I have no specific insider knowledge, but a decade+ ago they bought a company called the Echo Nest that was developing some of the best audio signal analysis algorithms around, I assume much of that influenced their recommender system.
Nowadays, they have a quite busy research department so I would imagine that recommendation is quite fancy indeed: https://research.atspotify.com
> This means there might be something universal about octave and fifth
There is! At least for the kinds of instruments that are conventionally used in Western music.
The harmonic series arises naturally from the physical properties of a string or wind instrument (e.g. violins, guitars, pianos, flutes, brass, organs, etc).
As a very rough description of the physical phenomena, the tones we hear arise from a full spectrum, atonal excitation (like a pluck or a reed flapping) bouncing back and forth along the length of string or tube, which is basically a one-dimensional "waveguide".
Frequencies that are aligned with the harmonic series naturally reinforce themselves, in the same way that putting energy at the top of the arc of a playground swing has more of an effect than in the middle.
Notably, musical instruments that are not strings or tubes, or more general sound-producing bodies, have more complicated patterns of sound waves dispersing through them, and don't typically follow the harmonic series.
Pitched percussion, drum heads, or bells have more complicated harmonic spectra than the standard harmonic series (as they are generally thought of as 2D or 3D waveguides where cancellation/reinforcement patterns are less straightforward), as do less musically conventional sounds like knocking two rocks together or striking an arbitrary surface.
From Edward R Murrow famous speech: This instrument can teach, it can illuminate; yes, and even it can inspire. But it can do so only to the extent that humans are determined to use it to those ends. Otherwise, it's nothing but wires and lights in a box. There is a great and perhaps decisive battle to be fought against ignorance, intolerance and indifference. This weapon of television could be useful.
The novel element is combining the passive medium with infinite content. In my circles, sitting slack-jawed in front of the TV for hours was something that only those with little mental energy or drive did[1]. By contrast, probably 75% have some non-trivial degree of slack-jawed passive social media consumption, even more so since IG and Tiktok.
To wit, I think what's interesting about this Era of media relative to the TV Era is the vanishing proportion of the population that's able to escape the habit.
[1] Not a value judgment: my sister and her husband consume massive amounts of TV but they're also both early-career doctors. I would be braindead at the end of the day too.
Idk how people do this, its just so boring. I tried tiktok and the first 200-300 scrolls were interesting, but then its just people regurgitating the same comedy/meme. Sure you can find a niche subject you're into like cooking, but most topics do get kind of dry after a while. I do think I'm in a minority though and know quite a few who spend hours a day on tiktok/insta.
It's enjoyable for 20 minutes a day, especially when waiting on something.
The key things are:
1) time offline is on your side. you can saturate yourself with current trends that interest you pretty quickly. You need to allow actual real world time to pass for those trends to update.
2) scroll with purpose and intent. aggressively dismiss things that don't immediately get your attention from any unknown source. (Helps the algorithm actually cater to your interests)
3) tell the algorithm when you don't like something. There's usually a "don't show me content like this" option somewhere. I felt dramatic about it at first, but it's the only tool you have to keep the algorithm from incorrectly assuming you enjoyed the content when you did watch the entire thing (out of sheer curiosity / hope / general inaction).
I noticed I now get a lot of low profile things in my feed that are actually pretty cool and fit the medium nicely. Lots of trade work stuff, before / afters, machines doing stuff, stand up comedy bits, etc. Those personalized things do not have room to flourish if I am giving too many things a chance.
One idea I had was being able to share your curated algorithm to others. ie. Your instagram explore page, or your specific tiktok recommends. People could subscribe to x person's recommends and see what they see.
In this spirit, I think Tiktok would actually be a tremendously good matchmaker for finding either friends or romantic partners. I think a lot of their recommendation AI actually figures out what you might like ahead of actually showing it to you, by trying to sort you into a cohort of people with very similar tastes. Which is why as a new user its sometimes scary how Tiktok can almost predict what you might like. E.g. people into cars, 30-40 yrs old, rural probably also like DIY.
That's a good point. A simple k nearest neighbors search of users would likely turn up very similar people to you even if you don't explicitly include any demographics.
