Nils Cremer

Building chips

[Nils Cremer @nilscmr]

CPUs suck. We're building a new general-purpose chip that scales to thousands of cores while being more energy-efficient.

We're hiring hardware design engineers, consider joining us https://t.co/8GCN0bN0xN

What we do differently ...

[Nils Cremer @nilscmr]

Oh and in addition to that we pay ARM principal engineer base salary

[Nils Cremer @nilscmr]

People underestimate the opportunity for generational whealth you have as a founding engineer. We for example offer up to 2% equity and expect to be valued at a few hundred million USD in about a year, let alone in 4. I encourage you to actually do the math

[Nils Cremer @nilscmr]

@ErnestiSario I hate it when people give you model output rather than model input

[Andreas Klinger 🦾 @andreasklinger]

FR8 is a 12,000 m² palace, filled with geniuses researching or building startups, it charges 0% equity and even pays for your food, living, and flights. One of their founders drank actual poison on stage to demo their tech.

Welcome to FR8.

Nothing about FR8 makes sense because it’s so over the top in their ambition, but they might eventually become the biggest thing for young founders globally. And it’s happening right here in Europe.

They are neither a hackerhouse, nor a startup accelerator, nor a classic research lab.

Instead they think of themselves as a university-like institution for the post AGI world that pushes you towards building companies, ambition, obsession, and bias-to-action.

Think YCombinator, Stanford and Bell Labs all wrapped into one thing for the most ambitious 20-somethings in the world to work, run by 20-somethings.

They just came out of stealth. Until recently people didn’t even know where their latest cohort is based. Because additionally on top FR8 is absurdly secretive. Their target group knows them and that’s about all they care for.

We visited last week to join them behind-the-scenes as they prepare for their first demo day in their new building - a 5 floor university building in the middle of Helsinki.

We knew them for quite some time so we were allowed to film them as the first team worldwide. But even we couldn’t film multiple floors and rooms of their building. This video gives you an insight into the ambitious craziness that FR8 is – but trust me there’s more to come in the near future.

The biggest new thing in startups – isn’t in SF – it’s in the north of Europe and attracts young geniuses worldwide. Welcome to FR8!

[Nils Cremer @nilscmr]

As porting software gets cheaper, hardware becomes more fungible

[Jarred Sumner @jarredsumner]

there will be a blog post about this. on what this means for bun, benchmarks, memory usage, maintainability going forward, and also the literal process of doing this (it wasn’t just “claude, rewrite bun in rust. make no mistakes”)

this is a 960,000 LOC rewrite, the code truly works, passing the test suite on Linux and soon other platforms. e2e I started working on this 6 days ago. this would’ve been a massive amount of work by hand.

[Nils Cremer @nilscmr]

I'm all for understanding intelligence

[Lukas Bongartz @lukas_bongartz]

I looked into something related recently. Models seem to learn meaning via collective coordinates between tokens. This is quite similar to how we describe phonon modes (lattice vibrations) in physics. (1/9)

[Nils Cremer @nilscmr]

@shipfr8 home

[Nils Cremer @nilscmr]

FR8 is the single one place to be if you want to create the impossible.

What the team has managed to put together in the past year is just unreal and what will happen in the next decade is unthinkable

[FR8 @shipfr8]

We took over a former technical university.

This is Hogwarts in real life.

For people who want to work on something too early, too weird, too ambitious.

[Arnie Ramesh @arnie_hacker]

nils is probably the smartest friend i know :)

would not miss the chance to work with him to build the future of chips

[Nikos @kalio_late]

I had the luck to briefly talk with Nils at the SPH last year. During our conversation I remember thinking “wow” this guy is actually going to change the world!
This is a one in a generation company.

[Nils Cremer @nilscmr]

@osmarks1 The hardware has some similarities but we have a sane programming model

[Nils Cremer @nilscmr]

CPUs suck. We're building a new general-purpose chip that scales to thousands of cores while being more energy-efficient.

We're hiring hardware design engineers, consider joining us https://t.co/8GCN0bN0xN

What we do differently ...

[Nils Cremer @nilscmr]

@Paralell_Shazli Going across CPUs is very expensive

[Nils Cremer @nilscmr]

@StefanZeiger No the opposite, lazy evaluation kills parallelism

[Nils Cremer @nilscmr]

@GrigoryEvko We actually support affine type systems :)
Have a look at the Fork trait of Vine

[Nils Cremer @nilscmr]

@EmphyrioLives DAGs are static while interaction nets are dynamic.

The hardware has some vague similarities with TRIPS. The biggest difference is that they relied on a smart compiler and unfortunately compilers are not smart.

[Nils Cremer @nilscmr]

@itsclivetime You still have to decide whether you have cache coherency across your NUMA slices. If yes then you again hit scalability issues, if not your accesses become incredibly slow unless you have very embarrassingly parallel workloads.

[Nils Cremer @nilscmr]

@itsclivetime A big advantage is that we can manage memory in hardware without relying on cache coherency because of the linearity of interaction nets (it's the same reason Rust requires neither a garbage collector nor manual memory management)

[Nils Cremer @nilscmr]

@itsclivetime Solving memory becomes very hard then. You'll probably not be able to scale cache coherency to that level so one has to go with software-managed scratchpad memories which are hard to program. You can only do that with a few applications (that's what Tenstorrent is doing)

[Nils Cremer @nilscmr]

@botir33751732 When's the next one?

[Nils Cremer @nilscmr]

@gatelevelanon Fun seeing you reason through this, I'd love to chat!