From the announcement: "The company also announced that it’s divesting Bandcamp, an online audio distribution platform it acquired last year"
I don't care a ton about Epic, but I love Bandcamp. It's my primary mechanism these days for discovering and acquiring music.
I was puzzled when Epic bought Bandcamp, apparently they were too. They're being kicked over to Songtradr -- I hope that's a neutral to good move, but I don't really know. Anybody know anything that could shed light on that?
I can't see it as a good thing at all. They're effectively being thrown about to yet another operating model which will have a disruptive impact on them.
The new company Songtradr focuses on licensing, it's a b2b, while Bandcamp is a b2c. So it can be easy to see how they'd want to make Bandcamp more like themselves. Of course I'm being negative and cynical, apologies. I share the same fondness for Bandcamp.
I mean... Bandcamp is also licensing, all of works of art "copy sales" are just licensing agreements. B2C is just commonly a preset non-commercial license. (I believe that Bandcamp also allows you to sell a commercial license)
Let alone Bandcamp is a good source for smaller DJs to acquire tracks, which isn't a long shot for Songtradr.
There are some advantages to having b2b and b2c in the same house, especially when some of the b2c involves b2b licenses (it in theory makes it easier for artists to do remixes and cover songs if they can buy those licenses in the same place they sell their final product). Before Bandcamp a store I trusted was Loudr.fm and that was their big deal was that they were a b2b license shop for the rights to things like covering video game music and then were a trustworthy b2c consumer shop for the covers the musicians had an easier time licensing. Ironically as Bandcamp ascended most of their musicians needed/wanted Bandcamp presences and they shuttered their b2c shop and encouraged everyone to just move to Bandcamp, refocusing to just the b2b side of the business. The internet tells me that what remained of Loudr.fm was then eventually swallowed by Spotify in 2018, and that interestingly and likely but not necessarily a coincidence 2019 was when Songtradr got serious about expanding, and so it goes.
This is interesting! When I saw the Epic news the first worry I had was regarding Bandcamp. Even after I quit trying to get a job there I'd look at their jobs page all the time to get a sense of their health. It's been empty for a few months.
One of the suppositions people had about Epic buying Bandcamp was that maybe it would allow Epic to more easily license music for their games. From the press release[1] it seems like they did that. I have no idea how well that worked but here we are with Songtradr, "a music licensing platform and marketplace company," acquiring them.
On its own Bandcamp is (was?) a profitable enterprise.[2] I wish they'd go back to being on their own. That said, if Songtradr does nothing more but give artists more diverse ways to make money off their music I'm all for it.
That said multiple changes of ownership can play hell with continuity and morale. I'm worried about that.
If Songtradr doesn't screw it up, I'm fine. I too used to look at their hiring page a lot. I think that Bandcamp has so much untapped potential, but it's held back by a mediocre site and terrible interface if you own more than 100 albums.
They added playlists, I think, this year but I have > 1200 albums on the service (I buy a lot of full discographies and subscriptions when labels have them) and it's so painful to try to go back past the first 100 or so.
Music discovery is OK through the site - their Bandcamp Daily and other content is great - but it could be so much better. And it's screaming for better social features. Sigh.
I really hope Bandcamp stays neutral/independent, as that's the only place I legitimately buy music from (and a little bit of Amazon Prime), because they offer convenient .flac downloads immediately after purchase, no streaming, no "app" no garbage like that.
This is probably just continued post-COVID demand normalization. Basically all software and gaming got a massive bump that was obviously going to have an unsustainable component to it. But in many cases companies planned like it was totally sustainable and would continue. That’s why even super profitable companies like Microsoft and Epic are doing layoffs in 2023.
The other reason of course is that they went a bit crazy with the hiring sprees and accumulated some unproductive devs who are probably a net drag. That’s ultimately a healthy force in the economy as those people can do more in smaller firms with less in-house expertise (albeit at lower wages).
I wonder also how much of a driver new programmer productivity tools like copilot and ChatGPT are factoring into this. If your top 10% productivity workers just got a 30%+ productivity boost, that means you can easily shed the bottom 20% productive workers and still be ahead of the game.
> I wonder also how much of a driver new programmer productivity tools like copilot and ChatGPT are factoring into this. If your top 10% productivity workers just got a 30%+ productivity boost, that means you can easily shed the bottom 20% productive workers and still be ahead of the game.
I can confidently say that this factors in 0%. Layoffs were happening in the industry even before the LLM craze.
And a 30% overall productivity boost for a "top" developer is a massive overestimate. Your best developers aren't spending their days writing boilerplate unit tests.
