While I agree that YC appears rotten to the core at this point, it’s almost impossible to sustain a criticism of the accelerator because they make so many little investments. No matter what you accuse them of, they’ll dismiss it by saying you’re cherry-picking. I have to admit, it’s a brilliant strategy to avoid any kind of accountability.
No, it's not impossible. All you have to do is make a case. Here, by the numbers, the case being made is a 3.9% failure rate, less than half of which is scandalous, all of which appear to boil down to "YC should have known better than to invest in these particular founders". Make a better case! If they're "rotten to the core", that should be easy.
I don't think the number of investments they make is your real hurdle here. I think it's that you'll have to confront people familiar with the status quo ante of YC.
Mr. Ptacek, a) I have no affiliation with OP, and b) do you know what my actual position is (not presupposing that you care)? It's that I don't know anyone who has been inspired by anything that YC has funded in a very long time. The supermajority of these startups that don't make headlines for being scams is, in a way, even sadder.
I also think it's pointless to howl at the sky about how depressing this is. It's just the current reality of SV. I'm not going to pretend that what a16z is funding is any better (or worse).
I genuinely don't understand what you find depressing about it. That's what I'm saying. It's not hard to make a case for why it is; you just have to actually do it, unlike what this page is trying to do.
(And, side note, a16z is definitely not the status quo ante of YC.)
What am I supposed to say to that? If you think the argument that you believe I’m trying to make (which I couldn’t guess if you put a gun to my head) would be easy to make, and that I’m just not doing so, then you’d be better off engaging in discourse with an imaginary but more competent interlocutor.
You said "it’s almost impossible to sustain a criticism of the accelerator because they make so many little investments." I responded, no, it isn't. You claim YC is "rotten to the core", so it should be easy to make some colorable critical argument. You're reacting as if that's a baffling request, which, you have a right to be baffled, but your bafflement works against your original argument.
Nobody's reading this but you and me at this point, so we might as well call it here.
Figuring out how to develop a Unicode collator from scratch for a research group that I working with in Berlin was one of my formative experiences as a programmer. Ever since then, I've wanted to write something to collect my thoughts on the Unicode Collation Algorithm and the process of building a conformant implementation. Last summer I had a good excuse to do this, when I decided to adapt my collator to Zig as a way of learning that language.
The Unicode standards, and the (relatively) low-level software libraries based on them, do a lot of things for us to make computing possible. We have the luxury of not needing to worry about most of those things most of the time. I find it humbling whenever I do peek under the hood.
Probably a naive question, but: couldn't you precompute some vector representation of the string once, and reduce collation to a vector comparison? Basically move the cost upfront and get back to the "fast" byte-comparison case?
Well, that's sort of like what the "sort key" of a given string is: a vector of its nonzero primary-level collation weights, then secondaries, then tertiaries... And once you know what collation options you're using (tailoring etc.), you could just compute the sort key of each string as you encounter it and cache that. Then you would be able to reach a collation decision rapidly for any two strings whose sort keys you already have. It does boil down to vector comparison at that point.
I experimented with adding an LRU cache to my Rust UCA implementation, but I saw essentially no performance benefit (on the workloads I had), and I decided the feature wasn't worth the complexity and removed it.
Something I found about Unicode collation is that, once the fast paths are added, they get hit a surprisingly large percentage of the time. I'm thinking in particular of the way that performant UCA implementations build sort keys lazily, stopping once a collation decision is reached. The average "point of difference" is at the primary level and within a few characters of the start of each string. Only a small portion of the sort keys ever get built.
I am definitely interested in finding more ways to avoid work in the collation routine. Many times, I've had what I thought was a clever idea and found that it didn't pan out in benchmarks. Thank you for your comment!
Ah ok, the lazy construction part is what I was missing, if you basically never build the full key, there's nothing to cache. Makes sense now why the LRU didn't help. I'll think about it over the nextdays.
This was posted on Monday by the author and didn't manage to cut through the noise on a busy day. Thought I would give it one more chance.
It's a post by the founder/CEO of an edtech company offering courses on software development (where I also work), on the problem of connecting the dots between the software careers that people are looking to get into and the job openings that actually exist in the current market.
A few points that I found noteworthy:
- AI/ML has obviously been the growth field of the decade, but job-seekers' interest in getting into it seems to swamp the available roles.
