There will be a paper coming out soon which benchmarks compilation time along with the compiled outputs against some popular compilers. The project lead is Oliver Flatt, I'm sure if you asked him he could let you know more.
There's also Chris Fallin's aegraphs (acyclic egraphs) which I believe is turned on by default for Cranelift. I'm unsure if everyone would agree that Cranelift is an industrial compiler but seeing as Fastly makes use of it I think it would qualify. aegraphs seem to solve a lot of performance issues while also getting a decent amount of benefit from equality saturation.
At the end of the day, and mentioned by the other commenter, cyclic egraphs really are still currently fun research tools. There's a lot of really smart people working on constraining the blow-up to make use of them in "real" compilers.
At a minimum, I expect they may become interesting offline or "super optimizers" (such as Souper) that run on performance critical vector/fp code. Because you can get rid of some of the phase ordering issues of traditional optimizers, egraphs can find some novel and really fast optimizations
I'm a software developer by trade, Computer Scientist by education. I spent much of my early career at startups working through scaling issues, and hardening their product. Most of my work has been done in Go, and it is my language of choice.
After graduating, I'm looking for a career move into more challenging low-level software and tooling development jobs (kernels, distributed systems, compilers, etc.). I wrote my senior thesis on formally verifying LLVM's ARM backend, have built a few compilers/interpreters, implemented Raft, hacked on xv6, and hack on the Go compiler in my free time.
If you are willing to take a risk on a passionate learner with a few related projects under their belt and you have an LLVM based compiler, LLVM backend, distributed systems problem, OS kernel, or Go compiler project, I think I could get up to speed quick and be a valuable team member.
Señor Wooly | Lead Backend Engineer | Remote | Full-Time
Señor Wooly is an eLearning company specializing in 2nd language acquisition of Spanish and French via story-based graphic novels and music videos. Students and teachers are incredibly passionate about our snarky and devilish stories and love acquiring language via our progression-based eLearning system.
We are looking for Lead Golang/Ops engineer. We are using a Go HTTP API hooked to Postgres,running on top of Google Cloud.
Needs:
* 3-5 years of Go experience developing APIs
* 1-2 years of production Postgres experience
* Someone who follows the philosophy of choosing solid and boring technology
* Good non-technical communication skills
Nice to Haves:
* Experience with a cloud technology (ideally GCP/AWS)
* Terraform experience
A little more about us...
* Our content is used by over 20,000 language teachers around the world.
* We have been in business for over 10 years, and our size and revenue has grown every single year. In 2021, our revenue increased by 30%.
* We make some of the most original, exciting, and demented content in the world of education. We’re the people who made a horror story about a girl who is terrified to go to the dentist only to find out the dentist doesn’t have any teeth. Our customers are passionate because, quite simply, no one makes educational content quite like ours.
If you are interested, email jennifer at senorwooly dot com with your resume.
So rephrasing what you said, unvaccinated individuals who are twice as likely to get reinfected than those with a who were vaccinated after infection means that vaccination doesn't always offer stronger protection than getting the virus?
This isn't exhaustive in the sense that it doesn't cover all permutations of vaccinated, infected, and, but it shows that at least infected + vaccinated is better than infected. That seems to meet the criteria that vaccination always offer stronger protection than getting the virus (at least in the vaccinated + infected vs infected populations).
But even a cursory glance at the study shows that the authors of the study knew this wasn't exhaustive, but they cite research [1] backing up OPs claim, and then add their voice to back up that vaccine > infection, vaccine + infection > infection, and make OPs conclusion in the Discussion section.
> This isn't exhaustive in the sense that it doesn't cover all permutations of vaccinated, infected, and, but it shows that at least infected + vaccinated is better than infected. That seems to meet the criteria that vaccination always offer stronger protection than getting the virus (at least in the vaccinated + infected vs infected populations).
It is consistent with that criteria, but generally “always” means something stronger than “we have evidence it holds in one case”. Especially if that case is the rarest permutation.
> But even a cursory glance at the study shows that the authors of the study knew this wasn't exhaustive, but they cite research [1] backing up OPs claim, and then add their voice to back up that vaccine > infection, vaccine + infection > infection, and make OPs conclusion in the Discussion section.
The paper you just linked was cited on the line “Reinfection with SARS-CoV-2 has been documented, but the scientific understanding of natural infection-derived immunity is still emerging” in the OP’s article. The closest line I can find to “back up that vaccine > infection” is an offhand “ Although such laboratory evidence continues to suggest that vaccination provides improved neutralization of SARS-CoV-2 variants, limited evidence in real-world settings to date corroborates the findings that vaccination can provide improved protection for previously infected persons” which doesn’t seem like a particularly strong stance for “vaccine > infection”. Especially when we get back to the original claim which used “always”.
And it appears that they may have been wise in not going that far, since now that we have studies in review that directly measure the endpoints we’re discussing it’s certainly not clear that this is true[1][2].
