as others have pointed out in other comments. the guys at a16z are guilty of this too.
https://x.com/garrytan/status/1506769562468958210
"If you own an NFT..." you are an innovator among innovators. - Garry Tan, 30 September 2021 [quote tweet]
https://x.com/garrytan/status/1443460589049704450
Web3 vs Earlier incarnations technology adoption curve [image tweet]
https://x.com/garrytan/status/1521530531568963584
"airdrops going mainstream could truly upend some % of the centralized ad-based economy of Web2" - Garry Tan, 25 Dec 2021
It's ironic how YC became the Google/Microsoft of its industry.
The tweets don't age well in hindsight but sometimes there are technologies which feel like they might break through but never do. Having been bullish on the concept of NFTs doesn't make a strong argument for or against having good intuition of breakthrough innovation.
NFT's for real estate ownership, container tracking etc. could still have some form of utility. But what people think of when they hear NFT's isn't that, it's shitty monkey jpg's.
NFT's were never the next big thing, except for a very specific subset of very gullible idiots.
(There was actually a short gap between the final collapse of the NFTs and ChatGPT; it's a wonder VCs were able to get out of bed in the morning)
> Having been bullish on the concept of NFTs doesn't make a strong argument for or against having good intuition of breakthrough innovation.
It makes a good argument that he is inclined to be overly impressed by whatever old nonsense people are currently pushing on twitter.
This is true for some technologies but not for NFTs.
Vibe coding is more like the Visual Basic of this generation. It makes it much easier for less technical people to create software or for hackers to be much more productive, but there's still going to be a huge need for professional software development. It's not like everybody is going to become a vibe coder and there won't be a need for SaaS or low code solutions. I think tech people overestimate the capability and willingness for the average Joe to vibe code or engage with technology beyond the minimum required.
Visual Basic has never been well-regarded as a platform for "professional software development," so the analogy doesn't fit in that aspect.
Hmmmm... https://www.ycombinator.com/companies/industry/crypto-web3
Looks like the entire thread was bumped down from 17th place to 130th.
https://hnrankings.info/46120728/
>We moderate less, not more, when a story involves YC or a YC startup. This is pretty much the #1 rule of HN moderation. I've posted about it many times over 10 years: https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...
> I thought HN policy is to moderate less, not more, when YC is involved.
As dang wrote years ago [1], "Moderating HN less, however, does not mean not moderating at all—that would be a loophole you could drive a truck through".
But the real issue with this post is... there's nothing interesting or scandalous here, is there? Just someone (sure, our president) expressing opinions – which are all consistent with his business and investment activities.
What's the issue?
1. Advertise our thesis by building a narrative
2. Evangelize our portfolio
By posting a narrative (that most likely went through some form of Strategic Comms team) comparing Zoho against a vibe coded product while also showcasing some of YC's star vibe-coded products, Garry is able to both craft a narrative that helps support YC's portfolio as well as bring a couple of people to start thinking about Vibecoding. It also acts as an indirect attack on G-Suite without calling out a massive organization like Alphabet by name, which YC needs to coexist with because a large portion of YC portfolio companies will either be acquired by Alphabet or will take or have taken some amount of funding from Alphabet and Alphabet related personal, and Alphabet personal are LPs in YC.
It doesn't matter if the take is right or wrong - it's started a discussion, and maybe one or two Zoho customers have now heard of a couple YC products to consider (outside of the HN bubble, very few people know about vibe coding or AI/ML).
All businesses do this, and knowing Zoho, they will probably leverage this as well as a way to market data sovereigninty and "make in India" by raising the specter of the big bad American capitalist trying to undermine a bootstrapped Indian company.
Anyhow, Zoho has built it's own foundation model [0] and offers an Agents marketplace for domain-specific agents [1]. I'm not sure if can compete head on against a LLaMa or DeepSeek on from sheer performance perspective, but it's good enough (something which a lot of engineers forget is more important than being perfect) to build a "data sovereignty" and "tech nationalism" story which they will absolutely run with as a result.
I have a theory but the primary one is not very flattering to those involved.
Of course he endorses slop coding (Pichai's endorsement is criticized by the Zoho founder for those who do not click through to X).
Zoho will now be a very interesting company for the vast majority of people who hate "AI". I'll have to check it out.
So people will follow those with money (or that they perceive to have money) without much critical thought about where that is going to lead them, they're hoping for wisdom but may end up being misled. That's why all of these ultra wealthy folk turned on a dime when the political weather changed, they don't really have principles, they just want more zeros.
This is perhaps what most non-dev people don't get. Maintenance is a far more harder thing than building something. So you want to go slow when building things, not fast. Either way building things fast has been a solved problem for a while, people don't go fast not because we don't have tools, but there are other fairly valid reasons to go slow. This is true with so many other things outside of software. I guess its called 'haste'.
