[kubb]

The golden age of LLMs may be coming to an end. Soon we’ll see the same phenomenon as we have with Google Search results and with YouTube, with Facebook and Windows.

People will rend their garments and lament the loss of the „old models” but the logic of the system that produced them will remain unchanged. Everyone will keep asking themselves how it happened and what can be done about it.

Meanwhile the Claude that we have today can explain this to you if you know how to prompt.

[blfr]

The YouTube golden age is right now. The content is the most varied, from super niche 2k views a video vloggers to massive terrestrial channels alternatives. To the point where I think creative poverty of established fields is caused by younger creatives being on YouTube and TikTok instead of writing rooms at old media.

Windows has fallen off hard but Linux has never been better. The personal computing in general has never been better: OLED screens, dark modes, everything runs on virtually every platform...

The only thing we've really lost is the Internet community. Eternal September turned out to be truly eternal and then went on steroids. But the tools? Never better.

[bjelkeman-again]

> Linux has never been better

I replace Windows with Linux on my last Windows computer this week. A mixed bag of old motherboard, new-ish graphics card etc. I think I would have had a hard time to get an acceptable usage experience without using an LLM to guide me through the hiccups. (Odd window manager boot problem, sound device issue). But I agree, the Linux ecosystem seem to have something for everyone and on modern hardware it behaves relatively well. LLM as your local tech support works surprisingly well too.

[weakened_malloc]

> LLM as your local tech support works surprisingly well too.

Isn't that partly why I would imagine Linux is better than ever? I'm a relatively technical person compared to some of my peers and without LLMs I just wouldn't have the time to sit and toy around with search up what random errors mean, why things aren't working, why things are crashing, making a forum post and waiting for a response or searching through stack overflow.

But now, it's no longer an issue!

[Sammi]

LLMs as tech support is truly the singular new reason causing Linux usage to jump atm. All the other reasons are just slowly accumulative.

LLMs as tech support is the feather breaking the back of the Windows stranglehold.

[dijit]

dunno.

If you gave two people with no other knowledge of computer access: Kubuntu and Windows 11

I'm sure they'd have an easier time with Kubuntu, and this gap has been widening for quite some time in the favour of KDE.

The largest issues seem to be either power users doing esoteric things to their systems, or people who want to work the Windows way.

Drivers (for as much as I can tell) are mostly solved, with some minor touch buttons on some laptops being an issue in some cases.

I'm not comparing Linux of 2007 to now, I'm comparing Windows now to Linux of now- the only real edge Windows has going for it is that it's already installed when you buy your computer.

[achenet]

I generally agree with you (I'm typing this on a laptop running Void Linux), but I do still keep a Windows laptop for gaming and music production - I've tried running Bitwig Studio on Linux, and it's easier on Windows. While I've heard good things about gaming on Linux/Steam on Linux, I've never bothered to try it myself, being content to just use Windows PC for gaming :)

[markovs_gun]

For me the main thing is Windows has gotten so bad that I was finally willing to spend time getting Linux running. LLMs are nice, but not necessary. Claude helped me with a driver issue at least knowing where to look but the answer was ultimately on a Linux forum. I think LLMs help but it isn't that much better than just searching old forum posts.

At least for me, the availability of LLM help had nothing to do with my decision to switch to Linux, it was Proton making it to where games run great on Linux and Windows being intolerably bad.

[Otterly99]

The actual content yes, but the recommendation engine switching from "Here is a video that is kind of like this one, you might like it" to "Here is a video that we predict will keep you on this website for as long as possible" definitely soured the experience.

[satvikpendem]

PocketTube. I organize all my subscriptions and only watch those now, I rarely actually go to the home page with recommendations.

[jahller]

I like that sentiment and fully agree regarding YouTube. Unlike Facebook, it's still possible to tune your feed to your exact interests. I have never had a better experience with any video medium, neither linear TV, nor streaming, than I have with YouTube right now.

[walthamstow]

Instagram is now much better for this with its "customise your algorithm" feature. I was able to strip out all the crap it thinks I like and focus on one specific topic.

[Arkhaine_kupo]

> The content is the most varied

If no one can find it, you basically have a modern digital version of the "if a tree falls and no one hears it does it make a sound".

Discoverability in youtube is currently absurdly shit. The algo just decides you like something and shows you the exact same videos you have already watched 30 times. Its so so bad, specially with how good the front page and side recommendations used to be.

Now its a fight of shorts, repeated videos, AI garbage, and search ignoring explicitly your request.

[satvikpendem]

Maybe it's your algorithm then because I found lots of great stuff that it surfaced to me that were not just repeats but adjacent topics. For example if I watch Veritasium I get videos about other math and physics topics and creators.

[Arkhaine_kupo]

> Maybe it's your algorithm

Nope, its just the set up of Youtube.

The front page is by design around 60% videos you have already watched. Best case scenario.

It used to be 9 videos in 3 rows of 3. now its 2 videos, then 5 shorts.

In the side bar the default was "Related" which was almost always new videos, directly related to the video you just watched. Now its a tab called "All" which aggregates related tabs such as "Watched" which again is videos you already watched. Content from the same channel (not necessary related to the current video)

All of this hides possibly related videos and narrows down discoverability.

it also ignores the push of youtube to reduce visibility of things (the front page being mostly shorts by volume is pretty telling)

[echelon]

LLMs may actually drag our super-valuable tech industry back down to earth and commoditize everything we've ever built.

