Thanks, I've been trying to build a digital garden (or a cemetery) for some time, but with interconnected sites/articles/toys on different domains rather than a single Obsidian Publish style site.
Tangent: I recently found a beautiful and extremely well structured digital garden made by Maggie Appleton (https://maggieappleton.com). So, now I'm just jumping between either being inspired by the amount of research she's shared or analysis paralysis since there are so many interesting approaches to the subject matter.
LLMs, I have an `editorial-team.ts` file containing their personas which I feed into the system alongside WSJ news headlines from an RSS feed.
Obviously with this kind of work, the good stuff requires luck (== quantity) or a manual curation process. I just wanted to run this on autopilot.
I might add a feedback loop at some point, e.g. poets with better engagement become more motivated and those who perform poorly start getting jealous... a little bit of drama here and there.
It sounds like you're against this. I'm not sure why.
SEO is a scam to convince search engines to show something other than most useful content. Content farms are a drain on human attention. Human intelligence had it's chance and we squandered it. Bring on the AIs, they're not going to be worse than the mess we've already made.
Sure, but you would as a generality have gotten BETTER information from the real sources from whom the SEO-oriented farms took the info and optimized for the search engines. They're repurposing someone else's original content.
No, the way it works is having people write the content. Now are there crappy ones that just copy content, write bad stuff, etc.? Sure. But there are also those with helpful content. Here’s a content farm I only recently found that has a lot of helpful original content: https://bootspy.com/
Prompt trainers and tweakers and optimizers will probably replace the SEO types.
I'm terrified of a world where I ask, what beverage will be served on Mars, and the answer will be engineered to be given as Coca-Cola is a popular drink possibly available on Mars, or right now. Would you like it delivered?
Why? The only difference is that everyone can afford it. It is probably better in the end than only having paid chills doing it, since people will realize they can't trust "internet consensus" as being a actual consensus.
The real problem for some sites, is the amount of believable spam on mindless rant feeds, like Twitter and Reddit etc.
Fire one of the many offline LLMs up on a university network or somewhere else that a single IP connecting many times won't be too suspicious and let it go to town.
> Hot take: SEO is dead, content farms are done, almost no one will leave Google/Bing to read more than what it generates.
I disagree. What incentive does a website have to produce content if it receives no traffic? Website owners want control over their traffic and the data too. They will wall off any search engine that doesn't give them anything in return (basically they're scrapers at that point). The search engines that do this will starve themselves of information they need to stay relevant and the first search engine to play fair continuing to share traffic will get an advantage.
More likely, Google and Bing will continue to play fair and send traffic. Perhaps AI features will become browser features instead, complementing websites, not replacing them.
I actually think this is what prevented Google from deploying generative AI before. If the search engine, which existed to direct searchers to websites, becomes an answer engine, which does not direct people to websites, it will take some time, but large quantities of the sites Google consumes will cut them off.
The trade is info to Google for visitors. It's not hard to guess how sites will react when that trade stops.
It’s a pain to having to prompt everything. People still want to see certain identity when browsing, for example browsing CNN or wired. Prompting “show me some interesting tech articles”, and you know a lot of that is Ai generated, it feels stale
Wrong. The low-quality content farm will be out and the high-quality site will be fully paywalled (no more data to AI).
For instance, when searching for information on setting up RDS, like high availability MySQL on self-hosted servers on-premises, neither ChatGPT nor Brad provided the necessary details. I would not make a publicly searchable blog or Github repository if I were well-versed in setting such RDS on-premises. Instead, I would make it available only behind a paywall. Why should Google and MS/Bing benefit from my hard work if they do not direct traffic to my site? Previously, low- or high-quality pages benefited from SEO, but that is no longer true. There is a wealth of knowledge that has not yet been documented. Unless Google or Bing starts directing many traffic, experts, and websites will not share updated and other valuable information. I anticipate more paywalls for high-quality content, similar to good news sites.
Besides that, be careful about wishing for one or two centralized sites controlled by big tech giants with AIs such as Google and Microsoft. The web developed because of many 3rd party websites.
I don't have answers on how good 3rd party sites will survive with Google/Bing sending traffic and Ad clicks, but the equation is about SEO changing. Either you create non-profit Wikipedia-style content or high-quality content paywalled. That is my best guess.
> I would not make a publicly searchable blog or Github repository if I were well-versed in setting such RDS on-premises
You won’t. But I will and I think others will as well. My goal is to help people set up RDS and I don’t care about commercialization. So happy if AI reuses all the material.
There are many technical blogs that are just hobbies, or fun, or brand PR, or book tie ins or whatever.