What a coincidence, me too. I think there should be a marketplace for them. Ability to lock changes, or go back to an earlier version of your algorithm. Power user tools for curating it better.
That's the next influencer game imo. Having people want to see your feed(s)
No, that's not what he's saying. Broadcast TV & radio is human-generated, audio and video, and not immersive. It's several steps behind the trend he's predicting. And they are also not equivalent in terms of getting user engagement and attention, which is why those industries have shrunk so much.
I wonder if there are individual YouTube compilations of similar videos (think "Funny Cats" montages) that have generated more profit than Funniest Home Videos shows.
are you sure? it sounds like the exact same playlist on every station with the same owner in the same market with such little care about the content in the playlist but only based on an algorithm.
You can't ignore matters of degree. It does sound similar, but just by reducing the content chunk length to ~20 secs, they've dramatically improved the algorithmic manipulation and ad insertion (from their point of view), making it all far more effective.
In the old model people left the tv droning in the background, in the new model, people are riveted by their phones (with 2nd and 3rd screens (and radio (and billboards)) droning in the background).
IMO the differ here is less about AI vs human curated, but curated/generated for a very broad audience vs one specific consumer, ideally live as their mood or interests change. We aren't there yet with the highly dynamic mood and interest changes (e.g. interest fading after 5 Dr pimple popper videos, let's throw in some wingsuit stuff), but it's on the same trajectory. Address broad audience -> address smaller, niche audiences -> address individuals -> address individuals in the moment
Yes, we already have "slow AI". It's the corporate paperclip machine. It's has a strict value function; make money to shareholders. And it will do what ever it takes to make more of it.
All these ppl scaring us with AGI are either distracting us from the clear and present dangers of slow AI so they can keep profiting from it , or are just duped by silly technooptimism.
Of course it will, but it's something happening now by slow AI. We don't need any research on some hypothetical.
We don't need to frame this as a technological problem, it's clearly a social one, of which we suffer the consequences today.
No, but it was undoubtedly produced with the intention to get the most people hooked as possible. Traditional media was just less effective at hooking people because of the limited number of distribution channels and the cost of producing content.
Copyleft / GPL licenses require copyright law to have any threat of enforcement.
Otherwise anyone could just ignore the terms of a copyleft license and use open source under any terms they wish.
This can cut both ways.
Even after the recent massive dip, Snap's valuation is an order of magnitude greater than Facebook's notorious buyout offer.
Twitter's Fleets feature lasted less than a year, and Im not sure that their Clubhouse-"inspired" feature has gained much traction.
Its not clear that even Facebook will be able to disrupt Tiktok's momentum.
Slack exists and is successful despite consolidated offerings from the tech giants (Google famously has a double-digit number of chat apps).
Generally, startups have drastically less internal momentum, bureaucracy, tech debt, or politics to contend with and are much better positioned to push fresh ideas, and be responsive to customer needs rather than fitting a larger corporate narrative.
But its fair to say that if a startup idea is really just a feature idea (even a really good, well-executed feature idea, like Calendly) and doesn't scale its ambition beyond that, the best outcome to hope for is a buyout, and at worst being built internally by a giant.
On the other hand, its never been easier to start a software business thanks to incentives from the tech giants.
Google Cloud is basically free to start up and gain traction with.
Granted, there may be downsides to this, but not having to think too hard about infra opens up a lot of opportunities at the same time.
Yeah, honestly I think the bigger negative impact on startups is the incredible profitability of these targeted advertising monopolies that has pushed market salaries so high that upper quartile programmers with 3-10 years experience inherently see their value as being $500k-$1M in total comp without any real notion of what kind of business structures can generate the kind of cash flow to justify those salaries beyond just burning VC capital in an oversaturated bull market.
Even though I personally have benefitted from the upward pressure to software engineering comp, as consumers I believe we would have a way better ecosystem of tech products if engineers salaries weren't tied to VC economics.
But engineer compensation has been larger pushed up by the FAANG companies which are highly profitable. There’s some VC fueled companies competing with them like Uber but on the whole engineer compensation went up not due to VC money but because big tech was able to generate millions in profits from engineers. Every FAANG company has a P/E ratio under 30 so these big engineer packages are from strong business fundamentals and profits, not VC economics.
