How is this related to Toyota? Toyota the car manufacturer?

Yes. Had to look it up, but apparently it was developed by TCNA (Toyota Connected North America) which does car software and such.

source code not available?

Whoa, I think GPT-5.3-Codex was a disappointment, but GLM-5 is definitely the future!

I find 5.3 very impressive TBH. Bigger jump than Opus 4.6.

But this here is excellent value, if they offer it as part of their subscription coding plan. Paying by token could really add up. I did about 20 minutes of work and it cost me $1.50USD, and it's more expensive than Kimi 2.5.

Still 1/10th the cost of Opus 4.5 or Opus 4.6 when paying by the token.

> I think GPT-5.3-Codex was a disappointment

Care to elaborate more?

I tested this pretty extensively and it has a common failure mode that prevents me from using: extracting footnotes and similar from the full text of academic works. For some reason, many of these models are trained in a way that results in these being excluded, despite these document sections often containing import details and context. Both versions of DeepseekOCR have the same problem. Of the others I’ve tested, dot-ocr in layout mode works best (but is slow) and then datalab’s chandra model (which is larger and has bad license constraints).

I have been looking for an OCR model that can accurately handle footnotes. It’s essential for processing legal texts in particular, which often have footnotes that break across pages. Sadly I’ve yet to encounter a good solution.

I found Mathpix to be quite good with this type of documents, including footnotes but to be fair my documents did not have that many. It’s also proprietary.

This reads like proper science fiction tech!

Not surprised, related: https://news.ycombinator.com/item?id=46920315

That's more than related

Hugged to death?

HN hug of death:

https://web.archive.org/web/20260207195051/https://theloop.e...

I read the article but can't make any conclusion from it. Someone can explain in simpler terms?

There have never been enough tenors anywhere, as the article acknowledges.

It's not clear what the evidence is that the problem is getting worse though? Or why it would be?

Declining popularity among men

Can't wait for Bloomberg's Odd Lots podcast have a supply chain episode on this topic. :)

Those unblurring methods look "amazing" like that but they are just very fragile, add even a modicum of noise to the blurred image and the deblurring will almost certainly completely fail, this is well-known in signal-processing

Not necessarily.

If, however, one just blindly uses the (generalized)inverse of the point-spread function, then you are absolutely correct for the common point-spread functions that we encounter in practice (usually very poorly conditioned).

One way to deal with this is to cut off those frequencies where the signal to noise in that frequency bin is poor. This however requires some knowledge about the spectrum of the noise and signal. Weiner filter uses that knowledge to work out an optimal filter.

https://en.wikipedia.org/wiki/Wiener_deconvolution

If one doesn't know about the statistics of the noise, not about the point-spread function, then it gets harder and you are in the territory of blind deconvolution.

So just a word of warning, if you a relying only on sprinkling a little noise in blurred images to save yourself, you are on very, very dangerous ground.

What I find fascinating about blur is how computational photography has completely changed the game. Smartphone cameras now capture multiple exposures and computationally combine them, essentially solving the deblurring problem before it even happens. The irony is that we now have to add blur back artificially for portrait mode bokeh, which means we went from fighting blur to synthesizing it as a feature.

I would love to see any journal showing how profitable an open source company vs closed source one (as a software house). imo terrible business idea?

Startups fail because of a lack of adoption far more often than by any other reason, including competitive and monetisation factors.

If your developer company gets popular you’ll be rich enough anyway. You might need to choose between screwing over your VCs by not monetising or screwing over your customers by messing around with licences.

But yourself as a founder will likely be okay as long as the tool is popular.

...I sense an animated svg of a pelican playing simcity benchmark is brewing somewhere

Funny you say that! When the two new models were released Friday I spun up mayors for each. (But didn’t do the prompting in the most scientific way.)

Mayor Compounded Wonder - Claude Opus 4.6

https://hallucinatingsplines.com/mayors/compounded-wonder-2c...

Mayor Bronze Offramp - OpenAI Codex 3.6

https://hallucinatingsplines.com/mayors/bronze-offramp-09941...

TL;DR: Opus won.

Have also thought about using openrouter and getting one mayor per model running the same prompt through all of them to create potentially the world's dumbest LLM benchmark.

> LLMs are awful at the spatial stuff,

Which LLMs are you specifically referring to?

Are any of them trained with Micropolis data?

