That's recipient testing based on mailbox name. I don't recommend that for spammers - its so trite and early 2000s.
I wont allow you to test deliverability to my email domains without you sending an email I can analyze and decide to allow or drop mid stream. I also get to drop it before you consider it sent. I obviously drop connections that just establish from and to and go weird after that.
The real problem is that OP is or wants to be an old school disruptor working at what used to be an exciting and disruptive employer (but isn't any more - its just a boring old money maker).
OP crank out a pretty decent and well received, by the community, product and get absolutely canned because they are well out of touch of how Google now works. You don't do risk (without reward) at Google and you certainly don't show a bit of ankle or look exciting. Google are well out of the market for being interesting (outside of the balance sheet and P&L for those who fetishise in accountancy.
Unfortunately: going viral isn't always a good thing as anyone who has experienced a nasty virus will attest.
> what used to be an exciting and disruptive employer (but isn't any more - its just a boring old money maker).
I feel sorry for this person, but I would be surprised if this would have been okay at Google in the past 20 years. It wouldn't have been okay at any company I've ever worked at, big or small.
I think there's a valid argument that this started as a simple DevRel script or trick, but due to the way you can write a lot of code very quickly with AI it expanded to something that resembled a full-blown product.
Maybe uncharted territory as the previous assumption was that an individual DevRel person releasing scripts couldn't be mistaken for a supported product because one person couldn't produce that much code in the past.
I would encourage this sort of thing in my company. I'm not google. I'm not legally beholden to anyone except myself and my business partners ... and my own sense (which is worryingly odd!)
Google can never be exciting or interesting evermore by design and intent. They dived on in and went "money" full on. They exist to generate revenue for their shareholders. They dumped the "Don't be evil" thing without blushing.
I think your encouragement is admirable but could be interpreted as naive.
For one thing, the author of this tool used Google trademarks (the logo) to represent the project.
If you are even slightly larger than a mom and pop small business you pretty much have to defend that trademark or else you risk losing it.
But, okay, fine, you can just tell them not to use your trademark and have them say it's not an official thing. No big deal.
The other thing I would say is that growing beyond even a relatively small number of employees fundamentally changes everything. Once you don't have that face to face with all your employees that trust level between you and them can't possibly be the same, no matter how good your intentions are. Even a modest company with 25-50 people...how well can you know those people, really? Even if you try your hardest to know them?
Once you have a certain number of employees you run into probabilistic realities.
Google has over 100,000 employees, which means statistically speaking a few of them have committed or will commit homicide. The idea of "we trust all our employees" can't exist from a mathematical perspective, even if the leadership happens to be the nicest people in the world who really want their employees to have freedom and autonomy.
>Google has over 100,000 employees, which means statistically speaking a few of them have committed or will commit homicide
Nitpick, but the set of people who work at Google is highly non-random. I am pretty sure Google employees are much less likely to confirm homicide - they are older (homicide rates after 24 drop sharply), more educated (strongly inversely correlated with homicide), richer (same), etc.
In a modest 25 people company you absolutely can just go and do this of your own initiative and it will be tolerated or encouraged. Any company that size where it's not possible will close in 18 months when it runs out of investor money without having accomplished anything. You should still mention you're doing it beforehand, though.
Dunno about 20, but 7 years ago, they fired a security engineer for forcing in a CL for their internal Chrome extensions to put a disapproving banner on certain anti-union websites. Wasn't a very harmful change, but because she left a clear paper trail of circumvented code/release reviews, she couldn't be trusted anymore.
That was a security engineer modifying internal security tooling without proper permissions/reviews.
The union piece was probably extra motivation but still you just do not do that to security infra, it should always be a firing offense unless it was a truly exceptional circumstance.
Conversely, this guy was in a DevRel role where it sounds like they released open source stuff all the time and the line was a lot more fuzzy (admittedly I've only heard one side of the story).
In this case I guess the only other option the person had is to quit to build something of their own or work at another more builder friendly company, but at least in this case they got to release a thing that was valuable to many people and got popular, so they got some good publicity out of that. So ultimately I don't see that their actions were wrong overall.
A nvidia spark thingie has 128GB unified RAM. They also have a dual port version of one of these things: https://www.nvidia.com/content/dam/en-zz/Solutions/networkin.... ie 2 x 100GB/s ports, they may even be 2 x 200GB/s. Once I've got my paws on one, I'll know more.
You can cluster these beasts too. Two and three (with two IP subnets) is fairly obvious. Four or more might need a switch depending on how much network latency affects things.
Apple seem to have forgotten about M series with gobs of RAM. I can't get the Apple shop to show more than 96GB of unified RAM and that costs a kidney.
I have one, and I love it. That said my buddies Mac smokes it for inference workloads in terms of tokens per second AND its more usable for other things.
If you are training and doing research it's great, if you want to cluster them it cant be beat, but if you just want local inference on a single box buy a mac or even a strix halo device.
Get your buddy with his smoking Mac to allow multiple concurrent connections and see how it gets on compared to your Spark. Don't ever use a single "chat" test to derive performance - try running say 10 or more.
You might also notice that your Spark has a pair of QSFP28 or DD (not sure yet) type interfaces as well as the 10Gb/s ethernet - that network card is a right old beast and adds quite a lot to the cost. It is capable of either 200 or 400Mb/s and can be split into two lots of four. Your mate's Mac probably has a wifi connection and is too cool for ethernet 8)
That NIC is there for a good reason - the Spark wants some friends to cluster with and you will absolutely spank any Mac when you spaff Mac style money on say three of these beasts and some cables and cluster them up. If you want four or more, you will need a switch and Mikrotik and others have them.