You aren't helping people, though, you're helping a for-profit company earn billions of dollars off your, and other's good will and knowledge. The for-profit company is helping people, in exchange for direct and/or indirect monetization.
Your blog post's knowledge-dump ends up never being seen.
This is why StackOverflow reacted so poorly to Stack Exchange's announcement they would be using all of your hard work to train a for-profit model.
It might not be illegal - but it sure does feel weird, at a minimum.
> You aren't helping people, though, you're helping a for-profit company earn billions of dollars off your, and other's good will and knowledge.
This has all been played out before. I’m cool helping companies along with people.
Imagine you discover free energy. Would you not give it away because you don’t want billion dollar companies to benefit?
I have a choice to give away info or keep it. Id rather give it away than not. I’m not going to sell info. It’s not worth the time for whatever negligible gains it would make.
Even if there was a magic way to only allow “good people” and not companies to use my info, I’d still allow companies because I think information should be freely given and exchanged and hope that others feel the same.
Imagine where we would be if Linux wasn’t used by billion dollar companies. The world would be way worse. Amazon would be poorer but the world would be much poorer.
If you discovered free energy then you and your family are probably getting whacked rather than disrupt the billionaire-class who depend on scarcity, artificial or otherwise.
Governments have basically subsidized energy so that it is as close as possible to free for generations now. Tons of industrial processes rely on super cheap energy, one reason China has such a powerful steel industry is because of how cheap energy is over there.
Energy is one of many inputs into industrial systems, and it is a fairly commoditized one at that.
Free energy doesn't mean you don't need tons of infrastructure still in place, tons of parts and labor to install that infrastructure, factories to make the parts of that infrastructure. And then when it comes to using that energy, large capital expenditures are still needed to get up and running, which means the rich will keep getting richer.
Or to put it another way, electric cars + solar panels are totally a thing you can do right now, and at no point was anyone's life put at risk to bring these about.
Odds are the first country to get free energy would see obscene economic growth, all sorts of absurd things become possible. For start, salt water + energy => fresh water, solving the droughts that our farmlands are facing.
All metal processing uses tons of energy.
Meanwhile, my local electric co, Puget Sound Energy doesn't even make a billion a year in profits. The nation's largest energy company, Exelon, barely clears 2 billion in profits per year.
Microsoft makes more profit in a week than the nation's largest energy company does in a year.
Whilst I don't disagree with a lot of the points you're making (they're great):
ExxonMobil's operating revenue is $413.7bn USD. I'll be surprised if someone isn't getting whacked if they let the free energy genie out of the bottle.
My family lived in South Africa post-WWII because my gruncle's skin grafts wouldn't take in the UK climate. The Kariba dam gave free electricity to the surrounding area, the idea that this was something people paid for elsewhere just blew the minds of everybody there. Free energy is not a new thing.
However, politics and socioeconomics play a big part in whether or not a country can make successful use of energy, but the possibility of something being possible doesn't mean it'll happen. Furthermore, today, water reserves are a problem for the dam, and power generation was being suspended late last year due to it.
> ExxonMobil's operating revenue is $413.7bn USD. I'll be surprised if someone isn't getting whacked if they let the free energy genie out of the bottle.
Also isn't ExxonMobil one of those strange vertically integrated situations where most of their revenue comes from themselves? They drill the oil, then process it through their own plants which results in them paying themselves lots of money, but I may be mis-remembering that.
Anyway, solar is coming, wind is coming, and electric cars are coming, and no one is getting gunned down in the streets.
Heck China would love infinite free energy! So would any other country that wants completely dominate heavy industry overnight.
> Energy isn't cheap in the UK.
It has been a long time since the UK highly subsidized heavy manufacturing.
Energy prices have of course been going up everywhere, as, most people, realize cheap energy that results in dead kids and cancer for everyone was, in retrospect, really stupid.
> Furthermore, today, water reserves are a problem for the dam, and power generation was being suspended late last year due to it.
Infinite free energy would solve all water reserve problems, although it might end up being an environmental disaster near ocean shorelines...
I feel similarly to prepend: I often write things up because I want to help people by sharing knowledge. I don't care whether people get that knowledge via directly seeing it on my site vs hearing it second hand from a friend vs reading it in LLM output.
It's not important that my contribution go directly from me to the other person; what matters is that they get the information they're looking for.
I don't have a strong feedback loop here, but I generally figure that if I make it public then search engines (or now LLMs) will give it to people who want it.
With a blog post - you get feedback via comments. With SO, you get feedback via votes and comments. With HN, you get feedback the same way.
We have feedback on all of these systems so that humans can more readily discern correctness. An AI model is not going to be able to readily understand what solution is correct, lest it wouldn't need to crawl the page in the first place.
