=0.0000 t=0.0001 [A] sensor_read ok t=0.0002 [B] sensor_read ok t=0.0005 [A] compute begin t=0.0009 [B] compute begin t=0.0011 [A] compute end t=0.0012 [B] compute end t=0.0013 [A] write buffer[0]=1 t=0.0014 [B] write buffer[0]=2 t=0.0015 [A] read buffer[0] -> 2 t=0.0016 [B] read buffer[0] -> 2 t=0.0017 [A] ERROR checksum mismatch expected 1 got 2

—

what do you think

They are trained on text to respond well to text based questions and do tasks involving modifying text files.

They are not designed for playing games, looking at games, or visual puzzles. Also translating games into text input for the LLM skews the test completely.

Imagine trying to get a human to solve visual puzzle but they can’t look at the puzzle but it has to be explained to them in textual format, we would be terrible at it.

But yet we persist in wasting time on this benchmark. It doesn’t mean anything.

Literally promising frontier-equal results but at 1/3 price, i.e. cheaper than Kimi K3?

Doesn't offer any real benchmarks or explanation of how this magic trick is accomplished. No credible team or notable scientists behind it...

Hey look! I managed to get the AI to fail at some basic thing (after trying 1000s of times), it means AI is failing!!!!!

It's a total red herring and mispresents the current state of the models completely. It's like meeting a child prodigy in maths and then saying he's actually dumb because can't drive a car yet.

The idea that AIs are "bad at math" because if you send it 100s of basic arithmetic questions, it can get one wrong, is frankly laughable.

It's like me saying Terence Tao is bad at math because I sent him 100 long division problems and he made a mistake in one. Yes, we know models think in tokens, and if you give it a bunch of math problems (and its not using tools like Python to deterministically work on the problems) then of course you cannot guarantee accuracy.

But no one thought tool-less LLM calls were a solution for arithmetic in the first place. So the author is constructing a great big straw-man and attacking it vigorously.

The reality is this, the frontier models are as good as (OR BETTER THAN) the leading mathematicians in the world right now. Leading mathemeticians (like Terence Tao) are using Fable and GPT5.6 as partners in doing research.

As for being stuck in past? Again, weird Anti AI/AI Doomerism because models have fixed weights. So what? They can use tools (and do so very well) if they need up to date data and information.

Again it's like the author wants to paint a picture, based on their own biased beliefs and chooses to represent the current state of AI in an entirely inaccurate fashion.

The models are beyond expert level in many areas at this point.

Do you really believe that adding extra junk to your prompt is going to make the model write code better than it does already?

Again, imagine going to Terrence Tao and "prompting" him to get better at Maths, do you think you can do it? What prompt would you give to him to make him produce better maths. Unless you're already a world-leading Mathematician I think you would find it hard.

It's like imagining you could "prompt" Richard Feynman to be smarter at Physics.

That is, for 99.9% of engineers, if you want the model to do a code review of your project, the best solution is to just ask Fable, "Hey Fable, do a code review of this project." Throwing in extra text like "think like a senior engineer", "ensure you focus on DRY principles, KISS, self documenting code, etc", doesn't make a difference.

These sorts of tricks used to work with dumber models, but now, like I said before, it's like thinking you can prompt Linus Torvalds into writing better C than he already can do.

If there is a huge demand for shipping goods internationally, investing in ships and planes isn't burning money.

There is massive demand for compute in the world right now, Google is investing in that area. That's a good thing.

I see everyone around me doing way more work, of way more depth, than they ever did before using AI models. I see my company and friends of mine all paying large sums of money to Anthropic, Google, OpenAI to use AI models, and do more work than we did before.

So Google is investing in infrastructure which is HIGHLY in demand, there is much more demand than supply, and then they are making money from this infrastructure...

That's a good thing for Google, and as an investor in Google, I am glad they are making these investments.

Yes the OP is completely biased.

Why isn't there an American company doing this? Why isn't an American company making the lasers, mirrors, metrology, vacuum handling tech, etc?

If the machines are so easy to build and integrate, why isn't it happening in America?

I could make this exact same thing over a weekend and post it on Hackernews. But I won't because I would be embarrassed to do so.

The bar for posting something to HN should be high, the bar for wanting people to read your code, your writing, should be putting serious effort and thought into it. Not just vibe coding something up with a vibe coded README and 100% vibe coded code and not even a novel idea or implementation.

That is, creating the scoring system/judge models etc for RL is not easy at all. You can easily create an RL loop which is getting better and improving its scores, but actually the result is totally garbage, because you're measuring the wrong thing.