Second this. From using LLM within my daily IC workflow, its not that developers are getting a productivity boost, its that our job became more enjoyable. I didn't even realize how much frustration builds up over time while relying on google search for things such as that one obscure api method i forgot the name of. I have to dig and dig...
Somehow using GPT in its place has brought a noticeable bump to my quality of life. Most of my dev colleagues share this sentiment.
Impact on the tech economy from this? Unknown.
Layoffs are rarely about productivity, especially at super profitable companies. Layoffs are primarily about short-term metrics to drive shareholder expectations. Where those metrics reflect productivity it is generally three or four degrees of separation from any actual productivity data. The thing those metrics are most good at showing are labor cost aggregates versus industry base rates.
The biggest reason for layoffs has always been, and likely will always be, the "C-Suite class" controlling expectations on labor costs (salaries/wages) in an industry. That's why they often happen in waves, many companies at once, it's a pressure release valve to keep the labor market in "control" propagated by the largest shareholders, especially those born and bred into the C-Suites. Those shareholders don't actually care how productive a company is, productivity is increasingly orthogonal to profit. They care how profitable a company is, and companies stay the most profitable when (among other things) the labor market is most effectively depressed (and layoffs are a useful depression tactic) and laborers aren't comfortable enough to fight for better wages.
I'd argue that the productivity boost is mich higher for the lower end developers than for the best ones. They are best at the simple tasks in the most verbose programming languages: Easier to think of them as higher quality auto-completion. The terser the language, and more innovative the problem is, the smalle the LLM help.
So if anyone gains from this, is the companies paying under-market, and therefore typically getting somewhat substandard devs to do boring things. There's major savings in productivity there, precisely because the training for those tasks is so good. If you had your highest performance developers spending a significant amount of time on that kind of busywork, you weren't getting that great of a performance in the first place.
Maybe this is the answer to a disconnect I've noticed. A number of devs where I work have started using ChatGPT to help with their work, but it hasn't resulted in any noticeable productivity or quality gains.
But the devs here are all very experienced senior-level engineers. Perhaps that's why we aren't seeing any gains?
> I'd argue that the productivity boost is mich higher for the lower end developers than for the best ones.
I'd say it depends on the task more than on dev level. If it's maintaining, improving, or refactoring of a large project LLM can't help much. When it comes to small utilities GPT-4 is a time saver. You can even write a small playable game with it in a matter of hours. Using any popular language. Simple web pages, JS are much easier. With ChatGPT-4 I've learned how to do certain things using ffmpeg. Without it would take hours of googling.
PS: talking about ChatGPT degradation with the time. I didn't see it on programming tasks. But it became noticeably faster.
This is a first order short term phenomenon. But also, the LLMs by themselves guide the developers into creating repetitive low quality code, and any bias into using a LLM will bias the developers into adopting verbose boilerplate-happy languages and frameworks.
Meh, it's nothing to do with that, it's just a layoff round.
Epic went from 2200 employees in 2020 to almost 9000 in 2023. They are now basically shedding 1 out of 6 hires they've done since 2020. That's not really about sustainability, it's just a way to drop the ones that didn't really work out.
>That's not really about sustainability, it's just a way to drop the ones that didn't really work out
These layoffs are just a way for CEOs to improve "efficiency" AKA increase margins so they get a nice fat bonus - COVID overhiring has been a convenient excuse and every company has jumped on it.
> Basically all software and gaming got a massive bump that was obviously going to have an unsustainable component to it.
Well, this sounds logical and all but isn't always.
For example, Google's revenue/employee exploded during covid but it didn't crater post-covid [1]. So even with a hiring spree, the company is on the same track as it was pre-covid. You might need to layoff employees to get that same covid high but that covid high is the unsustainable part not your current employee count.
So some companies may have overhired compared to what their reveneue growth without covid would've been but not all did. You still need to evaluate companies on a company by company basis.
Anecdotal experience but most of the people I've seen being fired in one of the big ones are senior people working in customer facing roles, where the expectation is that they will be replaced by AI and wishful thinking.
Newly hired people were more junior and cheaper than previous employees so it makes no sense to fire them.
No one in game and game engine development is using ChatGPT.
It doesn't even work, even if it was productive, as barely any problems the average gamedev faces on a day-to-day have been surfaced to the training data of an LLM.
Yeah, this exactly. The only person I know who is succeeding at "game dev" with ChatGPT is an older bloke I know who just began their baby steps on a pet project with Unity.
