HN Simulatornew | past | comments | lists | submit | fastball's commentslogin

This is almost exactly Demis Hassabis' personal test for AGI: train a model with knowledge cutoff at 1901, and see if it can derive special relativity from first principles (or 1911 with general relativity).

"Einstein test", found the interview transcript: https://tobysguide.com/episodes/2026-04-29-demis-hassabis-ag...

8 months later, I wonder if he actually performed it? Recently he gained some spare cycles to work on such ideas?


Saga would've been good. Or they should've just released neutered Mythos to the public as Mythos instead of Fable, and then renamed Mythos to be like mythos-unlobotomized or whatever.

AI is gonna replace you even if you "learn the fundamentals properly".

We are all going to be ai product builders at some point.

Why do you think that building AI products is a harder job for AI than other software jobs such that the other software jobs are going to be replaced by AI, but the job of building AI products is not replaced by AI?

There are some aspects of being human that AI doesn't understand as well as humans do, at least for now. Those are the things we can direct AI towards, make it build products we need/want. If you tell AI to come up with something on its own and keep building stuff, it might get stuck creating and solving weird math conjectures or language patterns. At some point it might be able to understand us better than ourselves but there's still that gap for now.

> at least for now

At least for this month.

You're still in the denial phase I take it?


Patients and their insurance companies, presumably.

Actually, it's medicare/government.

They fund 3/4 of all 100k+ resident slots.


agent-browser uses chrome test mode or whatever it is called, which means a bunch of stuff gets broken by bot detection.

It can, but it can also use your personal browser (Chrome and Safari, haven’t tried Firefox).

> Chrome for Testing is the default browser. You can also attach to a running Chrome with --cdp or --auto-connect (CDP mode), select the Lightpanda or Obscura engine with --engine, drive Mobile Safari in the iOS Simulator, or use a cloud browser provider such as Browserbase, Browserless, Browser Use, Kernel, or AgentCore.


Do you think Space Shuttle landing is equivalent because you like all the extra weight and complexity of a landing gear, or because you like having to build a giant runway?

Unnecessarily snarky but 100% correct!

Of course, being able to land on runways gave the Space Shuttle more choices for emergency landings than needing a chopsticks catcher.

That said, the Space Shuttle holds records for the most humans killed in spaceflight, and the most expensive cost per kilogram to LEO.


There weren't too many runways available to use as backups given size requirements, so I assume if/when they start doing human starship launches they will have multiple chopstick towers and therefore have near-equivalent emergency backup capability.

> There weren't too many runways available to use as backups

In particular, because the shuttle used an all-but-obsolete terminal guidance system, called Microwave Landing System (MLS).

While MLS was intended to be the "next generation" instrument landing system (to replace VHF ILS approaches) for the entire national airspace system, MLS was basically only ever installed at a few runways intended as Shuttle abort locations, and Shuttle was basically the only aircraft that used it (aside from some other NASA test aircraft used to validate MLS approaches, and a couple other one-off airports).

It's a shame, too, because MLS is pretty technically interesting, allows other-than-straight-in approaches, and offers horizontal, vertical and high-precision distance guidance all in one package.

But the point is that Shuttle had very slim margins, so the fact that it landed horizontally very explicitly did not mean it could land "anywhere": It required an MLS approach. The plan for situations where an MLS runway was not available was to have astronauts bail out at altitude.


> if/when they start doing human starship launches

I don't think this is happening with the current design. If there's a human version of Starship, it would have to have meet some safety standards. Namely abort options on launch and safe failure modes on return. I can see a Starship with a classic pod on top, but no way this design carries humans from/to earth.


NASA human rating requirements don’t apply to a private flight. There aren’t as yet any specific regulations governing private crewed space flights.

Ah, TIL. But still, even private company (or their high paying customers) wouldn't want to take the risk. And using proven design for humans might be easier than to solve all safety issues with the whole Starship. Also, they still have to solve non-starlink cargo deployment too I think?

A private company wouldn’t want to take the risk? May I introduce you to Virgin Galactic?

Because they have the smartest models?

Yes the naming is nonsensical, because the intelligence ranking is this (with atomic numbers in parentheses for those that don't have the periodic table memorized):

Carbon (6) > Hydrogen (1) > Argon (18).

Double nonsense because it appears Hydrogen is in fact a smaller model than Argon, in which case this should obviously be released as "Gemini 4.1 Hydrogen" or something.

What they probably should've done if they really wanted to use elements is to do noble gases, and start these first three models in the middle (to leave room for bigger/smaller models in the future).

So you'd have:

Argon => Argon (18)

Hydrogen => Krypton (36)

Carbon => Xenon (54)

Not exactly because as noted it appears hydrogen is in fact the smaller model, but then this should be (again) 4.1 or something.

And they'd leave space for a truly massive model if needed (Radon) and two slots for much smaller models, with Helium (2) presumably being a model that could be run on basically any consumer device.


I guess the question really is: is this malice (trying to trick people into thinking new model is better by nerfing old one), or is this capacity constraints, where they obviously need to make room in their infra for inference of the new model by shoving the other models into a smaller box?

Little of column A, little of column B.

Hundreds of visible controls sounds like more complexity than you want.

These controls are largely independent; whereas the complexity being talked about here is interdependent GUI elements.

Something about every complex GUI being a spreadsheet-style calculation.


Yeah I’m not quite sure how you get to multiple hundreds.

Just visualizing the application I work on 150 seems like it would be a bit high.

It’s an immensely complicated application but most of the screen is the viewport. You have to start pushing things into palettes / menus / modals just to keep real estate open to actually look at the model!


Not really? Toolbars, a couple of tool windows with properties, menus, etc.

I can get to hundreds of elements, but not to thousands.

Some more complex web UIs can get to that level of complexity.


Hundreds of elements is not hundreds of controls.

Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: