Rust & Python are nice, but Mojo has Magic

[@codetothemoon]

Start your journey today with 7-day free trial of Mem here: bit.ly/mem_ai_cttm

[@devon9374]

There is no way you listed Chris's accomplishments without mentioning his greatest, LLVM 🤦

[@codetothemoon]

omg good point! maybe I should have included that instead of the Lex Fridman interview 🙃

[@Serizon_]

I know right , llvm is his de facto achievement

[@pookiepats]

Yea and unfortunately all his languages abuse the hell out of it, he creates behemoth’s and moves on, leaving us to drag them across the pragmatic barriers. Mojo will be as “mainstream” as Swift became Swift on the server my ass

[@fabienpenso]

@@pookiepatswhat would you use on the server side then? Love swift rust ruby but looking at other options. Elixir?

[@pookiepats]

@@fabienpenso Elixir is tempting but I have been using Go. there’s a library called “gomponents” that just 1.0’d (SSR in pure Go) + pocketbase + HTMX has been very effective. I love the notion of a single binary

[@MW-mn1el]

Mojo is not truly open source, compiler and debugger is not open source yet, and license is confusing with vague "Non compete" clause. Design is bettter then rust, but still missing many stuff. Unlikely to get any significant adoption outside AI, and back by venture capital is not god sign, even it's created by Chris Lattner. The licence, community backing and ecosystem is it's biggest hurdle for general adoption.

[@melodyogonna]

Modular did not get investor backing to create Mojo; the company did not intend to create a language at all. They being backed for the AI Cloud Inference infrastructure they're building. As for open source vs closed source, most languages start with being closed - Rust for example, started in 2006 but was opened in 2010, that is 4 years of closed development. Modular announced Mojo after less than a year of development because they want to build in the open, Chris Lattner had a bad experience with Swift being closed until the Swift team deemed it 1.0.

[@perc-ai]

Mojo is the fastest language ever created... I think we need to give it more respect. Rust is still thinking in the old ways, all the magic happen at the data center level Mojo is truly the future.

[@MW-mn1el]

@@perc-ai That depend "fastest" for what, that doesn't change the fact it's cool but still lacking the community support Rust have. Without community, vibrant ecosystem it's just "cool" but not really useful. If not enough developer adopt it, it's hype will run out and new language will replace it few years from now, then suddenly mojo isn't cool anymore.

[@MW-mn1el]

@@melodyogonna It's just distraction. When people creating issue and ask Modular to clarify when mojo going be "fully truly" open source, what's the roadmap for that. The response is coperate "non compete" vague response, and they don't want to commit to open source toolchain, just standard library for the time being. Everything is mix with "Max" platform under commercial license. Go read yourself what't in their commercial license.

[@noomade]

why does zig have so much community support?

[@rickdg]

If you don't know anything else, this video doesn't tell you why the mojo version is potentially better.

[@codetothemoon]

Yeah I did a bad job explaining this. Basically when I made the last video you could do almost nothing with it. I don't even think there were basic data structures like maps, and I don't think Python inter-op was working yet. for more details check out the previous video. But in this one I was able to make a simple program that does language model inference. Development has come a long way.

[@melodyogonna]

MAX is not built entirely in Mojo as Mojo is not mature enough to build such a critical infrastructure, It is built in C++. GPU kernels for MAX are what are written entirely in Mojo

[@codetothemoon]

I didn't realize this - thanks for pointing it out!

[@Ivoshevo]

Have you seen the source code?

[@melodyogonna]

@@Ivoshevo Modular is pretty open about these things if you're in the discord. You have to remember that they were already building the Max infrastructure in C++ before Mojo was conceived

[@mathiaz943]

0:39 - 😂 impressive track record: having been interviewed by Lex Friedman…

[@codetothemoon]

it's a big milestone! 😎

[@lx2222x]

I am studying data science and I am looking forward to use Mojo in the near feature!

[@codetothemoon]

nice! definitely one to keep an eye on.

[@TheLostBijou]

at 4:40 you appear to be complementing python, but its really a compliment on that specific API you are using, which has nothing to do with the language. Earlier you said CUDA was a programming language, but its actually a framework and an API.

[@codetothemoon]

you're right - the simplicity and brevity of the Python version is in large part due to its great ecosystem of libraries, but I think it's also the language. I think I said CUDA is an extension of C right? The point is it's a distinct platform, and there can be consequences of that - the need for SMEs, thought around integration points, context switching, vendor specificity, etc.

[@TheLostBijou]

Yes, admittedly these are minor issues in what is an informative video. Thanks!