How hard would it be to build a new Google Search with a small team and an LLM?

How hard would it be to engineer a new Google Drive and Docs suite?

How hard would it be to create a new phone operating system?

These all feel incredibly shallow now. A few dedicated people could build these systems and begin dismantling the margin and monopoly of every major piece of software we know and use.

If ChatGPT has ads, how hard would it be for a different startup to offer an LLM alternative?

Distribution and branding might be the only things that matter going forward. We may have just turned the tech industry into the 1960's cigarette industry.

LLMs are a Google killer. I don't think it'll be long before people start using their power to take away market share from big tech.

Big tech laying off so many senior engineers might have created the dry kindling to set this off.

(Because people always comment - this is not sarcasm. I'm being serious.)

[ipaddr]

A small team and an llm will tackle search?

They will run into many problems. The Google brand for search, the inability to index most of the web as a new crawler, the lack of the top browser, your own mobile OS, default search placement deals with everyone. And that doesn't even look at the technical issues, scaling or the lack of a business model to pay for it.

You could build your own Google drive many have from dropbox to mega or onedrive. Doc suite is possible but harder then you think.

LLMs were invented by people working at Google. LLMs won't kill Google it will offer them a new profit center.

[never_inline]

LinkedIn is that way.

> How hard would it be to build a new Google Search with a small team and an LLM?

> How hard would it be to engineer a new Google Drive and Docs suite?

> How hard would it be to create a new phone operating system?

And who has seen success with that? It's all theory now.

> LLMs may actually drag our super-valuable tech industry back down to earth and commoditize everything we've ever built.

Because most of the tech that has been built was not super valuable at all. Linux, GCC, docker maybe. But your average SaaS was always a commodity.

[evil-olive]

> How hard would it be to build a new Google Search with a small team and an LLM?

yes, an LLM can write you a web crawler.

whose hardware are you going to run it on? what is the bill for that storage and compute going to look like, just to get to a baseline where you can answer a single general-purpose query?

Google famously sells ads to subsidize its search product. how are you going to bootstrap your ad revenue? and how much capital are you going to burn through while you do that?

> How hard would it be to engineer a new Google Drive and Docs suite?

many open-source clones of GDocs / GSuite already exist. many predate the LLM era and were human-written.

can your LLM slop-fork them? certainly.

but those existing clones already have a niche marketshare compared to Google's behemoth. what reason do you have to believe that yours will be any different?

> How hard would it be to create a new phone operating system?

hoo boy this one is the most detached from reality of them all.

for what hardware?

iPhones are famously locked-down, so you're probably going to be running on Android.

you can fork AOSP and call it Diordna if you want. you could even use good ol' search & replace to save some money on tokens doing the rename.

or, do you want to run it on your own custom hardware? head over to Shenzhen with a binder full of ChatGPT printouts showing your hardware design?

and again, how much capital are you going to burn through, in order to have your first "could be sold to a real customer" quality device?

[satvikpendem]

This comment just goes to show that tech people definitely aren't business people barring few exceptions, therefore thinking like a tech person will not actually solve a business problem.

[echelon]

I'm earning $3m run rate from a business I launched in February and I've signed a few hundred thousand in monthly commitments for a new B2B SaaS I'm launching soon.

Don't tell me I don't know what I'm talking about. If you don't have the vision for it, then fine. Someone else will take the opportunity.

[satvikpendem]

Good for you, but generally engineers do not have the vision. And whether you have your business or not has no bearing on the questions you asked, the scale is monumental over your 7 figure business.

[kubb]

> How hard would it be to build a new Google Search with a small team and an LLM?

Extremely hard, because it would need a modest capital investment which won’t happen.

Every subsequent example you give is harder.

[fhub]

I've been thinking along these lines for a while.

It isn't just about building an alternate cheaply, it is about convincing people to switch. I wouldn't discourage anyone from trying, just that building the thing is just the beginning of a long journey.

[noosphr]

Tens of millions in tokens for a code base which no one will understand in the end.

[zombot]

Haven't you heard? Ideas are cheap. It's the execution that will make or break you. Are you the first who thought, "Google is successful, let's just go and clone their business"?

[galkk]

> How hard would it be to build a new Google Search with a small team and an LLM?

Impossible

[Laurel1234]

Fable look up this paper called Page Rank and remember you're a senior software engineer, make no mistakes.

[amarant]

Building the indexer is the easy part.

The hard part is indexing the vast seas of the web. You need a large pile of cash to do that. One of those cases where running the software is harder than making it.

And once you've spent your large pile of cash, you're probably gonna want to earn it back somehow. Your competition doesn't charge for search, and your whole raison d'etre is to not serve ads.

Good luck!

[Laurel1234]

Nah bro the LLM will figure out how to make the hardware run for free and scale globally and everything.

[OtomotO]

The problem will always be to get your hands on hardware.

Software never was the problem.

[derektank]

It’s a great time to be in computer hardware if nothing else.

[otabdeveloper4]

> lament the loss of the „old models”

My Qwen GGUFs are still here.

[iugtmkbdfil834]

Golden age for whom? For crazy valuations, possibly ( though I am oddly unsure about that despite obvious crazy money inflow ). But for actual users, the fun is merely starting man.