1,075 karma · joined November 16, 2021
It is worth being aware of course, but to my knowledge we have no evidence that Federal law enforcement has used anything other than old fashioned detective work to solve any Tor related cases.
There's always the possibility of parallel construction, but I would expect some reasonable evidence supporting this in some high profile busts.
I bet I can condense this even more: it's yet another FAANG shibboleth that starts out as a good idea and the evolves into a ritual candidates need to perform until they can identify that they are "in".
Google Algorithm/Whiteboard interview has morphed into a generation of robots that can solve any LC problem but are incapable of building anything real in software.
Currently I think there are more books out there on "passing the systems design interview" than there are on actual systems design.
I guess this situation just universally sucks.
I'm not one to defend the actions of corporate leadership most times, but it does look like Salesforce is doing what the market is asking for: bringing up profits.
And if we assume an average TC of around 250k per engineer, that's 40 employees of 8,000 laid or half a percent of the employees could have been spared by letting go of McConaughey instead. In case they're mistaken it seems a lot easier to hire 40 $250k engineers again then get McConaughey back after breaking a contract.
Meanwhile the company I'm at is just learning that publicly traded companies can't live off runway alone and that they need to become profitable (something they've never done) and please investors. The "leadership"'s failure to act is causing increasing stress in the company, and personally I'm finding it more and more clear they don't really know how to run a public company.
There are so many companies that look like what I described I'm not remotely worried about anyone de-anonymizing me over this info.
I would much rather be at a company like Salesforce right now. At least there's some evidence the layoffs were effective and anyone surviving, after the recent earnings, is probably already starting to feel more assured about their job. When layoffs do eventually hit me where I'm at now, even if I survive I'm not going to have much confidence in the future.
The beauty of JAX is that basic usage is basically a single function: `grad`.
You just write whatever Python function you want and can get the derivative/gradient of it trivially. It gets a bit trickier when you need more sophisticated numeric tools like numpy/scipy, but in those cases it's just about swapping out with a JAX version of those.
In this sense JAX is the spiritual success to Autograd. However the really amazing thing about JAX is that not only do you get the autodiff for basically free, you also get very good performance, and basically GPU parallelism without needing to think about it at all.
PyTorch is an awesome library, but largely focus on building Neural Networks specifically. JAX should be thought of a tool that basically any Python programmer can just throw in there whenever they come across a problem that benefits from having differentiable code (which is a lot of cases once you start thinking about differentiation as a first class feature).
It's undoubtably a great time for startups to get talent, but startups as we know them today tend to be very hungry for VC money, which is drying up rapidly. I would be very hesitant to join one unless the match was really good.
That's why you use a distribution to model the distribution of future possible prices given the information you have today.
The entire point of modeling the market as a martingale: you don't know what the future price is, but you do know a pretty good deal about where the price might go at various point in the future. Perhaps the single most important thing you know is that the future price is expected to be the same as the current price (ignoring the risk-free rate).
The Martingale property is very important to understand about stocks because if you are certain that the expected price of a stock is less than its current value you should sell, if you believe it will be more you should buy. An entire market thinking like this means that the current price should be equal to the expected future price.
Additionally these models don't "predict" future prices, but rather represent what the market believes about future prices given the current price and other information. This is essential to properly modeling risk and pricing assets.
It feels like tutorials on tailoring the emperor's new clothes.
At the same time I feel it's only a matter of time until my job interviews involve questions like "can you provide a few example of where you've used prompt engineering to solve a problem? Describe you prompt engineering process?" and I'm just not sure I can fake it that hard.
It makes perfect sense why everyone, sincere and fraudster alike, would want to promote SBF as a boy genius. It's a great marketing tool, it helps convince people who don't understand crypto to have faith in the product, and you also get someone who believes the spotlight should be on them, which is particularly useful for any fraudsters who would prefer to remain in the shadows.
Clearly SBF never questioned that praise (who would at the time?), and still is operating as though he were smarter and more special than everyone around him. The sad irony is it reveals that he's markedly less clever than most of the people he associates with. Caroline Ellison, at least from what I've seen, seems to have realized quite quickly that the smartest choice was to quickly realize she wasn't a brilliant as she had been told.
