Hacker Newsnew | past | comments | ask | show | jobs | submit | Jyaif's commentslogin

> badly in multiple previous launches of ML-based products

Which ML-based products?

> It was convenient for Google that OpenAI acted as a first mover

That sounds like something execs would say to fend of critics. "We are #2 in AI, and that's all part of the plan"


> Today the weird stuff tends to get buried

"skibidi toilet" would like a word


Note that this wouldn't have been useful for Factorio, because Factorio deals with the harder problem of needing deterministic results with varying amount of parallelism, whereas from what I understand Dthread only give the same results if you run the program with the same number of "threads".

By cutting the middle man out, MS could pay TSMC more than nvidia per wafer and still save money.


They can even pay Broadcom to be a lower-level middle man instead. Despite the BRCM tax, it'll still be way cheaper than going to Nvidia.


This is the whole game right here.


Some numbers:

During winter, France uses ~50% more electricity per day than during summer. And during cloudy days in winter, solar produces 10%-15% what it produces during summer.

If you don't have month-long battery storage, in order to be fully solar based France would need to produce 20 times more electricity than needed during summer.


> France would need to produce 20 times more electricity than needed during summer

So, it's ~15 years away at current growth rates?

But they'll probably just get months-long storage at some point.


> And during cloudy days in winter, solar produces 10%-15% what it produces during summer.

This doesn't matter. If you look at the monthly stats, solar panels in France produce ~3x more in the summer than the winter at a month by month view. As such, you only need 3x extra overall, and some day to day storage.


Or just balance the mix with some on-shore and off-shore wind which is anti-cyclical with solar.


Or you use a different technology optimized for long term storage. Batteries are not that technology. Hydrogen (or other e-fuels) or long term thermal storage.

For the latter, see standard-thermal.com


> Or you use a different technology optimized for long term storage. Batteries are not that technology

I've heard this before but can you explain why? A cursory web search tells me batteries hold charge pretty well for 6 months. And the new sodium batteries from CATL are certainly cheap enough.


For long term storage, capex is king, not round trip efficiency. The capex of batteries ($ per kWh of storage) is much too high. There aren't enough charge/discharge cycles to amortize that capex. This is unlike with diurnal storage, where there are many thousands of cycles over which to spread that cost.


Thank you that makes a lot of sense.


To go into more detail: you want to really aggressively minimize capex, even at the expense of round trip efficiency, for long duration (like, seasonal) storage. The interesting example from Standard Thermal is something that could reach a capex of $0.10 per kWh-th of storage capacity (yielding heat at 600 C). This is three orders of magnitude better than batteries. Even tossing in 40% round trip efficiency from converting this back to electricity using steam turbines it would be vastly superior to batteries for seasonal grid storage.

There are varieties of batteries with somewhat lower capex and lower charge/discharge rates. Form Energy's iron batteries are of this kind. They would occupy an intermediate timescale, perhaps ~1 week, which could be good for leveling wind output.


The problem with batteries for long-term storage is the capacity. You would need an ENORMOUS amount of them to store months worth of energy.


France has 70% of their power provided by Nuclear.


> in order to be fully solar based

I don't think anyone is suggesting that.


> There is no one best way

I think that the laws of physics dictate that there is. If your developers are spawning the galaxy, the speed of development is slower with continuous development than with pinning deps.


We don't know the laws of physics though. We just have models which can both fit into human brain to a point that they are surprisingly good for what humans have an opportunity to experiment agaisnt. That is really awesome, but it doesn't mean we know.


It’s partial knowledge.

Saying “We don’t know.” feels more wrong to me than “We know.” (emphasis on the periods).


The offshoring has started happening in the last 2 years in some of the big companies, by for example opening offices in Eastern Europe.

I suspect it didn't happen before because these companies were more focused on growth than efficiency.

That being said, thanks to AI parts of the big companies are again focused on growth at all cost.


Have you considered using wasm as the foundation for Roblox, instead of Luau?


As it stands, we already have a high-performance, sandboxed VM that we are maintaining successfully, and our editor environment is decidedly _not_ multilingual for historical reasons/lack of investment. It'd be very, very expensive for us to see any of the advantages of wasm for the platform today, and it wouldn't really do anything about our existing need to support the millions of lines of code written in Luau today.

Also strategically, wasm is a massive project coordinated by a large number of companies, and it doesn't seem especially prudent to bet the success of a single multibillion dollar company and their entire platform on a project that they don't control the destiny of.


Agreed. I think an underappreciated aspect of choosing a script VM in the space Roblox is in (user-generated content where scripts are content) your product is at the mercy of whoever controls your scripting implementation.

The scripting engine is an integral part of your product, and you need to "own" it end-to-end. Any bugs that creep into new versions of your scripting engine, any API breakage or design changes that impact your usecase are things that you are responsible for. Roblox owns the entire toolchain for Luau, and it's relatively small compared to the set of libraries required to compile to and execute WASM in a performant way.

The nuances of your typical JITing WASM runtime or V8 are pretty hard to learn compared to a simpler VM like Luau, it's a big reason why I've used Luau in my own projects.


how sandboxed is Luau really?


enough that we run a multibillion dollar business entirely off of executing untrusted code on our hardware


LunarEngine is built on raylib, which compiles to WASM. FWIU it might be possible to compile a Luau game to WASM with LunarEngine eventually.

"LunarEngine: An open source, Roblox-compatible game engine" (2025) https://news.ycombinator.com/item?id=44995147


In 10 years the mentality evolves.

It used to be considered extremely rude to pull out your phone during a conversation, now all the under 20 do it.


“evolves” is doing an awful lot of heavy lifting there


In fact, we are devo.


20 years ago the world wasn't so tense and deeply polarised as it is now either. I think many people would object more now, not less.

Google glass isn't that old though. It started in 2012 with selected testing users and in 2014 it became possible to just buy one.


I could play mp3s on a 66Mhz CPU, though yes other tasks would make the playback stutter.


I was doing it on Win 95 or 98, maybe the OS was using more RAM than yours?


Possibly, I was on System 7.1.


Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

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

Search: