Was open-sourcing Linux a cynical, offensive move to devalue commercial Unix (a scheme hatched by duplicitous Finns)?
But more seriously, DeepSeek is a massive boon for AI consumers. It's price/performance cannot be beat, and the model is open source so if you're inclined to run and train your own you now have access to a world-class model and don't have to settle for LLaMA.
It used to be baked into Google search, but they seem to have mostly fixed it sometime in the last year. It used to be that "black couple" would return pictures of black couples, but "white couple" would return largely pictures of mixed-race couples. Today "white couple" actually returns pictures of mostly white couples.
This one was glaringly obvious, but who knows what other biases Google still have built into search and their LLMs.
Apparently with DeepSeek there's a big difference between the behavior of the model itself if you can host and run it for yourself, and their free web version which seems to have censorship of things like Tiananmen and Pooh applied to the outputs.
Boy do I feel for that candidate. As a manager, while this is not the kind of question I'd ever ask, if I saw a candidate work through a problem like this I'd take it as a strong signal for a possible hire. Ghosting is a dick move period, but especially after a performance like that. I would like a follow-up discussion probing how open/accepting they'd be to skilling-up in their weaker areas, how they go about asking for help, try to suss out if they're prone to getting stuck in "ruts". But if I didn't see any red flags my vote would've been yes.
As software becomes more essential to a business its reliability becomes more important. If your customers can tolerate defects or downtime it’s a signal that:
A) you’re not providing any real value
B) You provide so much additional value compared to your competition you still come out ahead in the wash
C) Your customers are hostages via vendor lock-in
A and C are the most common cases of persistent bad software.
Interesting trend I've noticed: Tiktok's users tend to like its algorithm, and its probably the app's most valuable assets, but western tech executives tend to hate it and speak of it with derision.
This stands in stark contrast with US-based social media companies, where both its users and content creators often speak like they're at war with the algorithm, yet to the tech elite these sites algorithms are tuned to perfection.
I'd guess that it comes down to differences between the outcomes that either algorithm is trying to achieve. When westerners advertise they tend to provoke a sense of anxiety and then position the product such that it appears to relieve that anxiety. So we hate "the algorithm" because it's trying to make us uncomfortable without letting us leave. We should hate the algorithm.
I couldn't speak for Tiktok's aims, but they seem different enough that its algorithm doesn't chafe in the ways that we've come to expect.
It seems pretty simple. The Tiktok algorithm is designed to push content you want to see. In the US social media platforms are designed to push content they want you to see and everything you care about gets pushed out of the way. With US platforms you always have to scroll past garbage to get to anything you care about. Tiktok just relentlessly shoves what you want in your face over and over and over again, and when it does misstep it moves on to something else before you even have the chance to consider what you'd rather be doing with your time.
> In the US social media platforms are designed to push content they want you to see
Who is they?
Anyway, you're wrong. TikTok pushes videos you want to see, while US app algos push content you are most likely to engage with. These are not the same thing. In fact, content one most engages with is content that generates outrage. Try not to get angry when you open Twitter. It's not easy.
Whoever runs the platform. That's the value they see in their platforms: the ability to control what you see, when you see it, and how you see it. You might only want to see what your friends and family have been up to, and in chronological order, but they are going to make sure you have to constantly scroll past shit you couldn't care less about to get to it, especially when that shit is advertising.
Not only do they do this for marketers, they actually think that making their users regularly disappointed and frustrated is a good thing. They think that forcing you to hunt for what you want makes finding it more rewarding and while you're scrolling past garbage/ads or searching for something you just saw, and cursing the obnoxious algorithm for hiding what you actually want they call it "engagement".
TikTok is also guilty of influencing what you see but they stay out of your way as much as possible. They bombard you with what you came for every minute you spend there. TikTok is massively popular and addictive because of that. US platforms could do that too, but their customers are advertisers - people who want nothing except to take your attention away from what you want to see.
You can go to twitter and get offended by bots, or you can go on TikTok and get bathed in dopamine. There's room enough on the internet for both experiences, but don't be surprised when people feel frustrated and annoyed by one of them and not the other.
Well yeah, you've got to stay angry at the right people at all times. As long as 49.9% of us are preoccupied with hating the other 49.9% of us, it's nice and cheap for the 0.02% to run the show. That's the social media value prop: division as a service.
> Anyway, you're wrong. TikTok pushes videos you want to see, while US app algos push content you are most likely to engage with.
No. US apps push creators, TikTok pushes content.
