lurker137

I wonder if OpenAI will release theirs soon. It seems very important to be the first one to IPO

In their thinking and if you ask them, I've seen Kimi K3 and other models confidently say that they are Claude from Anthropic, and accuse you of attempting to jailbreak them if you say otherwise.

Thanks, those are great points, now I'm definitely excited for the full release to try it out

How does it compare to Google's Genie 3 or other similar models? At first glance it does look much more impressive visually

I'm no economist but here is how I think of it. In an ideal world let's assume every worker is replaced by one AI / robot that is an exact clone of them. Then let's say there is an ideal government that taxes all the income each person's clone makes and hands it directly to them. In such an economy nothing would change, things would go on as they always have.

Now some will be unlucky and end up with a clone that makes them way less than average. So to fix this the salaries of all the clones are put into one pot and redistributed as "UBI". Given the wealth disparity, most people would end up better off than before, the high earners of the past will be a bit worse off though.

But of course we are not in a perfect world, in reality there are many inefficiencies, and no company would accept a 100% AI worker tax. However, these AIs are presumably not just our clones, but 100x better. So we are in a position of abundance not scarcity. There's enough eggs for everyone. It becomes a question of distribution and policy

I just have one giant structured md file per project and have docs in the codebase

I've done graduate level RF stuff for the same reason, but I'm here to tell you just like everything else once you learn it and its demystified, it won't feel like "magic" anymore. That doesn't mean it wasn't worth learning, it was just as interesting as learning all about computers. It's just that the magic will always be in the unknown, and once you know it you'll just want to know something else.

The good thing is it's unlimited and all reachable if you spend enough time. Currently my interest is in chemistry, for example, that stuff is a complete mystery to me

To stay alive

> The problem is that I know how to work around these things

Well I'm sure you could come up with a few ways to work around your locking container but why isn't that also a problem? I think it comes down to the level of "activation energy" it requires, which may include breaking it open and also getting a new one.

In a similar way your digital block needs to have a high enough activation energy (both in getting around it and then turning it back on) that it effectively blocks you. It doesn't need to be impossible, it doesn't even really need to be that hard since we tend to be tempted with these things when feeling at a low energy (or we would be doing something else). As an example, I have the leechblock extension where turning it off would require uninstalling it or typing in something really long, I havent touched it in years.

Think of it as an aid to will power, you'll need higher obstacles the lower your wall of will power (a limited resource)

> Another issue is that sometimes sites have valuable information.

You should try blocking the feeds and recommendeds instead. Setup some regex rules, use ublock lists, get extensions which block the home page etc. But you shouldn't block them permanently, set just enough time to catch up / remind yourself that most of the content is useless then move on

I think you should narrow down the scope of what you want to be exceptional at to the point where it won't take up all your time. Then understand that it will still take a long time but maybe knowing that you are still making progress towards it will be enough to convince yourself to relax and "brainwash" yourself in the other direction

Would you say you're worried of not finding any position at all with that level of experience or do you still have options / fallbacks that you're holding on to?

> When integrated with tools such as formal proof verifiers, internet search, and symbolic math packages, I expect, say, 2026-level AI, when used properly, will be a trustworthy co-author in mathematical research, and in many other fields as well.

I always imagined math would be the very last field to be disrupted by AI, but to hear this coming from Terrence Tao himself (who I consider to be the face of math as someone not in the field) caught me off guard.

That's fantastic, it seems like such an essential feature that should have been part of the regular OpenAi UI on launch