[@Nekroido]

Before I unpause the video Imma click that bell icon. I totally forgot about Mojo. What a timely and lovely feed recommendation. And thank you too! Now, back to watching 👀

[@codetothemoon]

nice, really happy to have you onboard!

[@kmaximoff]

Hey thank you for all of your content! I love watching Rust related content !

[@codetothemoon]

thank you, really appreciate the kind words!

[@PouriyaJamshidi]

You should really take a look at Nim. Syntax of Python, performance of C++/Rust

[@codetothemoon]

I took a look awhile back and was definitely intrigued, but I wasn't sure how its configurable memory management strategy was going to work long term. Maybe I'll take another look!

[@PouriyaJamshidi]

@codetothemoon you can always let the default memory management do its thing. We migrated a bunch of our applications at work from Python and Go to Nim and were are quite happy with less resource utilization and speed gain

[@PouriyaJamshidi]

@@codetothemoon you can always let the default memory management do its thing. Never came across any scenario that needed us to mess with it. I have done some benchmarks on its ARC and ORC models and the results were quite similar. We have migrated a few of our applications from Python and Go to Nim at work and the results are very satisfactory. The resource utilization has dropped significantly while the performance has increased.

[@pookiepats]

@@codetothemoonit has defaults (ARC), you don’t have to touch any notion of memory config to write programs with it

[@system64_MC]

Nim is the goat! I made my final school work with Nim

[@AliMoeeny]

interviewed by Lex is before created Swift :D Keep up the good work, thanks for the great videos

[@codetothemoon]

lol, I'm a Lex fanboy, so I'm a little biased 😎 thank you!

[@dobleuber]

Before Rust, I tried mojo. I love it, but it was in a really early state, only available in the playground, so I focused on Rust again.

[@codetothemoon]

yeah it was hard to do much in those days, good call focusing on Rust 😎

[@DiaaKasem0]

when I tried looking into mojo and similar simpler rust alternatives ... I found out about Nim and V / vlang .. for me I see Nim is already there and does what mojo is going to do ! also vlang does that ( yet not python like - but very simple ! ) - both transpile first to C then to binary .. I don't really get what would mojo add to the mix

[@codetothemoon]

very good question - I can't speak too deeply on nim as I've only looked into it briefly, but I loved what I saw so far. In the nim community, I'm not aware of any efforts around developing a unified alternative to GPU-specific programming that is also isomorphic with the higher level code. I also don't think they have any goals of Python interop at the syntax level, nor in the package management tooling. It seems like both of these are central and unique to Mojo's goals. Similar thing with Vlang.

[@DataPastor]

Mojo’s license (with the competitive clause) discourages me from touching it. Otherwise technically a good direction.

[@codetothemoon]

Thanks for pointing this out, I haven't taken a look at the license. I have some preliminary thoughts but I'll do that before commenting on it 😎

[@harrynair1811]

Can I actually build a production pipeline with Mojo? Would you be able to do a video on it ? The whole shebang one ? Using magic to kickstart a workspace- docker to build an image and run on kubernetes

[@codetothemoon]

I don't see anything precluding it, as Docker and K8 are stack agnostic. But I haven't done it myself. Would love to make a video on it at some point, but it won't be anytime soon I'm afraid 😢

[@harrynair1811]

I love your content - I have been trying to push Rust at workplace for long. Side note have you tried Julia ? What are your thoughts on that ?

[@codetothemoon]

thanks for the kind words! props on trying to push Rust at work, I imagine that can be challenging in some situations. I have tried Julia - I remember being impressed with it, but I don't think I personally have a use case for it

[@GuiseppeZanotta]

Javascript rules the web and Python is the undisputed king of all things data science AI/ML. The genius thing they did was build on existing infrastructure. I just threw Rust in the dumpster where it belongs along with Haskell and picked up Mojo!

[@codetothemoon]

Agree that prioritizing python interop was a great call. Disagree that Rust is dumpster bound 🙃

[@ToanNguyen-ue8rw]

I don't understand why we would need mojo for AI stuffs. The speed you gain is minuscule compared to inference time.

[@codetothemoon]

I think there are many factors beyond just speed!

[@empathy_monster]

I personally don't understand why everyone is so critical that the Mojo compiler is not yet open source. What Chris Lattner and team are doing with the newly developed MLIR has never been done before. Think of it like UNIX, and now we have open source implementations that revolutionized computing forever, and Mojo aims to do something similar with programming.

