kingstnap

Some of these Games are definitely in the RL environments.

Cloning existing games can be done with little to no work.

Building something new is challenging, with or without AI its orders of magnitude slower than copying.

Building something new is also risky. Copying something with proven vitality is less so.

Given this almost tautologically you should expect way more clones than novel things.

Eye watering prices as we expected.

The 512 GB ultra seems too expensive. You can buy multiples of the cheaper one. Like 4 of the 128 GB max, or 2 of the 256 GB ultras for less. (And end up with more total bandwidth and storage and compute).

Unless there is some sort of restriction on number per customer.

Argon 4 is really good in my opinion, especially in chat. Usage is limited, but one big difference is that it seems to better understand what is and isn't worth mentioning.

It has better tuning towards not adding unhelpful caveats and irrelevant details. When I told it I was unhappy with some slop being dumped into some repo markdown files from Astra and Sol it actually went ahead and tastefully cleaned them up in line with how I expected them to be. Reductions are normally so bad from agents.

One big thing is that it really feels like talking to someone smarter than me. Other agents just seem extremely knowledgeable but Argon felt smart.

It depends on your background. If you don't have a computer science background the statement needs layers and layers of buildup that can't happen in a single hacknews comment. But the outside implication is still accessible.

The outside implication for what is going on is that OpenAI has very clearly demonstrated insane capability in coming up with clever algorithms. You cannot deny some real level of super human like abstract reasoning here.

K-sat is the hard theoretical computer science problem. Nothing else is close in terms of how much it has been studied. There are journals and conferences for SAT.

It is really hard to come up with faster ways to solve it. So much so that people were convinced that you really can't go faster than exponential time. OpenAI has just shown a breakthrough construction better then decades of research by elite specialists was able to come up with.

I also haven't read ths paper but coding agents absolutely are a part of modern SAT solver coding challenges.

Probably the batteries included Standard Library.

Cross platform interpreter.

NumPy and SciPy really set the stage.

It's not a trivial feature. The Class A surface on an iPad is the screen.

If you plan to look at moving things, like scrolling, which is a central part of using the device, 120fps makes the scrolling less blurry and therefore more pleasant to look at.

This is like asking why people have a hardon for cars with nicer materials for seats and interiors if a normal cheap plastic interior could pass all the same crash tests.

Training is compute bound.

If its memory bandwidth bound you need to raise your batch sizes to leverage more parallelism.

> The coincidence is that they took the correct path out of “every possible one"

My point was explicitly that this didn't happen. I said that they took every possible path and published the one that worked. It's exactly throw shit at the wall and see what sticks.

> exactly after one of their users took it.

All parties agree that what happened is that someone had told someone working at OpenAI that Anthropic employees are working on Navier stokes. This is where the timing comes from.

---

What you are implying is this: OpenAI didn't do the throw shit at the wall approach as outlined in their release blog they actually secretely monitor everyones chat logs and found the idea in Tristan's Chat history and stole and published it.

Or some variant of this. You are only implying what you think is the truth, please state it outright. Don't beat around the bush.

It's not a coincidence... they literally just span up agents on every possible variant of the problem.

There are 10k agents and Tristan's approach wasn't even groundbreaking novelty, it was literally based on existing published literature from Diego Córdoba and Luis Martínez-Zoroa that the agents have seen.

It's not just probable that agents would persue his particular niche of trying to construct a blowup. It's guaranteed if you just consider the basic setup.

You can still gamble.

When it comes to hard drugs I can understand why there are reasonable arguments for legalization of personal use.

Idk how many reasonable arguments extend the overton window to thinking that meth should be advertised seemingly all over the place. Because gambling ads are actually everywhere.

It has little to do with gullibility and more to do with the way they do things.

In principle AI Agents are superhuman at codeforces but tons of these vibe coded apps have ridiculously stupid performance problems.

The problem is a utilization of expertise thing, it can tell you with incredible detail about pros and cons and caveats to a problem if you ask about it. Truly superhumanly intelligent.

Calibration when it comes to actual decision making takes an incredible reduction in capability after that. Maybe it has something to do with the sycophancy or something but its like someone with exactly zero opinions (and therefore, taste).

My experience is that you can brainstorm with it on "Do A or B?" and it will produce a coinflip answer A, then if you continue and ask about a specific concern "what about this aspect of B?", it immediately 180s and flips back to B being way better than A.

Acting on this expertise when decision making points arise naturally for itself is where everything really nosedives. It's just gonna do whatever.

So when you are doing this "financial advice please" questioning, it is entirely hitting these weaknesses squarely in the face.