Sure, if you're still learning how to move position from x to x+1, ChatGPT can help you with those basics, but anything more than that and they start hitting me up on Discord looking for real answers.
Something as simple as how the serialization of fields in behaviour classes was too much for ChatGPT. They read what it hallucinated up, got confused, and were calling me up to explain minutes later. For those who don't know Unity, this is something you have to do daily when building new logic into game components and ChatGPT was clueless about it.
The limitations of ChatGPT as a programming aid are very very obvious and it has a long way to go before it's really useful to seasoned professionals.
ChatGPT 4 may be better but CoPilot is terrible at making suggestions on algorithm work. It'll slip a -(x) in to parameters where it's supposed to be positive. Use the sim but wrong variable names. Swap parameters around. It's can be a nightmare and cost you way more time debugging than it saves typing.
Specifically with type-ahead suggestions I believe with typing you are more engaged in thinking about what is correct. With a CoPilot suggestion the tendency is to just skim it to see if it look good. And that's probably part of the problem; the LLM is trained on what looks good!
LLM code generation tools are producing output equivalent to a new grad who doesn't know much how to write code that integrates well in a program. It churns out a lot of text but every bit of it needs to be questioned and reviewed. As such it's not exactly a productivity booster where correctness matters, it's actually a liability. But for churning out code that isn't critical, it's great.
Yeah, I think it depends on your specialty. If you're an enterprise developer, you save on some boilerplate for sure, and it's a nice alternative to Stack Overflow.
If you're more of a generalist though? I can see productivity gains of 25% maybe.
You think the top 10% of workers are getting, at a minimum, a 30% productivity boost? You think they're getting in an extra 1.5 days worth of work a week because they're using Copilot?
You mostly have to ignore these AI solution hopefuls.
They're living in a momentary dream that AI LLM tools will solve all our coding problems. They don't and reality hasn't hit them yet.
I don't know any real seasoned dev who is getting a 30% boost because they installed Co-Pilot. The tools poop out too many errors/fake-garbage to be useful to devs working on battle hardened codebases.
Another way of spinning this - if "fancy autocomplete" is giving you a 30% productivity boost then either you're not actually in the top 10%, or you're doing working on "top 10% problems".
Regarding the second para, the interesting part for me is: how do these companies determine which are the unproductive devs? My impression so far is that senior management has no way of measuring this, given how little data is provided to back up decisions to RTO.
Not so sure about the third para. Why would one company be more ahead of the game than any other, if indeed LLMs are providing this supposed boost?
Sorry, but mass layoffs aren't exactly the most rational ones.
There's a lot of "you have to cut X off your salary budget", pick whoever you want. Then it's up to the managers to pick, very subjectively, who they want to work with the most.
FFS - I managed to dodge 4 layoffs at a startup, including my whole department being cut wholesale. My managers always kept me, even though I wouldn't call myself the "best" or "the cheapest". All of it just boils down to one thing - I'm the one that whoever manages me can rely on. Just that... being reliable.
> how do these companies determine which are the unproductive devs?
The senior management delegates. At every moment your direct manager knows whom they are willing to part ways with if asked. Even if everyone on the team is great there is always some criteria (costs, impact on important project, institutional knowledge) that sets some people apart.
It's not possible to measure with 100% accuracy, but let's not kid ourselves every company and every division can easily pick out the bottom 10-15% of devs who have close to zero real impact. The actual number in fact is likely far greater than that.
Developers usually follow scrum frameworks, so productivity could easily be measured by overall activity, number of issues closed, merge requests completed, etc.
If ChatGPT should be able to generate any code, a function that outputs a set of prime numbers longs in Kotlin should be simple enough (unless you consider Kotlin, the 12th-20th most popular language according to various surveys, too obscure). One that has a good big-O runtime that is compact as possible. After a bit of futzing I get this:
fun sieve(n: Long) = (2L until n).fold((2L until n).toMutableList()){p,i->p.apply{for(j in i * 2 until n step i)this[(j-2).toInt()]=0}}.filter{it!= 0L}
Also if the 12th-20th most popular language (Kotlin) being too obscure is the reason, how will they fare currently except with the most popular languages?
This is about as simple as a function as you can get. 95% of my programming work is standups, sprint planning and grooming, looking at Swagger or other API backend documentation for information, looking through Outlook, Teams and Confluence for information about my story, looking for UI on Figma, contacting various teams for more information - then someone wants me to review their pull request and stop to do that - but notice they have some business logic wrong because they don't fully understand the domain. Yet even something simple like a function generating a list of prime numbers, ChatGPT struggles with.