Startups have the ability to compete in ways like better equity deals but by and large they still highly prioritize their investor concerns over hiring concerns as reflected in things like liquidation preferences in ISO and stingy equity grants. Most of the founders would rather ride their startup to the grave then re-evaluate the “standard” terms of their equity compensation approaches which each year have become more favorable for VCs and less favorable for employees.
Even though big tech has many problems and things I disagree with, at the end of the day they respect engineers by paying them their value while startup founders prefer to make engineers second class citizens then bemoan their hiring difficulties.
> But engineer compensation has been larger pushed up by the FAANG companies which are highly profitable.
If you go back and re-read this is exactly my point. Google, Facebook, Amazon have structural advantages that make them more profitable than the majority of useful services ever could be, much of which is based on the unregulated and morally questionable use of massive amounts of user behavior data. This sucks the oxygen out of the room for startups who could provide an honest product for an honest price, but might not be able to get the economies of scale to pay $500k per senior engineer. In other skilled labor professions $100k-$200k is considered well compensated, and that creates a larger sweet spot which enabled more diversity of services. With software + internet the potential for near-zero-marginal-cost global scaling push everything towards a winner-take all mentality, and so small shops providing diverse and higher quality niche products get squeezed for talent.
Again, I have benefitted personally from this dynamic, but I'm also old enough to be saddened at the unfulfilled promise of lower software build and distribution costs that we envisioned at the blossoming of the web and the release of Microsoft's iron grip on software profitability in the late 90s.
GP said "senior", which can be interpreted either literally (as in the title "senior software engineer") or more broadly as software engineers with elevated seniority.
Some data for "senior software engineer" in the US indicate:
"The middle 57% of Senior Software Engineers makes between $117,200 and $203,000, with the top 86% making $375,000"
"The middle 57% of Staff Software Engineers makes between $107,389 and $262,186, with the top 86% making $572,331."
Since levels/titles are truly comparable across companies (nor across time), it's very hard to get a true sense of what those numbers even mean. But I think it's fair to say that you shouldn't expect a median nation-wide "software engineer" salary to match the expectations of the "right person you really want to hire for your position but you can't find".
That surprising to me (perhaps due to bay area bubble), how do I hire people at that rate? Is it entirely about the location? Are the skill set expectations different?
You can do a lot better than that internationally. When I was in the startup founder world my peers were paying their developers about $30K/year, and getting decent (but not great) talent in Thailand, Vietnam, India, Eastern Europe, etc.
The downsides are that all your employees are a.) remote b.) not great English speakers and c.) generally mediocre developers, and this all has negative impact on developer velocity and the caliber of features you can ship. If you're chasing a niche this is often a great trade-off, because it can make your business profitable even at revenue levels that can't support a big-company product. If you're chasing the next big thing this is usually a stupid idea, because somebody's going to raise $50M, hire top-quality ex-FANGs at half a million each, put them in a room together, and win the market before you can ship your first couple features.
I think you will find a lot of these people doing embedded, desktop, or enterprise intranet stuff rather than slick web/mobile products or massively scalable backend APIs. Using dated and uncool stacks. Dated and uncool practices. Windows centric. Hiring more about years of experience with particular framework versions than LeetCode or system design interviews.
Two perrenial favorite pieces on this website include Dark Matter Developers [0] and We Only Hire The Trendiest [1].
Amazon was break even for a long time on the retail segment, and nobody argues that AWS is profitable due to "unregulated and morally questionable use of massive amounts of user behavior data". Amazon's profits come from the fact that so many companies are willing to pay drastically more for AWS than for standard colocation services, even if they're mostly just using compute, disk, network and databases.
"the unfulfilled promise of lower software build and distribution costs that we envisioned at the blossoming of the web and the release of Microsoft's iron grip on software profitability in the late 90s."
Is it unfulfilled? Many of us don't use Windows these days, which is a nice improvement for those who choose other platforms, and the web/cloud has driven build and distribution costs much lower. When was the last time a software startup had to sink money into the logistics of CD burning and boxed retail deals?
