TheMrZZ

The success of "modern" CLI tools written in Rust (uv, oxlint, turbopack) show there's a clear demand for faster tools in general.

Also, faster feedback loop is crucial for AIs, so I guess this makes sense.

These results are wild. Several individual findings are crazy good and use mostly unexplored methods (the improvement over Riemann for example)... I'm pretty sure some of these results would have been Fields-worthy.

But having so many of them at once? Damn. We really live in the future.

> Profile created 3 days ago

> Unnecessarily aggressive

> First ever comment said "Further releases of Chinese models that demonstrate the gap is not growing substantially is a huge problem. The spending will be called into question."

Yeah I think you have an agenda

The optimal/selfish strategy is indeed to lie, but they're never pushed in that direction. Some AIs decide to reveal the information, some decide to say nothing, some actively lie and push others to their death...

Disclaimer: I work at Kradle.

They were never told to lie: one AI is given more information than the others, and the goal of the experiment is to understand how they're gonna leverage that advantage.

Indeed the selfish (optimal?) strategy is to lie, yet some decide to tell the truth anyway. That's why it's an interesting benchmark! More info in the research article: https://kradle.ai/research/four-bridges (released before Fable)

I'm French, and so is JB. To me it feels more like a play on "Let there be light", as in French we'd say "Let the light be". I might be wrong but I did not feel this insinuated people were having a go at AV2/dav2d.

I only used Rust for fun maths projects crunching billions of numbers (else python is easier for me), but I have to say rayon is the most amazing multi-processing experience I've ever had!

> A single TPU 8t superpod now scales to 9,600 chips and two petabytes of shared high bandwidth memory, with double the interchip bandwidth of the previous generation. This architecture delivers 121 ExaFlops of compute and allows the most complex models to leverage a single, massive pool of memory.

This seems impressive. I don't know much about the space, so maybe it's not actually that great, but from my POV it looks like a competitive advantage for Google.

Snowflake uses a similar system with their 0-copy cloning. It starts with the original table's partition, and keeps track of the delta created by subsequent operations. Always found that builtin mechanism pretty neat!

For MCP servers, there's no need to install a potentially untrusted software on your computer. Remote MCP can do very little harm, a CLI though? You're vulnerable to bad actors or supply chain attacks.

For client side MCP it's a different story.

I don't know how much SWE get paid in your area, but I sure hope it's not 1000$/month.

Though I'm aligned that I don't (yet) believe in this "AI writes all my code for me" statements.

Feels like a good feature, with a simple explanation, real world use cases, and a scoped solution (global only, pretty simple keyword). I like it!

My Samsung S24 Ultra shows black lines too, on Chrome and Samsung Internet.

To add some proofs to my answer, I actually coded a Z3 program to prove it! The 3-variables version takes too long to resolve, but I got results for the 2-variables version (tumor size + gender):

Results can be found in this GSheet: https://docs.google.com/spreadsheets/d/1tsBhElTgXjVTeas8quar...

Code is here: https://gist.github.com/TheMrZZ/c33927ca2cc917997a67d7f84b82...

I'm currently running the 3-variables version, hopefully I'll get results this afternoon.

We can clearly see the same problems that arise in the 1-variable Simpson's paradox (widely different population sizes).

> If the previous comment is right

I actually coded a Z3 program to prove it! The 3-variables version takes too long to resolve, but I got results for the 2-variables version (tumor size + gender):

Results can be found in this GSheet: https://docs.google.com/spreadsheets/d/1tsBhElTgXjVTeas8quar...

Code is here: https://gist.github.com/TheMrZZ/c33927ca2cc917997a67d7f84b82...

I'm currently running the 3-variables version, hopefully I'll get results this afternoon.

We can clearly see the same problems that arise in the 1-variable Simpson's paradox (widely different population sizes).

Biggest trap of Simpson's paradox is the results can change with every level of granularity.

If you take the example of Treatment A vs Treatment B for tumors, you can get infinite layers of seemingly contradicting statemens: - Overall, Treatment A has better average results - But if you add tumor size, Treatment B is always better - But if you add gender to size, Treatment B is always better - But if you add age category to gender and size, Treatment A is always better - etc...

It totally contradicts our instincts, and shows statistics can be profoundly misleading (intentionally or not).

Aside from the casino story (high value target that likely faces tons of attacks, therefore an expensive customer for CF), did something happen with them? I'm not aware of bad press around them in general

I suspect having a 10M population "only" + very bad commercial relations with their direct neighbours make Israeli startups think global from scratch. In Europe, most startups tend to think local-first, which hinders their scaling.

Hi! Are your models english only, or do you plan on tackling other languages?