On TikTok, its the content that goes viral. Some nobody with 700 followers can have a video explode. That is exceedingly rare on YouTube. Its usually the channels with 1 million subscribers.
Advertisers love that, and so do platform owners. Its much easier to control and squeeze a few creators rather than a big diffuse group.
> Advertisers love that, and so do platform owners. Its much easier to control and squeeze a few creators rather than a big diffuse group.
There’s most likely some middle ground for this. Too diffuse and it makes direct sponsorship difficult meaning the platform has more control over advertisements as they need to be algorithmically placed. Too centralized and you’re back in the land of cable where your advertising is less effective because of the broad cross section of the audience watching.
As far as “squeezing” creators goes… no, more centralized means higher demand which means higher pay and more control in creators hands.
> As far as “squeezing” creators goes… no, more centralized means higher demand which means higher pay and more control in creators hands.
Not at all. You forget that platform owners can suspend or ban someone without much recourse. "That's a nice channel you got there.. be a shame if something happened to it" and all that.
I must be immune to their algorithm because, even with an account, multiple searches and 1 hour of use, I didn't get a single video I wanted to watch. YouTube works way better for me. I might simply not like short form content.
I didn't get an interesting video even with the search function. I just couldn't find compelling content on tiktok for my interests.
Also, 1 hour without an interesting video is something I've never experienced on YouTube, even on new accounts (I create an account for new TVs). I think I'm simply not their target audience.
I couldn't even understand the point of 10-20% of the videos, like "What is this video trying to show here? A random person saying the word banana in a loop"
I mean 50% of TikTok is brain rot and 50% is normies doing normie stuff. If you're someone who saw the old Internet I doubt you'd give a shit about TikTok. It feels like a betrayal of what the Internet could have been
I mean the pre-smartphone era when there was a barrier to entry on the internet. It was a nice place. Then everything turned to shit, as it always does, when the ignorant masses caused a context collapse by diluting the previous subcultures into a homogeneous blob that is eager to get spied upon and to be served advertisements non-stop.
The point of no return was precisely the introduction of the smartphone because it lowered the amount of energy necessary for someone who doesn't care about technology to interact meaningfully with it. To this day most people don't own a personal computer unless they need it for work. But almost everyone has a phone.
People who use tiktok tell me it gets better when you have an account. All I know is whenever someone sends me a video it always tries to show the most emotionally charged stuff (e.g. war footage) right after. Now it's even showing pictures of Trump when you pause the video
i think they could if they started with "show users content they like" instead of "keep users staring at the app for as long as possible". both result in more engagement and more ad dollars, but optimizing for the latter becomes a race to the bottom with increasingly extreme, polarizing, emotion-inducing content.
the blatant algorithm manipulation around elections and politics is just the icing on the cake. sure, china is probably doing this too, but they're either being more subtle or playing a longer game. meta et al may have come out ahead for a few quarters but what's that worth if user count is declining long term?
It's pretty simple, the Tiktok algorithm recognizes the value of long term user satisfaction of it's users, and all the American Tech Algorithms are blind to anything long term, and plus the "user" is not the customer in the American tech product, the advertiser is, American social media is a product for the advertisers and not for the users. This is clear as day.
From time to time it will show you content adjacent to what you watch. After three or so viewings your feed will add the content you watched/engaged with regularly.
Press "not interested" twice (or sometimes once), and content will disappear.
American social media effectively ignores any input from users.
so much this. tiktok can easily be trained to provide you pretty much exact content you desire from it while no other social media platform (is willing) can do the same
> no other social media platform (is willing) can do the same
Are we sure? My experience with Instagram is the opposite, whenever you linger a bit on a reel you'll start to get more of that content
It's a pain because you have to constantly press "not interested" to curate your feed, I'd pretty much prefer to have content feeds, so that I can watch anything without having it go in my (im)permanent record
I don't use any of these services, but it's interesting seeing your comment and some of the replies.
Just this past week I met a friend who uses TikTok and he said the same: Really good algorithm. He said when he watches "intelligent" stuff in it, the recommendations tend to be as "intelligent" or even more so. Whereas his experience with Instagram was that it quickly starts suggesting brain dead content.
I have no issue with Tiktok remaining a platform in America, but as someone who has used tiktok all throughout high school until today as a senior in college, I think it is objectively bad to use this app. Some of my friends that also use tiktok can ONLY watch videos in 2x speed. They don't even have the attention span to watch a video at normal speed any more, which they proudly admit for some reason.