I feel like you've neglected that there was a massive lawsuit against big tech where Google, Adobe, Apple, Paypal, and others actively colluded with each other to suppress wages:
It really wasn't until after this lawsuit, along with Facebook actively making better offers, that tech salaries started to skyrocket. I'm the sure the bull market for 8 years didn't hurt either.
> It really wasn't until after this lawsuit, along with Facebook actively making better offers, that tech salaries started to skyrocket.
No, the lawsuit had nothing to do with it. The market exploded because Facebook refused to join the cartel. The lawsuit fined the companies involved risible amounts and no one involved got fired or jail. Irrelevant.
Management and executives are DEFINITELY underpaying us relative to what we bring to the table for them. This is true of every industry going on decades of increasing productivity and stagnating pay. What's new here besides greater awareness and labor churn? Will there be an actual change going forward? Of course the demographics favor workers now, but the laws and politics often don't. Immigration and outsourcing and automation can change that power dynamic very quickly. What're you going to do? Form a union? Start a company? Vote? Quit and work elsewhere? Interesting times we live in.
Maybe that is one of the reason headcounts have ballooned and complexity has increased so much -- an attempt to minimize individual impact. VCs and MBAs famously refer to this as the bus principle: what would happen if the key employee or executive got hit by a bus today? How would the company fare? The metaphor is quite telling in its priors, assumptions, and priorities.
All of my friends and aquaintinces in nursing, tax, retail, and trucking have a few people who are trying to break into programming because of the benefits and pay. Outsourcing, scope creep, and automation exist and are expanding in my industry as well. How long until this drags down programmers as well? With the potential end of this bull marker will programming compensation stop being the highest paid field? I do wonder and doubt.
I suspect that software engineers, as broad as that term encompasses, will eventually see the effects of pay and work which up to now they've been largely insulated from. More work being handed off to you for a minimal or nonexistent pay bumb. Or slower wage growth than we've been accustomed to seeing. That's my prediction anyways.
My other point is that, based upon the revenue and profit generated per employee for many of these mega corporations, software engineers should be paid more, but I think lately we have astronomical pay relative to other industries in the 3-10 year experience mark for most of the roles in the labor market.
This is one of the reasons why I'm skeptical that Apple opening up iOS to sideloading will lead to Meta, Google, or Amazon rival third party iOS app stores hosting apps with more invasive tracking and lack of privacy protections.
These tech giants are great at xeroxing products and features, not so good at selling them in such a way that users can be convinced to switch. The existence of so many cookie-cutter clones that never get anywhere, from entire cloud gaming platforms, to mobile payment methods, to slapping Snapchat-style stories into an app for no reason, shows that it's easy to envision and build, hard to get actual users.
Stripe's session had something unsettling in their presentation. They boasted about how well they're good at dealing with bureaucracy and byzantine laws in fintech so we, as their customers, don't need to worry about them. In a way, they're admitting that the moat is built by the bureaucrat class and they're the beneficiaries of being inside the perimeter. They deserve it but we should sometimes take two steps back and confer to ourselves – "Wait...why have we built a giant moat in the first place?". Large corporations have perverse incentives to increase the moat by working with the only coercive entity we have - the Government.
Future has in for us more state corporatism than I'd like. It is sort of dystopian, but that word has been diluted these days.
In the specific case of Stripe, Ill gladly take a larger bureaucratic moat in exchange for PCI compliance.
The cryptocurrency world is rediscovering this tradeoff all over again.
Stripe is more like the bridge over the moat, than the people inside the perimeter defending it. Unless you can point to them lobbying for more rules only they could meet?
Finance laws are indeed extremely complicated and many are very questionable, but I don't know how much to blame government for that. Money is at the centre of so many conflicts, crime and disasters that a lot of those rules are the scar tissues from previous wounds. You can argue to get rid of them and cryptocurrency has provided a useful sandbox for people to try that, but a lot of the problems those rules were designed to create came back (e.g. AML law proved quite effective at stopping ransomware, hence why it all gravitated to cryptocurrencies). I worked on Bitcoin for a long time and the experience definitely made my view of financial law a lot more muted. The downsides of the regulations are only occasionally admitted to by the authorities and they aren't that good at building convincing arguments for it (because they don't have to be). But it's not like the laws exist purely thanks to Dastardly Dick.