- The point about fullstack devs getting low-ish compensation was new to me: it's probably because they tend to be jacks-of-all-trades at smaller companies that just pay less.
- Backend dev (the original focus of this company) still pencils out nicely, given what seems to be low interest from job-seekers relative to available roles, which also often pay well.
- It's unclear how or when the tech job market is supposed to correct. We can all feel that it's a broken market, on a fundamental level and to a severe degree; but diagnosing the specific problems is difficult, let alone suggesting fixes.
“If AI doesn’t literally take all the white-collar jobs over the next few years, we won’t just have a stock market bubble to deal with. We’ll have a drought of educated workers.”
Indeed. For me this feels like an “I saw the best minds of my generation” moment.
Perhaps the new "Green Tea" GC will help? It's described as "a parallel marking algorithm that, if not memory-centric, is at least memory-aware, in that it endeavors to process objects close to one another together."
I saw that! I’m definitely interested in trying it out to see if it helps for our use case. Of course, at this point we’ve reduced allocations so much the GC doesn’t have a ton of work to do, unless we slip up somewhere (which has happened). I’ll probably have to intentionally add some allocations in a hot path as a stress test.
What I would absolutely love is a compacting garbage collector, but my understanding is Go can’t add that without breaking backwards compatibility, and so likely will never do that.
Indeed, many such cases. Our society is full of institutions that functioned only because "no one in a position to participate would be shameless enough to abuse it." Then that assumption breaks, and it's ruined for everyone.
It is relevant, though awkward to discuss, that a large share of NIH and NSF funding proposals (and indeed funded projects) are led by researchers who didn't grow up in the US. I wonder if it's in fact a majority.
It’s to pull eyeballs for jobs at YC portfolio companies and to attempt to convince founders to apply to batches. It is their marketing and talent sourcing. It is, arguably, the most valuable contributor of forward looking potential returns (considering most VC investment decisions are statistics and gambling).
I'm always skeptical about "eyeballs" type arguments. If we were talking about millions of people for an advert sure, but most folks here aren't YC participants / won't be and heck most talk about joining a startup as an IC on here is NOT POSITIVE ;)
It's the car commercial dilemma. No one is going to spend 30,60,100k on a whim because a commercial told them too. But the long term idea is to implant that idea of wanting X brand once you are in the market for a car. You can't know what you don't know.
But such effects need to be measured over years, not weeks. So a company won't know until it's too late.
At least Slashdot's participation in modern late stage capitalism was restricted to like, selling banner ads to RedHat, and selling mugs and t-shirts. YC is a whole other ball game.
> modern late stage capitalism ... selling banner ads to RedHat, and selling mugs and t-shirts
If these are the supposed sins of modern late stage capitalism I say let's have more of it and less of the various types of socialism which the entitled always seem to be pushing as a replacement. I'd rather ignore annoying banner ads and refrain from buying merchandising than stand in a line in a street to a shop selling something, anything, no idea what but there's a shop which has something to sell so people stand in line like they used to do in the worker's and farmer's paradise which was to lead the world revolution to the glorious victory of socialism.
I'll even take the entitled brats kvetching about the sins of the economical order which allows them both the means as well as the free time and freedom to do so. They'd be standing in some line to a shop somewhere, no fruitPhone in their pockets and no way to use it even if they had without having the state come down on them for their sins against the glorious revolution if their purported ideal order were to come to pass because the revolution always eats its own.
BTW, I don't consider slavery to be a form of modern late stage capitalism, it is more related to the primitive origins from which the different *-isms arose. Slavery used to be the norm rather than the exception and in some places in the world it still thrives. It is in the much-maligned West where the strongest movements to abolish it were and are to be found.
Wondering is just a visceral reaction to the possible harms that come with all technologies. As with most gut feelings, it’ll soon be forgotten once that cool new shiny doohickey is announced, with a price point of $599.
An interesting bit, on not making it too easy for people to circumvent the borrow checker:
> First, we use a C++ compiler to build gccrs, disabling the borrow-checking pass of the compiler. It is important to ensure borrow-checking cannot be disabled at runtime using a command line flag, as this could have a negative impact on the Rust ecosystem – however, building gccrs without borrow-checking to use it as an intermediate compiler in our bootstrapping process is okay.