I've found a lot of tech workers (including myself) really need to learn how to read an earnings statement. Coursera has an excellent Intro to Financial account course [0] that I've found very helpful in understanding what's really happening at these places.
If you don't count salaries and operations. From the linked statement
> Income from operations was $9.4 million and operating margin was 6%.
Last quarter was DO first profitable quarter. If sales go down it's going to be hard to maintain that.
While I generally agree with the importance of geography I think this is glossing over a major part of the history of India and Pakistan.
The partition of India [0] by the British upon their exit from the region should be understood as a major tragedy in human history and very much lead to increased instability in the region.
Writing this off as "mere geography" is to miss the major and well documented impact of British colonialism on today's political climate in the region.
Companies generally try very hard to limit the warnings that layoffs will be coming. Every time I've seen layoffs it's a week or so after the "macro is tough, but we're still strong" meeting.
Ultimately this is the hardest part of layoffs: the uncertainty and lack of control.
You can't really know layoffs are coming until the day you get that email, and you also can't really do anything to save your job no matter how much you would like to believe you can.
It's a stressful season in tech to be sure.
I recognized that this was going to be happen last year, pulled my manager aside and explained how dire the situation and how we needed more visibility in the company. He thought I was being paranoid, but agreed with the general benefits of improving things. We've made some promising changes, but layoffs haven't happened yet so no idea if it worked.
If layoffs are happening leadership and the board are already drawing up plans for layoffs and nobody is thinking about you they're thinking of an org chart and a list of compensation to figure out what makes the most sense from a big picture perspective. They're thinking of a product road map and who is needed to make their current strategy work. Nothing you can do now will change that.
The time to convince leadership you were mission critical is well before you realize that layoffs are coming.
All that said, layoffs are more arbitrary that most would like to believe. Don't stress yourself out too much thinking about what you can do to safe your self/team. You never had much control over this situation.
I predicted there would be mass layoffs in tech right like this last year, and more specifically predicted 6 months ago that we'd be seeing basically what we're seeing as far as layoffs.
With rising rates investors are asking companies to show a clear path towards profitability. Most of these companies are losing more money each quarter and industry wide this has to stop.
For the larger profitable companies it's likely because they realize a huge amount of their revenue is coming from these companies that don't make money.
While I'll be the first to criticize companies for working to remove power from workers, what we're seeing is simply investors waking up to the economic reality that growth without profits is not sustainable.
As I've mentioned frequently before, it amazes me how many people looking at tech company financials will only look at Revenue and completely ignore profits.
What the revenue/profit plot you showed for TWLO clearly demonstrates is that there is no magic point there suddenly you increase revenue and profitably starts to increase as well. I think an assumption many people have is "if revenue keeps rising you have to be profitable soon", clearly that is not always the case.
Another pattern I've noticed is that many companies are not like Amazon, which is the model of a growth company. While AMZN does often reinvest its profits (historically to the dismay of investors) it has repeatedly demonstrated that it can be profitable. For companies like TWLO and UBER it's not obvious that they even can be profitable.
When I see this TWLO chart the message I get is that the only way to increase revenue is to increase net losses. It's not even remotely clear that TWLO can generate a profit at any level of revenue.
JAX is able to differentiate arbitrary python code (so long as it uses JAX for the numeric stuff) automatically so the backprop is abstracted away.
If you have the forward model written, to train it all you have to do with wrap it in whatever loss function you want, and the use JAX's `grad` with respect to the model parameters and you can use that to find the optimum using your favorite gradient optimization algorithm.
This is why JAX is so awesome. Differentiable programming means you only have to think about problems in terms of the forward pass and then you can trivially get the derivative of that function without having to worry about the implementation details.
It's not just the comp difference (though I do expect most people with severance are hoping to get something close to their old TC), but startups are notoriously fragile and at the same time wildly over optimistic about their futures in down markets.
I've talked to a few startups about interesting roles over the past few months and no matter how obviously in trouble they are if the market continues the way it is, they all deeply believe that they have an infinitely bright future.