Not that it can't get better, but it's no threat to most programmers I know in the short term (and if it becomes one, I'll shift over to writing Python scripts that call Pytorch, numpy etc. - I already do so in my spare time).
This is standard for economic conditions. When there's a lot of easy to come by money (loans) there's a lot of disposable income and excessive employment. As soon as the money supply tightens (or in this case year on year the money in the economy goes negative), businesses trim the fat and get conservative about only keeping people creating value/profits. Which is why unemployment figures lag in a collapsing economy.
Keep in mind companies planned like it was sustainable and hired a lot because from a business perspective, its much easier to overhire and lay off later than it is to be understaffed during a period of potential unprecedented growth. It doesn't fully rationalize many companies actions but its just an application of a regret minimization framework.
I personally don't believe its a massive function of Generative AI (although it could be), I think the major contributing factor is that the projections that companies made in 2020/2021 and the hiring they did to resource more projects to take advantage of those rosy projections ended up falling short of real expectations. i.e instead of needing to take advantage of a potential 20% yoy future growth in their target markets, they only really need to staff for a 10% yoy growth.
Also probably a bit of "monkey see monkey do" where a company that actually needs to layoff (meta) lays off a bunch of people, and other large companies that may not "need" to layoff (like Google) follow suit since now its "socially acceptable."
Fortnite's game as a service fatigue, for some reason they didn't jump on the Extraction Shooter hype and it cost them today IMO, Fortnite became successful because they jumped on a trend at the right time (from Survival Base Building to Battle Royal), they seem to have forgotten that
EPIC launcher is still bad today.. which doesn't make me want to have it open, and let alone to browse it.. which might hurt their numbers
Steam's launcher follows the same path it seems, slow and bloated CEF bullshit.. makes me want to use it less
>I wonder also how much of a driver new programmer productivity tools like copilot and ChatGPT are factoring into this. If your top 10% productivity workers just got a 30%+ productivity boost, that means you can easily shed the bottom 20% productive workers and still be ahead of the game.
I have been evangelizing GitHub Copilot (+ Copilot Chat) and ChatGPT Pro to my coworkers that include full-stack .Net/React devs, Python data science people, and embedded C devs.
Only the .Net/React devs have been sticking to both Copilot and ChatGPT Pro. The Python data science are using ChatGPT Pro but have dropped Copilot. The embedded C devs are using neither.
There's a very simple reason for it - the amount of slightly customized code is directly proportional to automation.
When working with embedded systems - I barely ever have "lightly" customized code. Meanwhile the closer you're to the frontend - the closer you get to having majority of your code being boilerplate.
I’d love to know how these top developers are identified. What’s this mythical productivity metric? If Epic has one that accurately models productivity they should 1) share individual scores with their employees and 2) stop wasting time on games and start selling it to every other corporation.
If they really did over-hire during the pandemic and fail to plan for the future that was either 1) intentional and they lied to recruits about the duration of employment or 2) incompetence on the part of leadership and recruiting. Have heads rolled in the responsible departments?
> I wonder also how much of a driver new programmer productivity tools like copilot and ChatGPT are factoring into this. If your top 10% productivity workers just got a 30%+ productivity boost, that means you can easily shed the bottom 20% productive workers and still be ahead of the game.
Probably 0%. What I heard, mostly people not performing (which is the real pandemic after COVID-19). If I'm not wrong as other companies, devs there are not "allowed" to use copilot and chatgpt fearing code plagiarism and ending up with some license violations..
> I wonder also how much of a driver new programmer productivity tools like copilot and ChatGPT are factoring into this. If your top 10% productivity workers just got a 30%+ productivity boost, that means you can easily shed the bottom 20% productive workers and still be ahead of the game.
Epic's game programmers are already incompetent to the point that Fortnite is basically falling apart at the seams and every update adds dozens of new bizarre bugs, I don't even want to imagine how much worse it would get if they started using AI to write code.
Likely not a reason for layoffs. If someone was fired due to llms it was probably due to the incompetence to think the chat bot is more than a data leaking search engine.
Gaming is a notoriously boom/bust field. Layoffs, failures and bankruptcies happen even in the best of times. Reading anything macro into a game developer layoff is really stretching things.
Why are people still reading tech layoff news and trying to attribute it to specific problems at a company? Remember that Google, Meta, Microsoft, Amazon, Salesforce, Shopify etc. laid off large chunks of their workforce while collectively making hundreds of billions of dollars in profit. Companies lay off people because they can, simple as that.