In the absence of government, those same large private entities would simply build similar bureaucracies in their stead, and we would have even less recourse against them. Vote with your dollar or vote with your vote, the dilemma exists in both real and imagined societies.
Moats should be built through technological advancement, not appeasing regulations. I think you misread, I am not advocating absence of the government. But we should question decades of regulation that has piled up in a way that provides giant moats that do not do anything to build a better product (no strike for Stripe, speaking generally) but to satiate those regulations and prevent others from building anything.
To be fair, calling Snap, TikTok or Slack a startup is pushing the definition of a "start-up" a bit. This article _was_ written in 2018 when these were all smaller, but even back then these weren't too small.
Snapchat has been around since 2011, Slack has been around since 2013, and Musical.ly has been around since 2014. By 2018 they all had significant user bases; Musical.ly had been in acquisition talks with Facebook in 2016 [1], and was acquired for $1 billion by ByteDance in 2017 [5]. Snapchat had famously been offered $3 billion in 2013 by Facebook [2]. Slack was big enough to buy HipChat and Stride from Atlassian [3][4].
By 2018, I'd say these companies were already in a good position to compete with tech giants; but I wonder about the smaller startups. Say companies that have been founded in the past 5 years; have they had an unfairly hard time? Do they see any future where they can actually compete with a tech giant, and not just shoot for a buy-out? I don't have data on this, but I'd be curious to know how the trends have been shifting here.
> Im not sure that their Clubhouse-"inspired" feature has gained much traction
I have no idea what the numbers look like, but I'm aware of people using Twitter Spaces (even though I no longer use Twitter!), while I'm not aware of anyone using Clubhouse.
I'm currently using Google's chat and spaces at work. It's passable. They have a "threaded" view that I was really excited about at first but you can't collapse the threads so people just reply to whatever the first thread is that they see.
If they could get the UI right on the threaded bits and add a standalone desktop app, they'd be very close to "good enough" for most places that need some central chat option.
I haven't tried out their web hooks yet, but it's on my experiment list. The rest of the Google Workspace offering just makes life so much easier though.
> Even after the recent massive dip, Snap's valuation is an order of magnitude greater than Facebook's notorious buyout offer
you need to do the math for both companies.
If Snap took the offer and held the FB stock, it would be at the exact same valuation that SNAP is currently at
hmm, please explain.
is the ceo of snap still holding SNAP shares? it is liquid and the ability to sell off or keep is true whether he had FB shares or SNAP shares
TikTok required billions of dollars on ad spend to be able to compete. Definitely shows you can break into being a top-tier social product, but it maybe requires a lot of money to do in a short amount of time.
While I don't disagree with you at all, the current startup scene doesn't feel organic. So many of these "startups" are severely bloated and should have gone public ages ago.
A simple look at the employee headcount at most unicorns is startling. A startup that's years away from IPO shouldn't have 5,000+ employees. At that point, you lose much of the agility startups are supposed to have in the first place.
The constant inflow of private funding have allowed startups to acquire way too many bad habits and bureaucratic layers.
Even higher end brands like Samsung are pulling stuff like this.
Market forces demand constant growth, and in a saturated market this means some business unit manager will keep adding more invasive advertising as long as people keep buying the TVs.
I've had good luck with TVs from Sony, which would probably be considered higher end based on price bracket. They ship with near-stock Android TV, don't make a fuss about not being connected to the internet, and allow firmware updates via thumb drive. Have had a couple now that I use with an Apple TV, which is a pairing that works very well.
I wonder if Sony are releasing the Linux source code they run Android on and I wonder if their firmware updates are signed. If they do release source and updates are signed, you could probably install a regular Linux distro with Kodi.
Likewise, but the problem now with the built-in software is that Android TV is getting increasingly user-hostile as well. The major UI update they made which added an uneditable promotional section to the top of the home screen eg.
As a hiring manager I’ve worked with internal recruiters and a good one is invaluable.
- the transactional part of setting up interview loops, coordinating schedules, etc is extremely tedious for a field of 3-5 serious candidates for a given role. Internal recruiter takes all of this away from a hiring manager who has other duties.