Launching the invasion of Canada and Greenland perhaps..

Invasion of Mexico is also possible ...

>President Donald Trump said US forces will "start now hitting land" in Mexico targeting drug cartels

https://www.euronews.com/2026/01/09/trump-says-us-to-start-n...

(Jan 2026)

Aren't those countries thousands of miles away from El Paso, Texas?

Now this is the kind of content I come to Hacker News for.

exactly

> Appears to not age but also to never have been young

/me snorts

We rendered the one million part ABC dataset from Deep Geometry, and open-sourced the data. We also built a fun demo with the following pipeline: CAD > render > caption > embed.

Open-sourced dataset: https://huggingface.co/datasets/daveferbear/3d-model-images-...

Blog writeup: https://www.finalrev.com/blog/embedding-one-million-3d-model...

Interesting.

My go-to for CAD files is usually https://grabcad.com/library

I searched this for "WAGO" and "XT90", so I guess not the same use case. Some hits for "Raspberry Pi", though.

This isn't meant to be a commercially useful search engine- just a demonstration. You'll only be able to search for terms that the VLM could directly discern.

From the blog post: Our search demo proves that it works quite well. As anticipated, text search works well, returning sensible results for even irregular or poorly formed queries. It’s worth mentioning that this is very different from 3D part libraries like Thingiverse or GrabCAD. Search in those repositories requires users to tag or annotate parts with a description, the text of which is used in search. Our system takes only an unnamed part as input, requiring no additional labelling.

Or Rome is the set of a real-life Truman Show.

Fort Pulaski in Savannah, Georgia also has cannonballs embedded in the brick walls: https://en.wikipedia.org/wiki/Fort_Pulaski_National_Monument

Yup; it's a neat daytrip if you're in the area. But then Capt. Gillmore showed up with rifled cannons and showed why we don't use cannonballs any more. :-)

Changing the ISO appears to scale the noise differently from the rest of the image.

Nice, but I'm going to need some ND filters :)

This is missing a setting for the kind of light falling on the subject. Is it full open sunlight? Open shade? Overcast? Sidelight? Backlight?

It all matters.

[flagged]

The software defined storage company croit.io announced it in their Workation in May 2023. AI is just another tool and people have to understand that it's not going away. As a company, you still need people to make use of this tool.

"Please don't fulminate."

https://news.ycombinator.com/newsguidelines.html

So that's an in-cluster supply chain attack enabler? :)

I think it would be better suited to use the terms we use for natural languages. A natural language is dead when the last person who learned it as first language dies and are extinct when there is noone that would speak it at all.

In these terms, telnet has been dead for a long while, but it's extinct now.

The most interesting thing here isn't the CVE - it's the invisible coordination. A backbone provider acted on advance knowledge of a critical flaw, implemented filtering at scale, and the rest of us didn't notice until GreyNoise's data showed the drop. The vulnerability got patched at the network layer before it ever reached the application layer. This is what mature security ecosystems look like - the boring, quiet fixes that happen before the press release.

Stop spamming AI slop

I grew up in a snowy place and I still live in one. I tell myself every year that this negative experience “builds character”, that being stuck inside forces one to be more intellectual, read more, etc.

I kind of still believe that story, but as I get older it starts to feel like cope, and the sunny shores of Miami / Spain / Warm Place seem more full of life.

> starts to feel like cope

In a way you’re right. All the effort to reduce energy usage, go green, the savings are all negated with energy spent on generating heat and emissions.

Also, sunlight improves your health: https://www.youtube.com/watch?v=sYnUPxG7ONk

Is anyone expected to know what railway.com is?

What is this 'railway'?

I am assuming that a domain like railway.com should be about trains.

Why does every tech company have to name themselves as a one word .com website and what they do is unrelated and vague to their own name?

Does every tech company think they are Apple and have to register every word in the dictionary and redefine it as a technology company?

Really bad name for a company.

could be worse

could be called "entire" (https://entire.io/)

Sure all the people who somehow find themselves unable to find community, are neurotic as fuck, and who are lonely have some sort of theory for how community is formed. This is definitely a case of "those who can, do; those who can't, teach". This entire field is full of immeasurable guru-bullshit without anything of any value in it. It's just pseudo-science dressed up in the language of science with some pithy lines of how "there's more to it than numbers" and garbage like that. It's just made up bullshit from people who really shouldn't have received a college degree.