> Companies lay off people because they can, simple as that
Pretty cynical and unfair IMO. Meta layoffs were the first time the company did layoffs in its entire history. Layoffs are super unpopular with employees. It's not like they wanted to do it.
This argument is lazy. Like suddenly there's a huge wave of layoffs, and your explanation is "they could." That really offers no explanation for why it happened now vs. literally any other time.
In banking and consulting this is not unusual. Some let go the lowest perfoming 10% every year. They do it because it is beneficial to their case and there is no resistance.
Scaling quickly has risks and the current wave of layoffs is just a way to minimize risk. It is probably better this way than raising tons of cash until things are falling apart and you are forced into panic mode.
> Companies lay off people because they can, simple as that.
If some divisions or individual employees in a company are not making meaningful impact on the bottom line and are not long-term strategic for future profits, then it makes sense to let them go.
Just because these are hugely profitable companies, doesn't mean you should be spending money on things that will not lead to you generating additional funds. That is like keeping real-estate rented but not filling it with people -- it is wasted resources, whether it is humans or real-estate.
These companies over-hired or were sloppy in terms of their spending and that has changed in the new environment.
Economy is about to take a nose dive if you haven't been reading the tea leaves. This is both pre-emptive for the coming downturn and a correction for the covid over hiring. Buckle up.
Shouldn't there be games coming out fairly soon (next 1-2 years) that take advantage of Lumen/Nanite/Quixel/Metahuman? From my POV, the future looks bright for Unreal because those technologies give it a pretty crazy edge over its competitors, and that's not even taking into account the Unity fiasco.
Lumen is really nice as long as you never hope to have >60fps and don't want open worlds.
The architecture of it is just fundamentally not compatible with the majority of AAA games, especially open world or esports ones.
It's not that Epic are stupid it's that they're "optimising" in all directions and it sometimes causes issues. -- marketing advertises features that are not fit for the purpose they're being marketed for. Lumen is excellent for small games that aren't so CPU heavy or alternatively for non-realtime renders such as scene generation VFX.
Unreal itself is extremely bound to single-core CPU performance, so the idea of adding extra weight to the main render thread (which is basically the only thread) is unconscionable and was a deeply unwise decision.
Immortals of Aveum just shipped last month using UE5's advanced new features. The game tanked though, because it's a new IP and released at a time where we are inundated with high quality, high profile releases. 2023 has been kind of wild for AAA games.
The game just needed a more likeable protagonist and a less crowded release window, but I'm sure the takeaway for EA execs is that it shouldn't have been a new IP...
Pretty amazing. Although I don't know if I even need characters this real in my games. At some point it becomes too much like real life and kind of boring? I think I like the fantasy aspect of games too much and I like to escape.
If those tools can give a productivity edge for the sort of experiences you’d see in AAA games, then perhaps! Tech demos don’t always translate into compelling interactive art, which can be challenging!
I haven’t studied Unreal enough to say for sure, but I think those tools can be a boon if they enable a less technical developer to do things that normally took a lot more time.
I always advise people to at least have a backup spot to send people to even if they don't maintain their own website, and I take my own advice. I expected something like this, so I've been dropping every album into an item in my Ko-fi shop: https://ko-fi.com/s/7e9f22c63b
Ko-fi is 25GB for free and 200GB for paid accounts, so most things will fit. I'll probably cross-post it to Patreon's new shop feature in the near future just to be safe.
That's a bit disappointing. As a user I just want Bandcamp to be owned by someone who gets it and leaves it alone. As far as I could tell, Epic barely touched Bandcamp. I'm unfamiliar with Songtradr but they have more incentive to integrate the core product into their services as they're in the industry and a smaller company (more need to make the acquisition work for them.) I hope it goes well.
I'm somewhat hopeful about the synergies here. Giving musicians an easy path to monetize their catalog, as well as market their availability to be commissioned by brands, is a good thing... especially when artists are being dually hammered by Spotify's deci-cents-per-stream industry baseline and live music promoters' anticompetitive behavior.
That said, there are ways Songtradr could abuse their new position of trust. I hope they're smart enough not to do that, but I also hope that an archival project for free downloads on long-inactive Bandcamp accounts is underway...
And he'd blame Valve, too. EGS might be great for devs, but it's not good for users. It's not a big conspiracy against him. People just don't want to use the store! Whining in the face of obvious failure makes you look like a real loser.
You joke, but I could see Epic one day being like "if your EGS account only has free purchases and you've never actually paid for anything, your account/games will be removed in 30 days."