Almost all of my feedback today is via comments immediately after I post something, but I suspect more of the readers are people who come in via search. That's not that different from a future in which the searchers are intermediated by an LLM.
> Your blog post's knowledge-dump ends up never being seen.
I don't think this is true - it ends up being added to a soup of knowledge that is ladled out as people ask about it. It might not directly be seen, but it may contribute more to the overall knowledge about the space since OP's knowledge is now able to be sliced, diced, and recombined.
It sucks for the individual who wants to be credited, and it is annoying that megacorps will benefit directly. But for someone who wants to increase the knowledge in the world, it's not so bad...
> But for someone who wants to increase the knowledge in the world, it's not so bad...
You may actually be decreasing the knowledge of the world.
It's not difficult to imagine a future where people do not know how to effectively communicate, and instead rely on ChatGPT (or whatever) to communicate for them. Or a future where nobody knows anything, and instead blindly trusts the output of AI models.
There's already dating services that let your ChatGPT bot woo someone else's ChatGPT bot...
How does your example (bots chatting with bots on dating sites) decrease knowledge? I can see that it dilutes knowledge, but overall isn't the size of the pie decreased?
Isn't it possible for knowledge to be effectively different based on how much incorrect stuff there is?
For example if one person knows two correct facts about a specific sport, and another person knows those same 2 correct facts as well as 2 false facts, and doesn't know which 2 are true or false, isn't the second person less knowledgeable about that sport than the first person, despite technically having the same amount of true facts on the subject?
Equally, even if the amount of known knowledge in the world remains the same, isn't a world in which every human while studying spends every minute learning true stuff vs. spending half their time reading made up stuff, won't people on average be less knowledgeable despite the fact that no global knowledge got destroyed in the process? (Ofc there's never been a time when it's possible to spend 100% of the time learning completely accurate stuff, hypothetical exaggeration there.)
I think you are just hating on LLMs because its a new thing. You can similarly argue that internet will ruin our social life because we no longer have to go to the library and meet with strangers researching up similar topics. But I don’t see anyone arguing against internet?
Progress is here, old man. Thinking that people will deliberately go out of their way to grok through pages of PDF documentations hoping to find that one line of relevant information instead of just asking is stupid and misguided.
> think you are just hating on LLMs because its a new thing.
Along with your "old man" dig, the fact that you have to attribute hidden motives to someone who's got specific concerns elaborated in their comment doesn't do you any favors.
> You can similarly argue that internet will ruin our social life because we no longer have to go to the library and meet with strangers researching up similar topics. But I don’t see anyone arguing against internet?
I know smart people who argue that paper maps reduce social ties because they remove the necessity to ask for directions. They're not wrong. It doesn't mean maps aren't really useful, or that they even want maps gone. It means they're paying attention to dynamics and maybe even better prepared to fine-tune them by introducing something else to the system.
People do argue against "the internet", btw -- the problem of determining authority/reliability of an internet source, hidden incentives of search or feed algorithms, sand castles, and yes the capacity for isolation or echo chambers from virtual interaction. They're not wrong either.
> Progress is here
Change is here. Thoughtful people can recognize that comes with potential advantages and look with concern on hazards and disadvantages.
Sure you're helping people. The ai gets trained and thens tells people the stuff you learned
The billion dollar company was already profiting off that goodwill by putting ads beside people looking for it
As far as the blog is self advertising for your resume or company, it's just as good since you're linking it directly. Maybe even better since the ai will have opinions about you
They were upset when StackOverFlow, a for profit company announced using their answers to train a model. But they added content to the same for profit site for free?
Especially when they knew it was allowed by the CCL?
> My goal is to help people set up RDS and I don’t care about commercialization.
This feels like the same philosophical difference that leads to debates about GPL vs. Apache licenses. The "how dare someone else make money using my work" camp vs. the "if someone else is able to make money using my work, more power to them" camp.
Monetization is not what GPL vs. Apache license is about, even if that's often how the Apache proponents want to paint things.
The GPL doesn't prevent you from making money; it only requires you to contribute your changes back. That way others can also make money based on your work, like... you just did!
I agree and have been having this discussion for decades of people who don’t care and just want their work to help others (Apache/bsd/mit/etc) vs those that want to have a specific ideology furthered (gpl) that all software should be free.
It’s not exactly the same, but I’ve frequently heard “I don’t want someone to take my code and just sell it.” So want to use gpl or some other license that opens but prohibits selling. I don’t think the latter really works out that well, but people seem to ask for it quite a bit.
Ha, jokes on you, I’ll just be using LLM’s to churn out obscene amounts of plausible-sounding-but-actually-incorrect amounts of content on the same topics, hosted across a variety of domains.