I also do think there is a little bit of passive censorship about controversial topics. For example, if you lookup "free tibet" or "free hong kong", the posts there have at most 2-5k likes, and it seems like these posts never really "hit" the algorithm's sweet spot and get famous. Sure, this is entirely anecdotal, but I do find it a bit odd how the algorithm chooses what to show and what to hide.
Once again as someone who uses it, I think tiktok and its algorithm are definitely crippling the youth of America. At the same time, it doesn't sound right to ban it.
The "western tech executive" is a propaganda mouth piece. This entire situation is reeks of political manipulation and dishonest voices from every single media pundit. Tiktok is merely a break in the wall to wall American Media Nonsense, and the proponents of the continual American Gaslight don't like a break in their manipulations.
Ditching a native framework for something JS-powered and running everything thru a cloud server sounds like technical decisions willfully made by engineering leaders.
VSC is the least bad Electron app I’ve ever used, but (heavily subjective) it pales in comparison to Neovim + Tmux. It’s not even close.
Related: I was looking at WinRAR’s site last week after reminiscing about it with coworkers, and found that a. They haven’t really updated their UI since I last used it a decade+ ago b. The download is still 4 MB. THAT is why native is superior – if you know what you’re doing, you can get incredible performance with absurdly low filesizes.
Because I need the space for videos and games and that's why I have large storage. Not for tiny applications wasting 300 MBs because someone thought that an electron app would reduce engineering cost.
Aside from the fact that it shouldn’t take hundreds of MB to launch the simplest of apps, and that it’s incredibly wasteful on its face? Memory. Where do you think those bytes end up when you launch it?
I use VSC because some mature plugins only exist for VSCode, like Rust, and Microsoft pushing it for stuff like PowerShell, killing their ISE IDE.
And it only performs that well, because all critical code is written in a mix of C++ and Rust running in external processes, and they have ported text rendering into WebGL.
It's an interesting example, because the fact that it is js makes it trivial for most developers to make modifications to it by opening the Chrome DevTools. Even if you're not a js dev, you probably occasionally write some js.
I'm arguing that it's successful because any of its users can trivially hack something on top of it and distribute it, including things the original devs never intended or think is a good idea. In that way, its success mirrors Excel.
> From a documentation, examples, accessibility, tooling, and number of people you can get help from, JS wins
Maybe as a general purpose language, but for this specific comparison (extending editors). Elisp and Emacs wins. Vimscript is not the best plugin language, but the interation process is way better than VSC.
> Probably egged on by people telling them they had a much larger hiring pool if they went with JS (which is almost certainly true).
There are many VC funded companies here. How many of you felt pressured to pick a hireable language like JS/Python because if not you couldn't deploy your investor's capital? Like, if you had presented a plan of "I'm going to need 4 graybeards that know Haskell", you'd get denied for not thinking big enough.
Interestingly, Mercury [0] is VC-backed, and their backend is entirely Haskell. In an interview [1], their CTO mentions that it’s actually quite easy to hire for Haskell, as the demand is much lower than the supply, and, as he slyly puts it, “interest in Haskell acts as a decent proxy for baseline developer quality.”
So while the pool is larger for JS/TS and Python, that may not always be beneficial.
1Pass is a perfect example for my argument – v7 was amazing, and native. v8 has weird bugs, like refusing to open despite saying it’s working fine. As to memory, while I forget the consumption of v7, v8 looks to consume somewhere around 130 MB for the app, plus that again for a renderer helper or two, and then the browser extensions.
100+ MB for what is essentially an encrypted local DB with cloud sync and a GUI seems absurd.
That's an example of something that can be done well or done poorly.
AirBnB, UberEats and Facebook are all built with React Native and they have excellent performance.
Using a JS framework for your UI doesn't inherently mean it will suck. It can be done well.
If you expect it to be half as much work, you'll be disappointed.
If you expect it to be a tradeoff that makes some things easier and some things harder, and you're willing to invest in making it excellent, then it can be a very reasonable choice.
At least on iOS, the Facebook app is not and has never been built with RN. Some features were, but none of the core ones like News Feed. The biggest example of an RN-based feature is Marketplace.
We switched from Datadog to Grafana (do not recommend unless they got you over a barrel on pricing and you need to escape) and one nice thing Grafana gives you is the ability to self-host for local development so you can even run integration tests against your observability... an edge case need but if you need it you're glad it has it.