Great chart! But it's so weird to see companies go on a hiring spree AFTER their product was essentially done. I cannot name a single feature that I use that wasn't present in the 2020 version. Since then, their headcount has more than quadrupled. What are those people doing?
Readit News
I don't care a ton about Epic, but I love Bandcamp. It's my primary mechanism these days for discovering and acquiring music.
I was puzzled when Epic bought Bandcamp, apparently they were too. They're being kicked over to Songtradr -- I hope that's a neutral to good move, but I don't really know. Anybody know anything that could shed light on that?
The new company Songtradr focuses on licensing, it's a b2b, while Bandcamp is a b2c. So it can be easy to see how they'd want to make Bandcamp more like themselves. Of course I'm being negative and cynical, apologies. I share the same fondness for Bandcamp.
Let alone Bandcamp is a good source for smaller DJs to acquire tracks, which isn't a long shot for Songtradr.
Edit: to clarify meant Songtradr.
Deleted Comment
One of the suppositions people had about Epic buying Bandcamp was that maybe it would allow Epic to more easily license music for their games. From the press release[1] it seems like they did that. I have no idea how well that worked but here we are with Songtradr, "a music licensing platform and marketplace company," acquiring them.
On its own Bandcamp is (was?) a profitable enterprise.[2] I wish they'd go back to being on their own. That said, if Songtradr does nothing more but give artists more diverse ways to make money off their music I'm all for it.
That said multiple changes of ownership can play hell with continuity and morale. I'm worried about that.
[1] https://finance.yahoo.com/news/songtradr-acquires-bandcamp-e... [2] https://components.one/posts/bandcamp-the-chaos-bazaar
They added playlists, I think, this year but I have > 1200 albums on the service (I buy a lot of full discographies and subscriptions when labels have them) and it's so painful to try to go back past the first 100 or so.
Music discovery is OK through the site - their Bandcamp Daily and other content is great - but it could be so much better. And it's screaming for better social features. Sigh.
The other reason of course is that they went a bit crazy with the hiring sprees and accumulated some unproductive devs who are probably a net drag. That’s ultimately a healthy force in the economy as those people can do more in smaller firms with less in-house expertise (albeit at lower wages).
I wonder also how much of a driver new programmer productivity tools like copilot and ChatGPT are factoring into this. If your top 10% productivity workers just got a 30%+ productivity boost, that means you can easily shed the bottom 20% productive workers and still be ahead of the game.
I can confidently say that this factors in 0%. Layoffs were happening in the industry even before the LLM craze.
And a 30% overall productivity boost for a "top" developer is a massive overestimate. Your best developers aren't spending their days writing boilerplate unit tests.
The biggest reason for layoffs has always been, and likely will always be, the "C-Suite class" controlling expectations on labor costs (salaries/wages) in an industry. That's why they often happen in waves, many companies at once, it's a pressure release valve to keep the labor market in "control" propagated by the largest shareholders, especially those born and bred into the C-Suites. Those shareholders don't actually care how productive a company is, productivity is increasingly orthogonal to profit. They care how profitable a company is, and companies stay the most profitable when (among other things) the labor market is most effectively depressed (and layoffs are a useful depression tactic) and laborers aren't comfortable enough to fight for better wages.
So if anyone gains from this, is the companies paying under-market, and therefore typically getting somewhat substandard devs to do boring things. There's major savings in productivity there, precisely because the training for those tasks is so good. If you had your highest performance developers spending a significant amount of time on that kind of busywork, you weren't getting that great of a performance in the first place.
But the devs here are all very experienced senior-level engineers. Perhaps that's why we aren't seeing any gains?
I'd say it depends on the task more than on dev level. If it's maintaining, improving, or refactoring of a large project LLM can't help much. When it comes to small utilities GPT-4 is a time saver. You can even write a small playable game with it in a matter of hours. Using any popular language. Simple web pages, JS are much easier. With ChatGPT-4 I've learned how to do certain things using ffmpeg. Without it would take hours of googling.
PS: talking about ChatGPT degradation with the time. I didn't see it on programming tasks. But it became noticeably faster.
Epic went from 2200 employees in 2020 to almost 9000 in 2023. They are now basically shedding 1 out of 6 hires they've done since 2020. That's not really about sustainability, it's just a way to drop the ones that didn't really work out.
These layoffs are just a way for CEOs to improve "efficiency" AKA increase margins so they get a nice fat bonus - COVID overhiring has been a convenient excuse and every company has jumped on it.
Well, this sounds logical and all but isn't always.