Casual "tokens per second" in AI is a bit like gamers whittering on about "ping" when they are using TCP and UDP for their games. ICMP request/response is a handy way of testing network paths and can give some indications towards potential performance limitations.
can those macs boot linux? i've heard about Asahi but have no idea how far along they are. i've got my fleet configured with nix and sure, nix can target darwin, but there's a _lot_ of sharp edges there: i don't really want to pull that thread unless i have to...
I don't know. I think he just uses LMStudio most of the time on his, but that's one place I can say the spark really shines for me.
I'm a Linux guy, but also don't always have alot of time. The Spark comes out of the box with a nice Linux distro that's pre-configured to be easy to setup and the guides and online resources make getting up and running trivial, for even some complex tasks. You would have to do a LOT of tinkering just to figure out some of the things the nvidia resources walk you through natively. They have guides for a ton of stuff that include the optimal settings so you don't have to figure it all out through trial and error.
Check out these "playbooks" for some examples. [0] There's a lot to be said for not having to piece all that together yourself.
Not the new ones. Only the M1 and M2 have good support for Asahi. But you really don't need it. If you need Linux, use a VM (UTM is free and is equivalent to KVM/QEMU in speed, despite being a Type-2 Hypervisor.)
pretty much any of them, dude, as long as you have enough RAM, since it uses unified RAM and a powerful SoC CPU/GPU. Literally any M-class model, but the M5 is currently top tier.
The DGX Spark has basically the same memory bandwidth as a M5 Pro, and far more than a M5.
Only the M3 Ultra really beats it, and once you start scoping out the cost of a M3 Ultra with 128GB or 256GB, the DGX Spark doesn’t look bad after all.
> The DGX Spark has basically the same memory bandwidth as a M5 Pro, and far more than a M5.
I see ~274 GB/sec for the DGX Spark[1], versus 307 GB/sec for M5 Pro and 460 or 614 GB/sec for M5 Max[2]. One might call 90% "basically the same", but there are nominally two tiers above "Pro".
Yes, a MacBook Pro with 128 GB and M5 Max costs $5100 (14") or $5400 (16") versus currently $4700 for the DGX Spark, but the MBP includes keyboard, mouse, battery and portability. I believe its prefill is slower and you get 2 TB vs 4 TB SSD, but overall one gives up a lot to save 10% of the cost.
I looked, but a sibling comment just provided the links. ~274 GB/sec for the DGX Spark, vs. 307 GB/sec for M5 Pro, and max 614 GB/sec (!!!) for M5 Max? Why would you completely friggin’ lie about this, or at minimum, not double-check your facts before bullshitting? Plus, you get a full-fledged computer along with it!
Apple could actually be a good deal and you folks would still make up something to not justify it. In a way, it’s amazing what Apple has accomplished- Baseless negatively-tainted perception in certain influential tech circles.
(To be fair, they’re kind of earning it. I’m glad Tim “Sweet T” Cook is departing.)
Plus, my original comment got downvoted despite being factually-correct. Thanks, Reddit. Oh, wait…
Yep. Memory bandwidth is what decides how fast LLM's generate tokens (mostly). The DGX Spark has something like 270 GB/s of memory bandwidth, and the m5 ultra is ~615 GB/s. Theoretically DOUBLE the speed. In practice he only generates like 25% more tok/s, but that's still very impressive.
The spark can fine tune models in 1/4 the time and excels at other compute tasks in ways that Mac never can. Plus the high bandwidth ConnectX-7 ports would be like $1700 to buy on a card just for the network adapters... But for generating tokens, it just plain loses.
In case anyone was wondering my spark is basically silent as well. It's great at being ignored, if that's really important to you. I've run mine completely headless since I bought it, including setup.
It is 2x200Gb/s physically but the PCIe bandwidth is basically only 200Gb/s so it may as well be one, and actually its a weird 2xPCIe4 not 1xPCIe8 so it appears in software as dual 100Gb/s. Its a bit odd.
DGX Spark is ~273GB/s. That’s about M5 Pro territory, and twice as fast as the M5. You’d have to go to the M5 Max, or M3 Ultra, to get higher memory bandwidth than the Spark.
I don’t know about laws, but Teslas automatically record and save everything in the event of a collision. My brother was hit by another car while driving a Tesla a few years ago and it was very easy to retrieve video from the Tesla’s cameras and show who was at fault.
Espresso is so named because you "express" the brew from the beans and you do that with water because water is pretty neutral in flavour, is not poisonous and has quite a few other properties that we have evolved to exploit or live with.
Milk is a weird liquid associated with mammals nursing infants. We humans have evolved to be somewhat lactose tolerant post infancy which is rare in animalia (1)
Given that we are using the Italian word - espresso - then let's use their definition. If you add milk then you have a latte or a cappuccino or an americano con latte, a flat white or whatever.
Real weirdos try to milk oats. I've tried but I can't find their teats.
Hippopotamus does mean river horse and I was caught out by that (note the o instead of a in ...poto...). I think that word is really a joke - lol - a bit like floccinausilihilipilification, which I wont bother looking up the speling 4.
I managed a paltry 90/100. Some of those words require a classical education and probably a British one at that. I studied Latin at two posh schools and have O level English Language and Literature (that's two qualis at age 16).
I'm pretty well read and know exactly who Sandi and Stephen are. Ironically Sandi is Danish but notably erudite (that turned up for me) and navigates her way around English with remarkable aplomb.
reply