I’ll wait for those to get peer reviewed and more widely discussed before I’d be comfortable saying “in most cases infection > vaccine” (note I didn’t use the word “always”, which I doubt any researcher or clinician would) but the actual opposing claims in the papers you’ve cited are comparatively tangential to the original “always vaccine > infection” claim.
> Nikola was building barely functional prototypes of their trucks years before Tesla was building barely functional Semi truck prototypes, for the record.
If by "barely functional" you mean a metal skeleton on wheels that rolled down hills, then yes.
I'm not a shill for Tesla as they definitely have made very questionable claims in the past, but to claim that Nikola's semi is in any way similar to Tesla is a bit of a stretch.
This person is a troll. The Tesla semi is not “barely functional”. They have been spotted driving all over the place and doing real world testing on actual highways and in cold weather. Production was already to be started except for a shortage of battery cells (due to the extreme popularity of their existing cars and powerwalls/megapacks) which they are working to remedy.
And in 2022 it will be 2024. And you will continue to think that doesn't matter. If deadlines don't matter why can't Nikola claim the same thing? "It doesn't work yet, but it will?"
Because Tesla did not demonstrate fake trucks to get funding or take orders. The prototypes are not making materially false implications. Tesla does have the technology to make electric trucks, however bad they may be compared to Nikola which had a rolling shell.
Tesla and Musk do many bad things, but I'm sure if their truck was just a shell - they'd be in deep trouble too.
It still doesn't say how much it will be able to carry. They'll sell some for the marketing value, but no company will replace their diesel fleet with it
>They have been spotted driving all over the place and doing real world testing on actual highways and in cold weather.
Prototypes, with zero evidence they can achieve the claimed specifications. If you know otherwise, please, show us all. They are dependent on 4680 cells, which do not exist. They don't even have a factory, and they claimed the trucks would be ready for production 2 years ago. They will probably build something that works at some point in the future. Maybe Nikola could have as well? That is my point, where do you draw the line?
Now, what about FSD? And the solar shingles that "print money" that were faked on the set of Desperate Housewives and don't exist to this day? You know, they ones that Tesla shareholders are currently suing for? The ones that Tesla employees claim were fake in the deposition of said lawsuit? The ones you now buy are generic versions from China. What about the boring car tunnels; bit different from the old renderings, aren't they?
You have to be a troll if you are trying to convince people that a self propulsive vehicle, and an unpowered structure that can only move by being pushed or rolling down a hill are the same thing.
If you think the lawsuit is merely about rolling the truck down the hill, as opposed to lying to investors about their current tech, I'm not sure what to tell you.
There is zero evidence Tesla has a semitruck that with the range the claim at the price they claim to be able to produce it.
>That's nothing compared to offering a completely fake product.
I cannot wrap my head around this mindset.
So the truck doesn't exist in any manner in which they claim, be it speed, hauling capacity, range or price. But what they claim also isn't "fake"?
Let me spell it out, again: Tesla has raised money off of technology for which there's is no evidence they can or will ever achieve. FSD, solar shingles, the Semi's specs/price, the Cybertruck's spec/price (it's likely the Cybertruck as demonstrated isn't even street legal).
The fact that they have an unproven prototype roaming the streets is a long way from what they are selling. Why does the fact that they have one-off prototypes absolve them of this?
I've asked these questions multiple times, but the Tesla shareholders/astroturfers refuse to answer them:
1) had Nikola put their Semi cab on top of a golf cart frame to film the video (or better yet, a Model X frame ie no real underlying tech to match their claims), would that have been fine, because it "drove"? Sounds like it.
2) what about the Solar shingle demo? Employees involved in it claimed, under oath, that it did not function. It still doesn't exist today. The Buffalo factory is practically idle and shingles are generically sourced from China. Is that something you would consider "fake"?
> So the truck doesn't exist in any manner in which they claim, be it speed, hauling capacity, range or price. But what they claim also isn't "fake"?
Where did all those come from? You just said range and price before, and range is easy to fix.
And if a product costs more than expected to build, it still exists.
> shareholders/astroturfers
Shove off.
> solar shingles
They installed a couple dozen roofs, didn't they? Based on my existing knowledge I'd call it garbage but not fake. How long was this demo before those installations, and did any of the parts work?
>Where did all those come from? You just said range and price before
Who cares "what I said before". These are claims they've made that have no basis in reality. Claiming "we have a truck that can do all this" and can't is not that far away from claiming "we have a working truck" when you don't.
>They installed a couple dozen roofs, didn't they?
So you don't even know, you're guessing. They literally installed fake roofs that didn't function on a Hollywood neighbourhood set piece, then bought the company who claimed to be able to make these (a related party) within days. Is that much different than rolling a truck down a hill to raise money?