For example, Google's revenue/employee exploded during covid but it didn't crater post-covid [1]. So even with a hiring spree, the company is on the same track as it was pre-covid. You might need to layoff employees to get that same covid high but that covid high is the unsustainable part not your current employee count.
So some companies may have overhired compared to what their reveneue growth without covid would've been but not all did. You still need to evaluate companies on a company by company basis.
[1]: https://www.macroaxis.com/financial-statements/GOOGL/Revenue...
Newly hired people were more junior and cheaper than previous employees so it makes no sense to fire them.
Which industry? I’m seeing a lot of middle layers being reduced, but not the rainmakers.
It doesn't even work, even if it was productive, as barely any problems the average gamedev faces on a day-to-day have been surfaced to the training data of an LLM.
Sure, if you're still learning how to move position from x to x+1, ChatGPT can help you with those basics, but anything more than that and they start hitting me up on Discord looking for real answers.
Something as simple as how the serialization of fields in behaviour classes was too much for ChatGPT. They read what it hallucinated up, got confused, and were calling me up to explain minutes later. For those who don't know Unity, this is something you have to do daily when building new logic into game components and ChatGPT was clueless about it.
The limitations of ChatGPT as a programming aid are very very obvious and it has a long way to go before it's really useful to seasoned professionals.
Specifically with type-ahead suggestions I believe with typing you are more engaged in thinking about what is correct. With a CoPilot suggestion the tendency is to just skim it to see if it look good. And that's probably part of the problem; the LLM is trained on what looks good!
This may look good:
But it's supposed to be: Good times.If you're more of a generalist though? I can see productivity gains of 25% maybe.
Deleted Comment
Is that really how it works? It seems like the biggest gains are actually in the less skilled segment.
They're living in a momentary dream that AI LLM tools will solve all our coding problems. They don't and reality hasn't hit them yet.
I don't know any real seasoned dev who is getting a 30% boost because they installed Co-Pilot. The tools poop out too many errors/fake-garbage to be useful to devs working on battle hardened codebases.
Regarding the second para, the interesting part for me is: how do these companies determine which are the unproductive devs? My impression so far is that senior management has no way of measuring this, given how little data is provided to back up decisions to RTO.
Not so sure about the third para. Why would one company be more ahead of the game than any other, if indeed LLMs are providing this supposed boost?
There's a lot of "you have to cut X off your salary budget", pick whoever you want. Then it's up to the managers to pick, very subjectively, who they want to work with the most.
FFS - I managed to dodge 4 layoffs at a startup, including my whole department being cut wholesale. My managers always kept me, even though I wouldn't call myself the "best" or "the cheapest". All of it just boils down to one thing - I'm the one that whoever manages me can rely on. Just that... being reliable.
The senior management delegates. At every moment your direct manager knows whom they are willing to part ways with if asked. Even if everyone on the team is great there is always some criteria (costs, impact on important project, institutional knowledge) that sets some people apart.
fun sieve(n: Long) = (2L until n).fold((2L until n).toMutableList()){p,i->p.apply{for(j in i * 2 until n step i)this[(j-2).toInt()]=0}}.filter{it!= 0L}
and it works OK in some situations:
>>> sieve(100L)
res5: kotlin.collections.List<kotlin.Long> = [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97]
Let's try another test input -
>>> sieve(10000000L)
java.lang.OutOfMemoryError: Java heap space
Oops.
Also if the 12th-20th most popular language (Kotlin) being too obscure is the reason, how will they fare currently except with the most popular languages?
This is about as simple as a function as you can get. 95% of my programming work is standups, sprint planning and grooming, looking at Swagger or other API backend documentation for information, looking through Outlook, Teams and Confluence for information about my story, looking for UI on Figma, contacting various teams for more information - then someone wants me to review their pull request and stop to do that - but notice they have some business logic wrong because they don't fully understand the domain. Yet even something simple like a function generating a list of prime numbers, ChatGPT struggles with.
Not that it can't get better, but it's no threat to most programmers I know in the short term (and if it becomes one, I'll shift over to writing Python scripts that call Pytorch, numpy etc. - I already do so in my spare time).
I personally don't believe its a massive function of Generative AI (although it could be), I think the major contributing factor is that the projections that companies made in 2020/2021 and the hiring they did to resource more projects to take advantage of those rosy projections ended up falling short of real expectations. i.e instead of needing to take advantage of a potential 20% yoy future growth in their target markets, they only really need to staff for a 10% yoy growth.
Also probably a bit of "monkey see monkey do" where a company that actually needs to layoff (meta) lays off a bunch of people, and other large companies that may not "need" to layoff (like Google) follow suit since now its "socially acceptable."