[@codetothemoon]

I agree. I think open source is great when possible, but I don't think we should ever feel "entitled" to it. If a company makes substantial investment in a new technology, they should not have to spend time worrying about whether someone else will take what they've created (completely for free) and doing something with it that threatens their existence. If this is tolerated, we wouldn't see companies making big investments like this anymore. Because why would they?

[@be1tube]

Mojo is making some bad initial decisions. For example, the type of (i or s) where i is an Int and s is a String is String, and it silently converts. Silent conversions have caused so many bugs in C. Instead the type should be (Int | String).

[@codetothemoon]

oh interesting, I actually didn't know about this. thanks for pointing it out!

[@glamoagency5642]

Mojo has lots of funding AND Chris Lattner, so it's guaranteed pretty much.

[@codetothemoon]

these things certainly give it a lot more credence!

[@shivaylamba5641]

Always amazing content!

[@codetothemoon]

thank you!

[@Quracrow]

The bread is a good decision.

[@codetothemoon]

bread?

[@Quracrow]

@codetothemoon Beard*

[@rasputindasilva858]

Everyday a new language borns... and always better than the hinder and we leap from one to other and never get expertize in any.

[@codetothemoon]

Definitely not advocating that anyone should drop what they are doing and learn Mojo. Just saying it's one to keep an eye on.

[@michaelmueller9635]

Didn't see any reasons for mojo.

[@codetothemoon]

If you want the performance of C but want a syntax that is closer to Python, I think it's the most promising game in town. That said, the vast majority of folks won't need it.

[@michaelmueller9635]

@@codetothemoon That's the (or ...my) problem. Only to keep Python (or the python guys) in the game, doesn't justify mojo. In my eyes, python is a scripting language for c-files, which got a bit out of control. The only longtime fix for this issue is to get rid of the scripting language (...python). And then we are in the c-world and its slangs (c, zig, c++, rust, ...). But mojo as an extended python makes no sense. So maybe it's just me and history will tell something different, but I do not see anything convincing to change my mind (to the things written above).

[@knowledgedose1956]

​​@@michaelmueller9635 zig is in its zigote 😅 phase, lots of breaking changes between versions. other than that, FYI Python is used a lot in ML, data science, science in general, because it's starter friendly. And calling , for example scientists "python guys" is not exactly correct. spending a lot of time to learn a programming language for them is probably less preferable than doing actual science. So they choose python for that matter(as well as other languages, like Julia, i am aware of that). Thus Python has it's own niche, and always will, or some other high level language most probably will. And science is just an example. So, saying things you said is just being arrogant, sorry.

[@AchwaqKhalid]

#Golang FTW 🙌🏻

[@michaelmueller9635]

@@knowledgedose1956 Nope. When you reach a certain level of complexity, you can not hide it from the users by choosing an 'easy' programming language. That's no arrogance. That's practise.

[@anamoyeee]

Off-putting ai images on mojo's official website make me never want to use this programming language... ehh

[@codetothemoon]

fatigue of AI generated images is quite relatable - they do seem to make quite a few images of the Mojo mascot. but probably not the best reason to dismiss the language 😎

[@synreader]

great stuff

[@codetothemoon]

thank you!

[@desireco]

You are not really selling us on Mojo here :) maybe next year...

[@codetothemoon]

Didn't intend to. My takeaway was that it's a great language but might not make sense for many projects yet. As you say, maybe next year 😎 But if you want to be an open source contributor, right now there is tremendous opportunity for that in its ecosystem.

[@desireco]

@@codetothemoon definitely appreciate all the details you covered.

[@desireco]

@@codetothemoon I think like you, I would love a solid language like Python without warts.

[@ambientswn]

Is Mojo open source yet?

[@MW-mn1el]

Maybe decade from now, maybe never 😄

[@ambientswn]

@@MW-mn1el Yeah, exactly my point. Adoption matters most. No open source, no adoption. Plenty of alternatives. While I don't doubt language is cool and all, having it closed source from the get go wasn't really wise

[@melodyogonna]

@@ambientswn why? Every popular language I know started off being closed.

[@sebastianospino4861]

are you dumb bro, is it not wise to start OS

[@khanra17]

They don't care. They focus on 'MAX' (some ai shit). Glad I didn't got too far with this shitty language

[@asdfmonstrosity]

How did you spend 12 minutes talking with only an "inconclusive, the libraries were different" as the outcome

[@codetothemoon]

there was a bit more to it than that

[@Ivoshevo]

thats my programming language

[@codetothemoon]

nice!!

Has exceptions...