This information is free on the internet, people have been reporting on it for years. But the Tesla PR machine is strong. Try reading the law suit deposition, and then go try to order the shingles they advertised during the event. Then go visit the Buffalo factory where they were to produce these shingles (received hundreds of millions in subsidies for it). Then go look at what they are actually installing, years later, at jacked prices (generic shingles anyone can order from Chinese producers). And if you're still keen, you can read the reporting on the lawsuit happening right now about how all this came together to bail out Solar City.
Not sure about you, but slapping an inferior product together, years later at a higher price, that isn't what you demonstrated when you took the money, is fraudulent:
"We can't actually deliver the product for the price we told you". Huh.
We can also talk about the "Paint it Black" FSD video, which was heavily edited to demonstrate full self driving capabilities which still don't exist today (and may never, but for which they have received billions in revenue) as a marketing piece. After that, maybe the battery swap (earned millions in subsidies from California), which also was not a feasible technology with a faked demonstration. Here's a little read on that one:
>Based on my existing knowledge I'd call it garbage but not fake.
I guess one could argue the Nikola truck was "garbage" because it had no drivetrain, but not "fake" because it existed? We did see it roll down the hill, after all.
Nikola was a fraud. Other companies are doing the same thing, because it works.
>Shove off.
Boy, you were pretty quiet when people were accusing me of trolling and being a shortseller above, eh. Why is that?
I'm going to let you in on something cool: I'm not a shortseller, I have no financial stake in Tesla other than what I own, long, in S&P500 ETFs. Meanwhile, there are Tesla shareholders and owners in this thread and others, who don't offer any disclaimer when they praise the company.
Interesting how your preconceived biases can lead you to conclusions that are in fact the opposite of reality. Fancy that.
Because it's really hard to talk about whether a product is fraudulent or not if the list of problems keeps changing.
> Claiming "we have a truck that can do all this" and can't is not that far away from claiming "we have a working truck" when you don't.
To an extent, yes. With the huge caveat that production cost is not something a truck 'does'.
> So you don't even know, you're guessing.
I'm going by the best news articles I could find on short notice. If you have better, link it. I don't think it's reasonable to demand I go do physical tours before I can talk about the subject.
> They literally installed fake roofs that didn't function on a Hollywood neighbourhood set piece, then bought the company who claimed to be able to make these (a related party) within days. Is that much different than rolling a truck down a hill to raise money?
So with this I was able to find some info. It's true, there was a fake demo of a particular kind of tile.
But the product does exist now. Your claim that it still doesn't exist is just wrong.
It's also important to note that they had different-looking tiles that already worked, and by the time they took preorders it looks like things were working.
So that's bad, but it's not on the same level as nikola.
> Not sure about you, but slapping an inferior product together, years later at a higher price, that isn't what you demonstrated when you took the money, is fraudulent
If you charge more than you promised, yes. That's different from being unable to meet the originally planned production cost on something. If you let people cancel orders, which tesla appears to be doing, it's not fraud.
> the battery swap (earned millions in subsidies from California), which also was not a feasible technology with a faked demonstration
Wow, that's pretty bad! Thanks for the info, I'll bring it up to people in the future.
Still not on par with a fake product though.
> Boy, you were pretty quiet when people were accusing me of trolling and being a shortseller above, eh. Why is that?
I don't respond to every comment that is rude to a third party. Almost none of them, really.
> Interesting how your preconceived biases can lead you to conclusions that are in fact the opposite of reality. Fancy that.
You're making wild assumptions about what I believe and getting them wrong. Fancy that.
And you do/did come off a bit like a troll in some of your comments.
>but to claim that Nikola's semi is in any way similar to Tesla is a bit of a stretch.
Okay, how much of a stretch? If they would have stuck a nowhere-near-production ready battery and drivetrain system into the Nikola trucks and pretended it was complete, that would have been fine? That probably wouldn't have been difficult, but equally fraudulent.
https://github.com/egraphs-good/eggcc
There will be a paper coming out soon which benchmarks compilation time along with the compiled outputs against some popular compilers. The project lead is Oliver Flatt, I'm sure if you asked him he could let you know more.
There's also Chris Fallin's aegraphs (acyclic egraphs) which I believe is turned on by default for Cranelift. I'm unsure if everyone would agree that Cranelift is an industrial compiler but seeing as Fastly makes use of it I think it would qualify. aegraphs seem to solve a lot of performance issues while also getting a decent amount of benefit from equality saturation.
At the end of the day, and mentioned by the other commenter, cyclic egraphs really are still currently fun research tools. There's a lot of really smart people working on constraining the blow-up to make use of them in "real" compilers.
At a minimum, I expect they may become interesting offline or "super optimizers" (such as Souper) that run on performance critical vector/fp code. Because you can get rid of some of the phase ordering issues of traditional optimizers, egraphs can find some novel and really fast optimizations