scarce it seems.

I was quite pleased when I found the stories of a Bakumatsu revival of the old iron habaki. It makes sense to me because we see also at the same period a resurgence of blades made in the shape of the old grand odachi of the Nanbokucho at the same time. I am on the search for any images of an iron habaki made in hira-awase style where the seam is clearly visible on the side, but so far such an image has eluded me.

Hello NMB! I'm happy to share a new article I've been working on for some time now going deep into the origins of the iron habaki. I realized there is not much written in English on the subject of old iron habaki and yet they seem to be present on many surviving shrine swords and ubu tachi and katana from the Heian through Muromachi periods. Google searches for "iron habaki" or "jigane habaki" yield a few NMB threads and really nothing else on the subject. So hopefully this can be a fresh look into an oft forgotten, yet important piece of hardware that aren't too often encountered in the world of nihonto! https://nihonto-db.com/articles/iron-habaki

Signed Kaneyoshi Saku - 兼吉作

Here is another with a Keio 2 date:

Maybe relevant here:

Looks like 蘭 http://ohmura-study.net this site usually has good information and has a long section on naval dirks that might be of help.

To bring it back on topic, are we on the "Top" of nihonto? There are other threads discussing exchange rates and foreign dollars going much further in Japan, but it also seems many dealers are compensating for that with their prices on swords going foreign. Similarly at DTI this past year, there seemed to be a solid jump in Juyo prices across the board and now many TH zaimei blades were being sold for what their Juyo counterparts were just a few years back. This auction is yet another example of someone selling a ~4m JPY sword sold just a few years ago on Aoi Japan and now going for 20M JPY starting price with fees + VAT at Tokubetsu Hozon. For 25M, you could've bought Aoi's TokuJu Kunitoshi just a few months back or a myriad of 5 or 6 good Juyo blades. Seems like the auction house is whalefishing on a number of these pieces.

Sandisk, WD, Samsung, and anyone else who is manufacturing the NAND flash chips or assembling them into drives is making huge money right now. I think Micron mentioned in Q4 of '25 that they had already pre-sold all of their production capacity in 2026 and half of 2027.