Zeruxe

I think that tool is actually great, and nice that someone deployed it again.

But it is too much if that makes sense, it's hard to fit in so much stuff in one little web app, so it ends up just being a cool visualizer but not so good educating tool.

The intended flow is actually to select a tooth first and then inspect it at a deeper level, including its internal tissues. It sounds like that interaction isn’t clear enough from the UI though, so I’ll look into making that maybe more straight forward, thank you for the feedback.

Thank you, im glad that you found it useful. Especially given your experience around the industry.

The base anatomy comes from BodyParts3D, an open anatomical dataset from DBCLS which is openly licensed under CC BY-SA 2.1 Japan.

The tooth internals were modeled to fit the actual teeth so it’s a mix of open anatomical data, processing, and additional modeling.

The 3D models you see are rendered in a javascript framework called three.js, an awesome framework if I may say.

Thank you again for the kind words!

Thank you for the kind words, im glad that you like it!

Hi, I have filled in the form and would love to have a chat with you, email starts with 'y' if someone tried to impersonate me. Otherwise just reach out to me via github discussions on the repo and you can have my email.

Yeah you are totally right and I appreciate the words. I sometimes tend to think I want to rather try to guide them in my POV, but some people refuse to give in.

You mean MRI scans? I could technically try to do something similar, would be more on the machine learning part, don't know if a web app could manage that. But definitely something I would look into if there is interest.

Thank you, you too!

I really appreciate you coming back, taking another look, and saying this. That is very mature of you.

The FDI numbering system can be confusing at first and I see how you could have mistaken it.

More than anything though, I do appreciate the change of tone, and I hope all these lurkers hating on AI were like you.

For what it's worth, I probably could have been less aggressive in my response too, the coffee explanation was funny.

Thank you for actually taking the time to look through the project properly.

No and that is a bug that you just pointed out which I will take care of, thank you!

Thank you for pointing it out!

Okay I will try to, thank you and please do the same if you don't mind.

Here is the product hunt for those of you who would like to support - https://www.producthunt.com/products/dental-scope?utm_source...

thank you for the kind words, I barely know how to use this as im very new to hackernews, even though it seems pretty straightforward haha

Thank you for the kind words, Im happy you enjoyed it!

Just a clarification for the people hating on clankers, I want to know what you expect from todays devs? do you genuinely think that people should manually handcode everything for it to hold any significant value?

I just don't understand the point of view where - "a clanker made it therefore it's trash", it's always the same people they lurk on reddit aswell.

Even in the case of putting something together manually, you are competing against people using AI and you will never match their production level or speed.

I mean the people who hold the highest priority say in this are the dentists/dental students, and so far all the feedback from those have been nothing but positive?

yeah I just noticed that, I have no idea I guess ill just write them again, ive seen this before

Thank you even though it definitely took more than 3 days haha

I have no idea what that means

haha thank you! Appreciate the kind words!

I actually have been in contact with Ashe just appreciating the idea of her work, and that was definitely a part of what inspired me, I just wanted to make extensive work that focused rather on a specific part of the body. Thank you for pointing it out.

Humans have 32 permanent teeth, and that is exactly what Dental Scope models. It is even pointed at the first screen.

If you found somewhere in the UI that says 38, point it out and I’ll fix it. That is useful feedback.

The rest is just a long rant based on the assumption that using AI means I neither understand nor review what I build. You know nothing about my education, my engineering background, or how this project was developed.

Sounds like an old grump that is falling behind to AI.

That is a pretty confident description of a process you were not involved in.

AI helped with implementation. It did not do the research, architecture, product decisions, debugging, validation or review for me.

You can dislike AI assisted development. Saying I "built nothing" is just lazy.

Thank you, I appreciate that.

Just to clarify the 3 days part though, that is the implementation period currently visible in the GitHub history. The research, planning and architecture work started roughly 2 weeks before that.

But I agree with the broader point. The useful part is whether the tool actually helps someone understand the anatomy better, and that is what I care most about improving now.

huh?

I actually changed the background myself as the initial recommendation from AI was a flashbang.

Thanks for the feedback about the bug I will look into that.

Also thanks for the suggestion and the compliment :)

The implementation that is currently visible in the GitHub commit history was done over about 3 days, yes. But that does not include the research, planning and architecture work that happened before I started committing the implementation.

I spent roughly 2 weeks researching the domain, figuring out the structure of the tool, planning the architecture and deciding how the different anatomical layers and interactions should work before the repo history you are looking at began.

That is very close to what I would like this project to become over time. There are some excellent anatomy tools out there, but many are behind subscriptions or institutional licenses.

My goal with Dental Scope is to see how far an open source version can go, especially if dentists, students, developers and 3D artists start contributing to it. Full body anatomy would obviously be a much larger undertaking, but I would love for the project to gradually expand in that direction.

The Wikipedia comparison is probably much more ambitious than where the project is today, but the idea of a freely accessible educational resource is what makes me do this.

Thank you!

Jev and Laya answer typed questions about text in one forward pass.

Their confidence scores aren't error rates: a model can say "critical, 78%" and be wrong, and nothing says which answers to trust.

Zet sits on top and marks each answer sure or unsure. Sure answers are meant to stay within an error budget you choose (5% here), set on labeled examples and tested on held-back ones.

Benchmark: MASSIVE 1.1 (human-labeled), six-topic task, 350 held-back requests. Both rows use the same Laya predictions.

- Laya alone: 350 used automatically, 45 wrong. - Laya with Zet: 227 used automatically, 4 wrong (1.8%), 123 sent to review.

Zet flagged 41 of Laya's 45 mistakes (91%) while leaving about two thirds of requests automated. The catch: 82 of the 123 reviews were for answers that were already right.

What it doesn't show: - The six topics were chosen as an easier task. With 18 topics, Laya made 228 mistakes on 602 requests and Zet sent all 602 to review: no automation, but no automatic mistakes. - The cautious 95% upper bound on the automatic error rate was slightly above 5% (5.2% and 5.7% per language), even though the observed rate was lower.

This project is under development, I just would like to share it for those of you who want to test it or help me contribute, there is still a lot to be done!

Site: https://yoosseph.github.io/Zet/

Feedback welcome.