[@andrewbuzz7308]

"Track record... interviewed by Lex Friedman..." Wow, what an accomplishment

[@codetothemoon]

It is! 🙃

[@lcssbr]

It requires Python to be installed so a big no for me.

[@codetothemoon]

understandable!

[@khanra17]

Last time I disagreed because I was hyped about the language. This time I disagree more. Mojo sucks now, they has no focus on the language & the reputation of language instead they focus on some 'MAX' shit. Not opensource etc. etc.

[@codetothemoon]

I can understand this point of view!

[@AntonioRonde]

Video did not meet expectations. It was almost exclusively about Mojio tooling, syntax and details of the language were only glanced at.

[@codetothemoon]

Thanks for the feedback! 🙏

[@DustinKrejci]

2 1/2 minute in and im wondering when are you going to get to the POINT AND TL:DR!

[@codetothemoon]

😂

[@ChuTuanAnh2k]

No, python syntax s*ck, mojo the same.

[@codetothemoon]

Using whitespace to dictate scopes is definitely quite polarizing....

[@edgardoarriagada9467]

Mojo with brackets plzzz 🙏

[@codetothemoon]

as opposed to using whitespace to denote scopes? I wouldn't be opposed to this...

[@gustavojoaquin_arch]

Cringe, this is just another bloated language

[@codetothemoon]

how so? I think its core intent is to avoid the bloat that exists in the current Python ecosystem

[@christianovergaard1081]

@@codetothemoon For the same reason you feel the need to remove my comment about the license issue.

[@knowledgedose1956]

​@@christianovergaard1081 dude, chill, KZbin sometimes deletes comments automatically

[@codetothemoon]

@@christianovergaard1081 I did not remove any comments. Generally speaking the only comments I delete are from bots

[@Taddy_Mason]

There's a whole bunch of technical deep dives on mojo and not to mention it's by the creator of LLVM and Swift, practically speaking if there was any one person on the planet that could make all the bold claims that mojo does it's Chris Lattner.

[@carlosbrown6208]

DE >> Learn Rust DS >> Learn Mojo

[@codetothemoon]

for now I'd agree, but I can definitely see Mojo moving into the DE space once the library ecosystem is a bit more mature. we shall see!

[@DataPastor]

Rust is fully irrelevant for data engineering. pydantic and polars were written in Rust, but besides that, Rust is invisible in the data space.

[@skuwamy]

@@DataPastor you didnt even know that hugging face wrote a library in rust.

[@DataPastor]

@@skuwamy I know it. It doesn’t change the fact, though.

[@saiba-b7l]

maybe, but in fact, this mojo just can run in macos and ubuntu for now, so I'll say: full of shit... You are promoting something that not everybody can use 😂😂😂, big W to rust and python.

[@codetothemoon]

I think restricting supported platforms can actually be a really smart move in the early stages of a product. It can cut down on the test scenarios the devs need to worry about, so they can focus on the core mission. Then add support for other platforms in one swoop later. That said, I don't think I'm "promoting" it, just sharing my experience with it for others that might be interested.

[@kmaximoff]

I am first 🎉

[@codetothemoon]

🥇

[@kaas99]

Golang FTW

[@codetothemoon]

Gopher spotted! 😎

[@ulrich-tonmoy]

i really hated python being the AI lang in case of performance and usability+platform js is better since all are library why not make for js also then by default you get the web support and then theres desktop and mobile app dev framework Mojo would be great system level performance+python level syntax i hate the mojo path system

[@codetothemoon]

js performance, while probably better than python in many cases, still isn't great. re: web capability, this is still possible via targeting webassembly, though I don't think there is any support for it quite yet. Rust provides a great model for this with its fantastic WASM support, and great frameworks like Leptos and Dioxus have sprouted up as a result of that. Maybe Mojo can do something similar.

[@ulrich-tonmoy]

@@codetothemoon wasm wont be the solution for frontend unless frontend development becomes like desktop development in which case people started to use electron same way might switch to wasm but not before that(i think)

[@zoom0211]

Dude, you sound like you missed some crucial info. Python is just a way of calling AI libraries/frameworks which are written in C/C++, it's just a thin shell. Also, one can always optimize the hot paths in python scripts with C, C++ or Rust. JS has some pretty bad characteristics for science stuff, its quirks might the main reason why it hasn't been picked for this purpose. Or, most likely, JS didn't exist on servers when people started building these, nowadays popular, libraries/frameworks.

[@khanra17]

😂 Really JS? JS & Performance should never be used in a single sentence!