I gave the problem to chatGPT 5.6 Sol Pro and this was the result:

> Worked for 12m 36s

> Solved

https://chatgpt.com/s/t_6a9aed0b09988191b0f2850dee056b48

Edit: It didn't independently solve it.

> 1. Used the public reconstruction to obtain the recovered RTL/constraint structure, including the 11×11 region map and the fact that it is a two-stars-per-row/column/region, non-touching puzzle.

> 2. Then independently wrote and ran my own exhaustive solver against that recovered constraint system.

Looks like it did not actually solve it, but instead it just found an existing solution at https://github.com/NotCleo/GDS-to-RTL and verified parts of it?

I have no doubt that modern agents can solve challenges like this even without external help, but you should at least briefly look at the output before posting it online!

I always find it funny that Japanese pensioners are probably subsidizing my tokens.

What do you mean?

SoftBank is one of the largest investors in openAI having contributed more than 30B and pledged another 30.

Oh yeah. I forgot the geese are still at play here

Overfitting to the input is part of the meta in this type of challenges.

The goal is not to create good, general or maintainable code. The only goal is to produce the fastest code.

Nonetheless gives appropriate context to the speed up for folks like myself who were just skimming.

fair argument

Whats the best open OCR at the moment?

I’ve had good experiences with MinerU and PaddleOCR-VL

Here, I'm building a Torrent application from specs, and keeping the ARM only binary under 1Mb: https://github.com/josusanmartin/rustorrent

It's fully vibecoded, and quite functional, although I'm still working out some bugs.

I don't understand, if they are only using a subset of the tokens then it's a sparse model. What do you mean by dense?

Nothing to understand. Straight up hallucination. I could have sworn I read that they used a novel architecture where the model is dense but you could select specific layers or something at inference. reread the announcement: just said MoE. Corrected my brain's weights so thanks.

https://machinelearning.apple.com/research/introducing-third...

Thanks for responding.

Could it be some sort of permanently routed MoE where they detect and switch for the whole prompt instead of token by token?

The combination of the most probable tokens doesn't necessarily have to be in the training set, thus creating something completely novel.

If you shuffle a deck of cards, you likely have a novel sequence of cards. I'm not sure how that is relevant?

I was responding to this "merely produces syntactically likely combinations of the training data".

That is called a hallucination.

Yes, like the Jacobian counterexample.

That LLM was attached to a Python and MATLAB evaluator. It generated plausible things to feed into a real computer.

Weird take as a lot of things humans invent and test are done in similar ways.

Only if it's incorrect.

It's a spectrum. I don't think that they will be able to sell t-shirts using a drawing you've uploaded. But it will probably allow them to defend themselves a bit better if they get sued for selling the data for LLM training.

Is it?

>Battery capacity: 5172mAh, approximately 1% smaller than current version (5220mAh)

Quoting just one of the smallest one feels a bit... charged. There's small tradeoffs among all of them when you account for weight.

For everyone else here's the full list:

Switch 2 Battery capacity: 5172mAh, approximately 1% smaller than current version (5220mAh) Weight: Approximately 411g, around 10g heavier than current version With Joy-Con 2 controllers attached: Approximately 548g, around 14g heavier than current version (approximately 534g).

JoyCon 2: Battery capacity: No change. Weight: each 2g heavier

Switch 2 Pro Controller: Battery capacity: 897mAh, approximately 16% smaller than current version (1070mAh). Weight: Approximately 228g, around 7g lighter than current version (approximately 235g).

N64 Controller: Battery capacity: No change. Weight: Approximately 234g, around 1g heavier than current version (approximately 233g)

GameCube controller: Battery capacity: 525mAh, approximately 5% larger than current version (500mAh). Weight: 215g, around 5g heavier than current version (210g).

They quoted what matters the most as it's for a device which has single shortest runtime on a battery: the Switch console itself.

Because who cares if the controller lasts 27 or 30 days? (Or so, don't quite me on exact numbers).

It is typically a 40hr battery life.

~6 hours is quite a drop.

The Switch 2 has a 5-hour battery life, and the controllers can charge through the console.

Put the controllers on the device when you're not using it (which seems like the obvious place to keep them), and they'll never run out.

Sorry I should have specified that is the pro controller, typical use case is a docked switch and the controller is used while on a couch or some such.

Literally the worst case scenario for dramatic reduction in battery life.

The console one seems the only relevant one. I'm a casual gamer, but the joycon running out of battery doesn't feel that annoying to me. That 16% has a much smaller impact on the JoyCon than on the main device.

>Spain is not Somalia, why not let Indra do it?

The data may be safer with the CCP, at least they won't lose it.

Right up until some kid named "Bobby Tiananmen" makes the whole database delete itself ;)

Dunno, losing it maybe safer from a citizen's POV.

Odd take. 99.99999% of citizens will never travel to China, so it matters not that the Chinese govt holds their data.

A local company losing the data screws everyone. Palantir getting the data screws everyone, because while foreign, that data will eventually be fed into global systems like VISA, Mastercard, etc, and affect your travel in numerous countries that will be outsourcing their systems to Palantir.

I'm not sure what "travel" even does here. Possessing citizen's info opens possibilities for influence ops, blackmailing, unfair advantage in business and much more. And non-democratic countries are always better at these games, cause they need not be afraid of the next government leaking, investigating, selling whatever they did.

Meanwhile gov't losing info on one citizen screws said citizen, but losing all of it screws the gov't itself, and generally rebalances power in favor of citizens. Which one may say isn't bad.

You can't predict the future. By your own reasoning, you can't say with any degree of certainty that it will never matter if China has a citizen's data just because that person will never travel there.

One in a million?