- an internal recruiter can plow through LinkedIn to find solid candidates, identifying ones for the hiring manager to review and, if qualified, reach out to. Again, tedious work for a hiring manager with other responsibilities.
You are right, although there is a risk that internal recruitment teams become a bottleneck. Particularly when the requirement for new hires is more elastic than recruitment capacity. Something has to be given a lower priority.
That should be your job. Even the "tedious" parts are too important to be left to a non technical person. Time of day for scheduling interviews really does matter and the initial communication of scheduling this can tell you an enormous amount about the candidate.
Your job should also include lots of time on LinkedIn searching, as many HMs do today. Letting a recruiter do this for you will give you many more bad interviews (expensive) than you'd get otherwise.
But apparently "tedious work" is above a HM. Have fun being on conference calls all day, pretending that you and not the engineers are the ones who do the work. Worker self management needs to become more of a thing in this field in general if we can't trust HMs to take an active role in candidate selection. Companies that disvow managers and empower their employees produce far more excellent work (e.g. valve) even if they don't always do it on the schedule that execs want - so this would be a net positive for the industry.
Strong nope on pretty much all points, and its rude and presumptuous to tell me what my job should be or what other work it entails.
We’re going to just have to disagree here.
> So, what makes a job a "bullshit job"? Well, the defining criteria would be that they only exist to the benefit of their employer. They don't generate any value as far as the stakeholders of an employer is concerned: clients, customers, members, patrons, patients, visitors, other employees, etc.
Lets not forget the most significant stakeholder from many business's perspectives, investors.
Many of these "bullshit jobs" enable businesses to operate at the scale required for continual economic growth.
Except for very tiny businesses, execs are generally going to be too busy to answer the company phone line or greet guests.
Graeber sees these as inefficiencies, leaning on his experience as an academic lifer, but academia is a very different, if adjacent, market to the capital-driven industrial world.
He seems to miss his own point that so many of these jobs are in the service of ever-increasing economic expectations.
VCs aren't primarily looking for brilliant visionaries, even if its a popular narrative--the (high) risk/reward profile of a business venture is at least as important to them as the character of its founding team.
The thesis behind Moth Minds seems to be that there is a lot of interesting/valuable* work that could be done that doesn't fit any of the typical funding models (VC, small business loan, burn through personal savings, Patreon, tips, etc.).
Like Xerox PARC type projects or PhD research without the gatekeeping.
This kind of work doesn't exactly require a singular brilliant visionary, but it helps to provide a financial incentive to lure someone with a unique vision from a comfortable tech salary.
Way back before the digital age when software was still an insignificant component of an overall technology spectrum, this was so long ago that it was a completely different economic landscape too.
There was a prominent recognition in major companies that for positions where a visionary ability was needed, they offered a considerable amount of compensation intentionally as "temptation" to keep the moths drawn to the flame rather than spreading out in an unforseen direction to do their own thing.
Whether that was starting a new business or going to a place like Bell Labs where you would be more likely to pursue your own strongest interests, this was all balanced out based on the need by the biggest payers for the most visible and persuadable high-performers.
This was actually a limiting factor.
Now the remaining less visible & persuadable high-performers of the same caliber are way harder to come by precisely because of the visibility issue, even though there are many more of them.
This is a vast resource that could potentially be leveraged to overwhelming benefit by an alternative paying force that has been absent from the landscape for all practical purposes.
This is putting a lot of undue blame on product managers who only have so much autonomy in a business regime set up for perpetual "growth" that is necessarily quantifiable and ultimately aimed towards economic growth.
Constant growth is the underlying objective for any investment-driven software business (pretty much any of the businesses regularly talked about on this site, and certainly ones supported by this site's operators).
"Making a great product for the customer" is necessarily secondary to investors getting a return.
But its refreshing to hear of notable exceptions to this!
In my field of music tech, Ableton is one of these, for instance, having semi-famously refused significant outside investment to be able to focus on making a great product: https://www.billboard.com/pro/ableton-founder-gerhard-behles...
Nowadays, they have a quite busy research department so I would imagine that recommendation is quite fancy indeed: https://research.atspotify.com