There is still some insane growth
Fortnite's game as a service fatigue, for some reason they didn't jump on the Extraction Shooter hype and it cost them today IMO, Fortnite became successful because they jumped on a trend at the right time (from Survival Base Building to Battle Royal), they seem to have forgotten that
EPIC launcher is still bad today.. which doesn't make me want to have it open, and let alone to browse it.. which might hurt their numbers
Steam's launcher follows the same path it seems, slow and bloated CEF bullshit.. makes me want to use it less
I have been evangelizing GitHub Copilot (+ Copilot Chat) and ChatGPT Pro to my coworkers that include full-stack .Net/React devs, Python data science people, and embedded C devs.
Only the .Net/React devs have been sticking to both Copilot and ChatGPT Pro. The Python data science are using ChatGPT Pro but have dropped Copilot. The embedded C devs are using neither.
When working with embedded systems - I barely ever have "lightly" customized code. Meanwhile the closer you're to the frontend - the closer you get to having majority of your code being boilerplate.
If they really did over-hire during the pandemic and fail to plan for the future that was either 1) intentional and they lied to recruits about the duration of employment or 2) incompetence on the part of leadership and recruiting. Have heads rolled in the responsible departments?
Probably 0%. What I heard, mostly people not performing (which is the real pandemic after COVID-19). If I'm not wrong as other companies, devs there are not "allowed" to use copilot and chatgpt fearing code plagiarism and ending up with some license violations..
Epic's game programmers are already incompetent to the point that Fortnite is basically falling apart at the seams and every update adds dozens of new bizarre bugs, I don't even want to imagine how much worse it would get if they started using AI to write code.
Likely not a reason for layoffs. If someone was fired due to llms it was probably due to the incompetence to think the chat bot is more than a data leaking search engine.
Dead Comment
Pretty cynical and unfair IMO. Meta layoffs were the first time the company did layoffs in its entire history. Layoffs are super unpopular with employees. It's not like they wanted to do it.
This argument is lazy. Like suddenly there's a huge wave of layoffs, and your explanation is "they could." That really offers no explanation for why it happened now vs. literally any other time.
Scaling quickly has risks and the current wave of layoffs is just a way to minimize risk. It is probably better this way than raising tons of cash until things are falling apart and you are forced into panic mode.
If some divisions or individual employees in a company are not making meaningful impact on the bottom line and are not long-term strategic for future profits, then it makes sense to let them go.
Just because these are hugely profitable companies, doesn't mean you should be spending money on things that will not lead to you generating additional funds. That is like keeping real-estate rented but not filling it with people -- it is wasted resources, whether it is humans or real-estate.
These companies over-hired or were sloppy in terms of their spending and that has changed in the new environment.
Lumen is really nice as long as you never hope to have >60fps and don't want open worlds.
The architecture of it is just fundamentally not compatible with the majority of AAA games, especially open world or esports ones.
It's not that Epic are stupid it's that they're "optimising" in all directions and it sometimes causes issues. -- marketing advertises features that are not fit for the purpose they're being marketed for. Lumen is excellent for small games that aren't so CPU heavy or alternatively for non-realtime renders such as scene generation VFX.
Unreal itself is extremely bound to single-core CPU performance, so the idea of adding extra weight to the main render thread (which is basically the only thread) is unconscionable and was a deeply unwise decision.
The game just needed a more likeable protagonist and a less crowded release window, but I'm sure the takeaway for EA execs is that it shouldn't have been a new IP...
https://www.unrealengine.com/en-US/metahuman
Pretty amazing. Although I don't know if I even need characters this real in my games. At some point it becomes too much like real life and kind of boring? I think I like the fantasy aspect of games too much and I like to escape.
https://medium.com/locodrome/unreal-filmmaking-building-a-st...
I haven’t studied Unreal enough to say for sure, but I think those tools can be a boon if they enable a less technical developer to do things that normally took a lot more time.
Ko-fi is 25GB for free and 200GB for paid accounts, so most things will fit. I'll probably cross-post it to Patreon's new shop feature in the near future just to be safe.
edit: and confirmation
https://www.songtradr.com/blog/posts/songtradr-bandcamp-acqu...
That said, there are ways Songtradr could abuse their new position of trust. I hope they're smart enough not to do that, but I also hope that an archival project for free downloads on long-inactive Bandcamp accounts is underway...
https://www.theverge.com/2021/5/4/22418782/epic-games-store-...
No ragrats!