A shocking number of post-seed but still awaiting series A startups have started reaching out to me and none of them have sane business models. On top of that it looks like VC funding is way down.
Any senior engineers who just got the shock of being let go from what they thought was a very stable job with great comp simply aren't going to be excited to jump onto a high risk startup just to relive the whole experience again only this time with less severance.
I've seen a lot of Python people sprinkle imports all over the place in their code. I suspect this is a bad habit learned from too much time working in notebooks where you often have an "oh right, I need XXX library now" and just import it as you need it.
The aggressive aliasing I do get since in DS/ML work it's very common to have the same function do slightly different things depending on the library (standard deviation between numpy and pandas is a good example)
But I personally like all of my imports at the top so I know what this code I'm about to read is going to be doing. I do seem to be in the minority in this (and would be glad to be correct if I'm make some major error).
If you keep an eye open on who's hiring on linkedin you'll see a surprising number of companies posting new positions will have layoffs the next week.
The reason for this is that you want to butter up the employees you're going to keep before layoffs. Getting promoted after you saw coworkers you respect get laid off makes you feel gross, and companies do need to do something to increase morale.
The funny part is many people at my co. believe promotions are evidence that we couldn't possibly get layoffs.
We've just gotten so used to the idea that "tech companies don't need to make profit" that we don't even question it any more. I still hear people shocked when a companies revenue goes up but their stock price drops while completely ignoring the reality that costs also went up.
The truth is all companies need to have profit and many of these tech companies have never showed that they're even capable of turning a profit.
What's happening is that investors are going to boards and telling them you need to make serious progress towards profitability. If sales are growing slowly or even declining there is no other option but to do layoffs.
IMHO this is still just the beginning, there's a lot of feedback in the tech ecosystem that are going to start playing out as more and more companies start scrambling to show profitability.
Similarly most non-marriage romantic relationship can be trivially ended "at will", but people get plenty upset when those end early than one party had anticipated.
More important though is the inherent asymmetry: it hurts employees much more to be terminated from an employer than the employer hurts when employees leave. You know, unless those employees organize and all leave at the same time. And we know well that companies put a fair bit of energy keeping that from happening.
Historically the way to resolve this asymmetry was an implicit understanding that you won't arbitrarily get terminated unless the employer has no other choice. This implicit assumption has been weakening over time.
The growing concept of "IC" goes hand in hand with the increased focus on "leveling" for engineers, as well as the rise of delegating all product related decisions to the rising class of PMs. All of these create a structure in which every team, even across companies, starts to look more and more the same and any individual can be trivially swapped out for another with the same role and level.
In the early days of the startup boom, ~2008, engineers at startups had a lot more control over the product process. Typical startups where a CEO with some general vision of the product and a knowledge of fundraising and a technical cofounder who would handle the real implementation of the product. CEO would make sure the company was health, CTO/tech cofounder would make sure the product was built.
But this lead to a class of labor that was a bit too powerful. As you likely remember engineers used to contribute a lot more than purely churning through tickets.
Today you can think of IC as a prefix to a more complicated part number: IC-SRE-4, IC-DEVOPS-2, IC-DS-6. Then you just have to find someone claiming to be the part and send them through a leetcode grinder to verify the build quality, then plug them in and go!
I'm still shocked that I hear non-tech people claiming they should put their extra cash into crypto. I guess there's still suckers out there to extract money from.
I know this comment will likely be downvoted/flagged by someone who still has some money to make from this con, but there hasn't been a coherent justification of the current crypto ecosystem is years. All of the early hopes have been repeatedly demonstrated not to pan out in reality and all that's left is grift.
Right because this time we're in a much larger asset bubble than we were back then.
At this point it wouldn't surprise me if we don't go into a recession because the larger the bubble gets the more desperately we need to keep some air in it. It would not surprise me to see cheap money return only because the alternative might, at some point, be the collapse of the entire systems.
We'll keep the bubble going so as long as we can, but the longer we punt this off the more extreme the breakdown is going to be.