Almost anything shortcut LLM’s provide can be a countered and aggressively wrapped in return.
N.B. I’m not actually going to do this, but it’s still a straightforward attack.
Yup. Commercializing things is hard. If I want to share info, I'll make a Reddit post or maybe a blog post if it's something lengthy.
If I want to paywall it, I don't know where to start. I imagine there's keys-included solutions, but I doubt they can just send me e-transfers for each payment. Maybe they can send money to my PayPal? I've never used that for receiving money before. Do I need a business bank account? Will transparency rules require me to share my address publicly? Maybe I need a PO box. I just wanted to share something helpful.
It seems to me like the options will be 1) publish free content on your website and make zero money because it will get crawled and displayed to readers elsewhere or 2) publish content on your website behind a paywall and make zero money because no one will pay for it (even if they could somehow find it, which they couldn’t).
> may work for some bloggers and news and news-adjacent publications
Yes, for creators who have managed to establish an audience that had high intent to consume that creator’s content. But good luck to people who are trying to make money just by delivering useful niche content where the consumers don’t even know of the creator.
Why wouldn't Google [replace with whoever becomes owner of leading AI platform] just hire experts themselves to generate for LLMs?
Seems like the obvious next step. For clarity, I'm not saying I particularly like that future, but I have a hard time grasping why you think a paywall would be an obstacle here. Especially not for the wealthiest corporations (ever). Additionally because most paywalls to information have very hard time keeping information from leaking. Otherwise news orgs would have figured that out by now!
I'd guess that google is currently the single largest funder of "content" in the world, between adsense on websites and payouts to YouTubers.
They're certainly not getting the content for free. They're running ads against it, and making money on it, but they are also paying for it. It's a relatively small shift in their funding model to pay out to creators for their contributions to an AI model instead of paying out per pageview.
> I'd guess that google is currently the single largest funder of "content" in the world, between adsense on websites and payouts to YouTubers.
Those are advertisers funding the content, not google. google is just a middleman. They generate no content and pay none of they money, they just facilitate the transaction and take a cut.
While I upvoted you, I think there’s still a lot of room in this world for human created content. In fact, I wager it’ll be more important than ever. People will be actively seeking out human-attributable information so there’s someone to hold the buck on where information originated from.
I’m generally pretty pessimistic about the effects LLMs are going to have on the world, but I tentatively agree with you on this point. In fact, a silver lining might be that SEO content marketing becomes worthless, so we’ll have fewer product review listicles that didn’t actually review products, but just did a roundup of other articles in hopes of getting some affiliate link clicks. Genuine writing is usually less directly profit driven, so maybe there will be more room for that in traditional search results hereafter. Still a raw deal for anyone that blogs or writes tutorials and now OpenAI makes money off of them.
is this a hot take? it seems obvious that 95% of searches for answers will end at chatgpt/llm (even if no good answer returns) which yes will kill SEO and traffic to content websites to begin with.
I work for a company who's main traffic comes from questions that google doesn't spit out the entire answer to except below the first few results.
I asked chatgpt and of course it gives the answer right away. We also have searches that land on pages with fancy ui's where user has to enter info to get answers. Took me about 1 minute to get chatgpt to spit out the answers and once I knew how to prompt it I could shorthand the input and get answers much faster.
I tend to agree, because if Google doesn't, OpenAI et al will. The real question is how does Google balance this against the loss in advertising revenue?
Promoted links or native ad formats in AI search results would be an almost too obvious guess.
Then putting advertising in whatever format LLM outputs are dressed as.
My biggest fear is, abuse eg how LLMs for search can work with targeted ads, eg making you slightly prefer Coke vs. Pepsi, or One Political Brain Parasite(tm) over another.
For the past 10 years I've been using google as a search engine and nothing else.
I type for the page or content I already know in advance I'm looking for and that's it. I never browsed or clicked on promotions. I found out about new sites and products, always through social media, or news pages. I will continue using it with the same purpose, I can't imagine people spending 15 minutes on a search engine trying to discover something new.
Why do people post worthless anecdotes like this when we know from trivially available data that Google makes over a hundred billion dollars per year from search.
With enough anecdotes, that 100 billion dollars starts to smell like click fraud. Google owns both sides of the ad market and is less than transparent with how their ads perform or how bids are calculated.
I will be very curious how the promoted stuff works with LLMs. When you ask a LLM what the best brand of shoes are, can folks pay to influence the result?
I personally know people spending 6 figures at the minimum on SEO, because they see huge conversion rates for their own businesses (which are not web-based stuff by the way, more of a niche in person technical services).
Thank you content writers for your service training LLMs. Here's 10 amazing ways to find a new career in a LLM world.