Thanks to @brextonpham, @devlordone & @yashhsm for the conversation these past weeks to help me clear my thoughts on this topic.

The most recent Y Combinator cohort was asked, “What percentage of your code base is generated by AI?” - One quarter (25%) of the group said that 95% of their code was AI-generated.

Andrej Karpathy (ex-Director of Tesla AI and Founding member of OpenAI) has, for better or worse, coined the phrase “Vibe Coding.”

In a nutshell, this just means that idiots like you and me can turn words (prompts) into code. This has now inspired Ideas Guys to put off getting a job and try their hand at creating the next Facebook.

Pieter Levels, the highly successful solopreneur, has recently shown what is possible with Cursor, Grok, and Claude in his flight simulator game, which is almost entirely AI-generated. The game’s recurring revenue is now reaching $85k a month through in-game advertising alone.

At the minute, there are still quite a few bugs that would require a base level understanding of wtf is actually going on when you are slamming your head into the keyboard and asking it to create an application.

https://x.com/drummatick/status/1901868964621467696

Naturally, some actual developers have popped up to provide some consulting and debugging services for the next generation of AI-only devs (yes, that front end was vibe-coded).

As these frontier models advance and become even better at generating code from natural language (text and speech), we will see their outputs become increasingly effective. Dario Amodei, Antrophic CEO, recently stated that in “three-to-six months” we will be at a point where AI is writing 90% of the code. In twelve months, it will be “writing essentially all of the code.”

https://x.com/slow_developer/status/1899430284350616025

So, what does the world look like if this is the case?…

Who loses?

Well, the cost to build applications converges to near zero. SaaS subscriptions that you have begrudgingly paid for can be replicated at a fraction of the cost.

If a SaaS company does not have such a strong moat, either through network effects, difficult-to-replicate products, or tight legal IP, then they are toast.

I am looking at you, Docusign, and Typeform. Cya, buddy. Also, it will be more efficient to build native products in-house that can connect to teams’ workflows, knowledge bases, and existing databases through MCP.

I’d argue an efficiency hire who knows how to build AI is probably one of the most sought-after roles currently for SMEs.

With some of these incredibly powerful smaller models coming online, too, it will be possible for companies to run everything internally on a local machine/model, which is great for privacy and OpSec. Teams that still believe “build it and they will come” to be an effective way to go to market will also die through smarter teams simply cloning their product but doing a better job at marketing it.

Who wins?

As code generation models develop, the playing field gets levelled from a technical perspective. Teams that can dominate in mindshare and attention will do extremely well. Great marketing comes from knowing your core audience because you are your core audience. If all products are created equally, apps become (even more so) fashionable products that people will peacock over. “I use Claude because I know better than to use the famous ChatGPT, which is for normies who don’t have a clue.” For apps to become something people want to tell others about, they have to be fucking cool. The branding, positioning, and cultural awareness have to all be on point.

And guess what? This is not something that LLMs will be able to keep up with in the near future. Why? Because humans are far more complex than AI can even imagine.

Sub-cultures are hard for those outside of their walled gardens to comprehend. What is impressive to one may not be impressive to another. We may see a world where each individual is greeted with a personal, bespoke user journey, UI, and even brand for the same underlying product based on their preferences without even knowing it. Graphic designers, heads of creative, marketers, and UI/UX wizards all have roles in an organization that require taste and an eye for design trends before they happen.

These roles will excel, and this will always be the case. Models don’t know what they don’t know. They, in their most basic form, are word prediction machines.

The App-Coin Thesis

Note: I am extremely aware that the phrase “app-coin” is very silly; sarcasm doesn’t deliver well on the page.

Well, unless you are giga-connected and are already on the cap tables of the likes of Anthropic, Cursor, OpenAI, Perplexity, and at the app-level Lovable, Replit, LindyAI, etc., then you are gonna have to have fun staying poor. If you are a hyper-gambling super nerd who still believes crypto is the future of France, then there are some teams that I have spoken with recently that are building with this future in mind. The idea is App-Coins. These may reflect and act as memecoins towards the founder or small team of builders that churn out the apps of tomorrow. Tokens are the greatest tool for overcoming the cold start problem, which has sent more ideas to the graveyard than anything else. Founders and small teams that are unable to attract their initial 10-100 users can meaningfully overcome this by leveraging tokens and speculation. There is no denying that humans want to speculate. They also want to speculate in a way that is fair. The idea that you can find the next cracked-out kid who is going to make waves with his latest idea is meaningful.

You see this thing here?

Cal AI takes a picture of your food and gets an estimate of the calories within +/- 10% accuracy. Sounds like everyone’s first business idea, right? This was a team of four teenagers who took one product available on MyFitnessPal and created a $20 million-a-year business. How? By understanding branding, positioning, and marketing to the right audience. They have aggressively reinvested the monthly revenues back into influencer marketing (which definitely works for the fitness industry for discovery), and they have positioned the product for the average user: it has a super simple, clean, yet interesting design.

They have also made it clear that if you are training for Mr. Olympia, then this isn’t going to be for you… How many times has this product not worked? If I were to guess, it is in the hundreds. What we will see next is apps that also launch alongside a token, which, for better or worse, will enable them to get some initial traction at least. Will it be gamed? Absolutely. Will some of it complete vapourware? Probably most of it. Is it going to happen anyway? You bet. What excites me is allowing anyone, anywhere, at any time, to take an idea to market through code generation models and the idea of vibe coding (which, for what it is worth, is already making me cringe at the name). Introverts who don’t like their carrots touching their peas can get a product to market without having to make eye contact or talk to anyone. That is bullish.

Teams I am watching:

• @KaitoAI - Mindshare and attention are number one if any smooth brain can launch a product.
• @devfunpump - Vibe coding platform taking solo founders and small teams from 0-1 on Solana with tokenizing the idea. Love the team and want them to succeed. Brownie points for @buidldao_
• @tryoharaaAI - Another idea-to-project in a few clicks project building on Base with a very slick UI/UX. Very based team again, Brexton is a G.
• @sendAIfun - It started with agents but is now looking to have an app environment that anyone can use. A recent interview with Send is available here.
• @0xtarobase - Quite under the radar, but I believe Alliance incubates them.

Those are a few off the top of my head, but I imagine some brilliant solo creators will come out of nowhere and blow us away. I think we’ll also see a $1 billion project with a solo founder in the next 12 months. TLDR: AI enables anyone to push an idea to production even if you can’t code. Tokenization creates speculation, which gives ‘devs’ a much higher chance of attracting initial users. An app explosion is coming. Finally, if this fantasy I have dreamed up in my head comes true, there will be one of the largest wealth creations we have ever seen. It will not only allow founders to create untold riches but also give the little guy who is paying attention the opportunity to invest in an idea before anyone else. We have seen huge demand to level the playing field and to make it more fair for retail investors. Pumpfun was an equal and opposite reaction to low float, high fdv bullshit, but that still wasn’t it. Investing in the garage band ‘devs’ who know how to market a product is the new opportunity staring everyone in the face. If you think this is you, my DMs are always open. Anyway, I best get back to doing some actual work. Idea Guys, rise, for you shall inherit the Earth. Let me know if you are building anything towards this weird and wonderful future.

Other related reads/listens.

• How to Use MCP & Vibe Coding to Create Your First App w/ SendAI
• A blocmates Guide to Vibe Coding for Crypto
• Is Model Context Protocol (MCP) the Next Big Unlock in AI?

Disclaimer：

1. This article is reprinted from [X]. All copyrights belong to the original author [@Grantblocmates]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.