This is true for most things. Especially where money and life are at stake. But Im guessing you could extend this to anything where reputation is at stake.
Im guessing it doesn't apply to some start ups, but other wise every one is subject to this.
I always thought that NFTs were completely ridiculous and essentially nothing but hype. But then again, I thought that amazon wasn't going to work either, when I was there building it, so I'm not sure that even in a given individual "good intuition for breakthrough innovation" is a unitary thing.
NFTs are just as stupid, if not more so and this time at least it looks like sanity prevailed. But the problem is more complex than just boolean 'made' or 'fail', and I think that's where the investment angle comes in. Investors bet on 'the next wave' all the time. And NFTs looked to the clueless as much as 'the next wave' as mobile phones or the transistor did at some point in time. The big differentiator to me is whether or not a thing like that requires a belief system or not. If it does then I don't give it much chance. But then we have all of crypto as a counterexample and quite a few people got stupidly rich peddling that.
I recall seeing inventory management systems, airline booking apps for travel agents, custom CRMs, internal LoB apps, check-in kiosks, vending machines, etc. with the tell-tale VB UI, especially the typical VB error dialog after a crash!
I messed around with several other UI toolkits of that era -- AWT, Swing, Qt, Flex/ActionScript -- and none was as productive as VB for simple apps. It was just the right amount of simplicity and development velocity for the myriad simple use-cases that were perfectly happy with rigid layouts.
LPs. Your LPs want to hear about how you are using their capital for A.I. If A.I fails 2 years down the line you will get to hide out in the crowd. Contrarianism in an institutional setting is actually very hard to do.
1. A number of media and substack articles about a Twitter beef that becomes a submarine article for Replit, Emergent Labs, and Taskade
2. A couple executives who may have not heard about these startups (this is actually very common outside the tech IC bubble) and will now ask their tech teams to contact them to see if they fit their needs
3. LPs who invested in YC by further reshoring Garry and YC's entire investment thesis.
> it wouldn't work on you...
It doesn't matter that it doesn't work on me - everyone does this form of narrative building (that's the entire point of Strategic Comms teams) so you have to play the game because that's how a Nash Equilibrium be.
Building a zeitgeist is a core part of demand gen, and yea it is transparent and tacky, but even 1 conversion for what was basically 1 hour of drafting makes the RoI positive.
The press nearly always obliges in the first year. It works for wars and pandemics, too. In the second year the first dissenters begin to show up. We are now in past the stage of opinion reversal, where the press and the public mostly hate "AI".
This is also why it is wrong to compare "AI" to the early Internet. After the Internet bubble burst, the public still liked the Internet.
The comparison is from a business perspective, and I absolutely stand by the comparison with the early generation of Internet companies with the early generation of AI companies.
Anyhow, the general public doesn't matter. It's capital (private and public) along with business cycles that create markets, and most of the public doesn't have the knowledge or the capital to make a difference.
The crux of the whole AI boom is exactly how much value it can materialise given the capital expenditure being absurdly astronomical, if it doesn't become the next trillion+ US$ market it will be a huge misallocation of capital.
AI is now a loose used term that is encompassing 3 loosely connected markets that have now fallen under the same umbrella:
1. Construction/Land Speculation: a large portion of the AI story you hear about is a DC construction story
2. Hardware: a large portion of the AI story is just a rebranding around GPU fab and design, especially due to issues around subsidy disbursement under the CHIPS act
3. SaaS/Applications: a lot of products being derisively called as "LLM wrappers" are not cool from a technical perspective, but from an RoI perspective are good enough - $30k for data entry automation that is 80% right is cheaper than hiring a data entry team of 4 who cost $60k each.
Much of the bubble is due to 1, but 2 and 3 are somewhat insulated because of FCF and adjacent markets and narratives to pivot to (eg. For 2 it was "Chip Wars" 2-3 years ago and before that it was "ML" for 1-2 years and before that it was "Precision Medicine"...)
> without the public to consume...
The AI story really isn't a B2C story no matter how much people try to shoehorn it. The Cs are perpetually broke and margins are shit. The value that arises from automation like AI is around workflow and workforce augmentation in some shape or form, which makes it a B2B play.
I don't get his suggestion. I'm paying Zoho for email hosting. What am I even supposed to vibe-code successfully for me to drop Zoho? There's no shortage of open-source IMAP/SMTP servers, in fact Zoho is probably using them too. Sure I can spin one myself, but I'm paying for Zoho for the *service*, not the software.
No idea how good Zoho is, but if you can pay $X/month and never even think about the problem ever again, then that is compelling, and the value of that depends on the customer.
If zoho go bust I'm sure another company will replace them.
You have to remember that no one gets fired for selecting an established software provider. Zoho is a 'safe' pick for SMBs because you know it will work compared to some custom CRM that Bill from accounting vibe coded in an afternoon.