I’ve historically been a pretty big fan of Grafana, I’ve advocated for the cloud solution at more than one company.
But it seems like business development has utterly hijacked the experience.
The flow you want out of the box is Prometheus, Loki, and Tempo with one button that drops you the config for grafana-agent (now alloy which seems good technically but brings a whole new config language with some truly insane discoverability problems) that makes graphs on screens, you build up from there.
But these days everything is some complicated co-sell, up-sell, click farming hedge maze through 90 kinds of cloud vendor rip-off half baked thing.
Graphs, logs, traces out of the box. Put all the works with Snowflake shit behind an icon. A small one.
Not OP, but looked at doing grafana self hosted for similar reasons. The tooling is too spread out across different installables, the common golden signals and other monitoring metrics have a high learning curve / cliff more like it, and there's not good enough documentation to cover the user from "I want to do a synthetics test on a service to see if it is alive and show the results of that test in a graph"...a journey like that involves 4+ different tools.
DD was just easier to use for everybody, has lots of useful baked-in things we liked to use (apdex scores), and was intuitive enough that non-devs could design their own dashboards. We also found it way easier to collect metrics, traces, and other things.
FWIW none of these things are insurmountable and I suspect you'll eventually reach parity. Datadog lost our business for two reasons 1) Lack of billing transparency and 2) an incompetent account rep who managed to piss off our finance department while also embarrassing our CTO. And to be clear, while we aren't a "whale" our spend was over $6M/yr with Datadog - and the CTO along with the rest of eng leadership were all huge DD fanboys and yet they still managed to burn that bridge to the point where we'll never go back.
The deal still has me scratching my head. They tossed out the brand name and logo. Elon already had the X name and domain. For much less than $44B I feel like you could clone Twitter and come up with a strategy to acquire users. Hell, for $1 billion you could probably pay a good number of influencers to move to your platform. $44 billion is an absolute fuckton of money to kill Twitter and move those people to your pet project.
> $44 billion is an absolute fuckton of money to kill Twitter and move those people to your pet project.
What's the point of having Fuck You Money if you can't say "Fuck You?" Your value assessment isn't taking into account the value of destroying old Twitter, of removing a major bullhorn in the information environment away from people that Musk probably considers adversaries at best, and malevolent actors at worst. Simply standing up his own competing platform would not have accomplished this.
It's weird that everyone and their mum "could" clone twitter easily. And yet the only products of note that's more than dismissive hackernews comments/slideware with something in production at similar scale was meta with threads and that's still inferior in terms of search and discoverability and scaffolded with the guts of Instagram and bluesky which has the advantage of being founded by Jack Dorsey and has been around for years now. For all the big talk musk et al have I'm not sure they could actually have built a clone. You can pay influencers but if we look at how dominant tiktok is/was making buzz and content is more of an outcome of a bunch of incentivisations and the kind of non technical community management stuff people dismiss as marketing than throwing money at all the big names you can find.
he bought presidency with it so it is not that bad of an investment. it is short-sighted to look at “twitter” value since he bought it, one should look his net worth since he bought it. even if closes it down tomorrow it will be money well spent
not bias but facts :) he took over twitter, made it into right-wing propaganda machine, changed algorithm during the election campaign ... it is short-sighted to not take into account that he basically bought dissemination of public information and single-handedly swayed the election. if you think he bought twitter to make money off it I have some crypto to sell to you :)
That's less an issue of a lack of manpower and more the result of one person up top pushing everyone away. A less annoying owner could lay off 70% of the staff without gutting revenue.
Though Elon Musk's purchase of the company was entirely in his own self-interest, and unfortunately, it's serving him well so far. Buying a company for a fraction of his personal wealth and leveraging that into having the president as his personal servant is one of the greatest investments in American history.
I feel like Twitter is a bit of a Rorschach test. If you want to “keep the lights on” you can get away with mass layoffs, but your revenues and product development sure do suffer.
I think you’re going to see a lot of tech companies adopt a less acute version of this strategy unfortunately
It's another form taken by enshittification to squeeze more money from the business. Employees are the largest cost center. If a business can get away with less innovation then they will get rid of and hire less staff to maintain their margin.
Realistically, this is a delicate balancing act. As others have noted, the Sales and PR game then (and always) drives the new and existing monies that come in.
But more seriously, DeepSeek is a massive boon for AI consumers. It's price/performance cannot be beat, and the model is open source so if you're inclined to run and train your own you now have access to a world-class model and don't have to settle for LLaMA.
reply