Out with this garbage. Defund the bullies.

These are the exact same arguments people make against immigration and diversity. I do not want this far-right drivel on HN, flagged.

You might want to look at the HN commenting guidelines. Namely you should avoid such uncharitable interpretations and show a bit more curiosity.

Amazing!

There is a whole genre of youtube videos with AI generated Feynmann voice based on those lectures. They are of surprisingly high quality.

Do you have a link to a good one?

I presume the original videos of Feynman are lost, or never existed?

Yeah, clicking unverified links in a markdown document to launch an executable....

Clicking unknown links is always a bad idea, but a CVE for that? I dunno....

Notepad was the epitome of a single, well functioning app in Windows for the last eternity of two.

Rewriting it to integrate AI and some bells and whistles recklessly and having a CVE is tragicomic if you ask me.

What AI great job!

I don't really understand the complaint here. It seems for most of those extensions have it in their literal purpose to send the active URL and get additional information back, for doing something locally with it.

And why does this site has no scrollbar?? WTF, is Webdsign finally that broken?

> And why does this site has no scrollbar

Seems someone decided it was a good idea to make the scrollbar tiny and basically the same colour as the background:

    scrollbar-width: thin;
    scrollbar-color: rgb(219,219,219) rgb(255,255,255);

Just create an AI service and users will voluntarily send you all their data.

No need for such complicated attacks /s

bluntly: a lack of regulation mandating that consumer goods manufacturing responsibilities cover the lifecycle of the goods (including end-of-life).

yes i'm fully aware that recycling components is difficult and costly; if you truly believe in the market as an innovating force, you could stand to be a little more optimistic that we could make this a reality :)

You obviously are not in manufacturing.

USB-C charger reuse is now common (Apple chargers still gets the UK/EU law exemption)

RoHS prevents Pb content in recycled parts (less toxic waste)

Lithium battery recycling drop bins are next to the store entrance (financial incentives)

ATX12V/EPS12V power supply in your PC is a standard component between motherboards

Aluminum and steel instead of plastics is common (consumers like the aesthetics too)

Under the guise of recycling, problems arise when third-world people use vats of acid to strip trace gold/platinum from electronics. Others strip, relabel (laser marking), and resell aged chips as new stock... this can cause safety/reliability problems.

Some firms now use solder centrifuges to extract RoHS solder off parts, and resell the tin bar-solder back to manufacturers.

e-Waste can be a desirable resource, but few people want old Lead contaminated CRT or mixed plastic filled with inserts etc.

Companies like AMD with AM5 compatibility across chip generations should get an award for their great work reducing waste. Linux <6.0.8 kept a lot of laptops out of the landfills too, but now kernel >6.0.15 will no longer support old GPU/Laptops as NVIDIA ends legacy driver support. =3

recyclable and reusable aren’t profitable for companies. They want you hooked on buying the latest incremental/minute change.

If companies like Apple cared truly cared about the environment. We would have phones, laptops with easily repairable and upgradeable hardware.

Framework is the closest we have come to having a thin profile laptop and easily repairable and upgradeable hardware.

Why did we make just an infrared telescope then? Why don't go into even lower frequencies, surely we would detect something too if we just look?

Lower frequencies are microwaves and radio waves. We already have the square kilometer array.

"just an infrared telescope"

how about you go make yourself conversant with "just" the technical requirements of the main cryogenic pump onboard, leaving out the rest of the thermal management systems for whatever remains of your life, which will have to be long in order to fail honorably.

The author of this post could solve their problem with Cloudflare or any of its numerous competitors.

Cloudflare will even do it for free.

Does this author have a big pre-established audience or something? Struggling to understand why this is front-page worthy.

I think the point of the post was how something useless (AI) and its poorly implemented scrapers is wrecking havoc in a way that’s turning the internet into a digital desert.

That Cloudflare is trying to monetise “protection from AI” is just another grift in the sense that they can’t help themselves as a corp.

Is 90% of his brain actually missing or is the volume reduced by 90%? I.E. are the mass and connections still mostly there but just squished by extra fluid?

From what I can tell googling about this, it seems it is mostly just squished, so volume is down 90% but mass or neuron count is not missing 90%

But how many wrinkles?

Thanks for posting!

Do you have slides from your FOSDEM presentation? The video has this funny angle :(.

Thanks for sharing!