Zoho has its flaws, but for small businesses, Zoho is a godsend. If an SMB doesn't want to pay for Zoho, I'd seriously ask them to recheck if they're making any money whatsoever.
Moreover Zoho is one of the few platforms out there that's really intuitive. Do this, do this and that, and you're good to go. Compared to setting up something like Zendesk or Freshdesk or Xero or QuickBooks , etc.
These statements are so out of touch with reality, I generally wonder where will YC be in 5-10 years.
Why do you need 3 platforms to vibe code a solution? Are each one of them not good enough?
Using same logic, why use Replit, EmergentLabs, or Taskade when you can vibe code your own vibe code platform?
I doubt you'll get all this for $8K/year (India) or $22K (USA) even after you vibe-code once (not counting the ridiculous costs of developing with AI - you still need your engineers to do that even before you start running the token counter).
Garry somehow gave free publicity to Zoho with this post. Maybe that was his intention?
It’s the same reason why vibe coding a better version of Airbnb (even if it’s just a simple CRUD app) wouldn’t actually threaten Airbnb as a business. The product isn’t the moat; the ecosystem is.
Seems ironic to post that - doesn't that same logic imply that most YC companies are worthless?
And I don’t think the friction in vibe coding is that different.
I guess this is just yet another YC ceo/whatever that I care little for (See also: Sam Altman).
Explains a lot.
Feels very disingenuous. I'm a huge proponent of AI drastically increasing efficiency of creating software but we're a long ways away from nontech people replacing and supporting collaboration tools used by medium sized businesses.
In such a hypothetical world, it would actually be much easier for us to fund companies simply because now the only thing we are funding is just sales, demand gen, and projected compute.
We provide funding so businesses that are the right fit can scale out the functions that they need. In some cases it's expanding engineering, in other cases it's expanding sales and demand gen, and in other cases is to subsidize a major purchase such as cloud credits or GPUs.
I'm already vibe coding complex things like GUIs, my desktop environment (NixOS), and last week a Wayland layer shell client that would have taken me quite a lot of work if I had to do them myself from scratch from docs, and I have 20 years of software experience.
The things I spend time building and polishing today with my own time are just next year's vibe-coded minutia.
Some people are going to have a very hard time swallowing this pill, though.
Longer term AI platforms such as Replit could offer easily deployment of ready made app templates e.g. a CRM. However you still need to pay for them much the same as paying Zoho, prices could be lower. But you still need to pay for them and on a monthly basis too. Vibe coding platforms will still be a SaaS business.
Before it, I was using https://wisprflow.ai/ for ez transcription since I would never have had the energy to build it myself nor work out the kinks.
We really are the last generation of software engineers, aren't we.
Surely AWS's overpriced lockin stuff will be "vide coded" away far faster than smaller companies
If I think about who's "site" HN is then PG or Dang are front of mind. I know Dang is /just/ a mod but he is who is visible. I've never once cared who actually ran YC, I didn't care when Sam Altman ran it and I don't care about Gary Tan either, they are 100% immaterial to the discussions that happen on this site IMHO.
Unfortunately, truth be told, most software is like that. Take Salesforce for instance - same issue. Granted, Zendesk niches at this painpoint you mentioned (and I've seen some enterprises switch from Zendesk to Salesforce and face the exact same issue as above), but their insane price increases after the PE acquisition does not inspire hope. UI-wise, Zendesk is still the best.
Speaking in metaphors: someone still needs to build the fixings, bricks, wires, pipes and tiles and paint and tools to put them all together someone has to still come up with the building.
I used to work in construction and I've watched the sophistication of all the pieces get better and better over time. Insulated floors, heated floors, push fit pipes, RCDs instead of fuses, better materials, resin driveways. I live in a new build house, I'm not nostalgic for a sandstone cave.
I've built some crazy things these past few months, I've got a voice control computer with head tracking mouse. Apart from the actual voice transcription model and the neural network model that works out where my face is, I could have built it all by hand. With enough data, I could probably have even made the face position model, but I simply wouldn't have taken these on. They just sound too monumental.
And then we've got the perfect is the enemy of the good. As programmers, we want to produce things that are better for everybody else. So when we start building a library we start considering how all the people are going to use it. Now simply don't care. If it's a bit rough around the edges it doesn't matter because it's for me. And if you want one for yourself, you can prompt it into existence.
I'm a component builder as well. And now I can build polished components that are great for future use. Not only that, they're fully documented. I just wish I wasn't pushing 60. I'm feeling the joy of being with my new computer - 15 again seeing more possibilities than ever.
I've faced serious burnout in this industry. Fed up of another round of read to database, write to database, present data from database. Spending days fiddling with forms, add this box, make this box bigger, that button's got the wrong text. Now I can just say those things out loud into a text box and switch to another text box and build something more complex.
As for whether he's clueless, lets see how investing in tons of vibe coding startups works out.