Relay was never just about building software faster. The master plan was an AI venture studio that finds opportunities, builds products, markets them and backs the winners.
I get the overall idea of an automated software/business factory. What I’m rather sceptical about is that turning an average idea/business (or breaking out of the initial bunch of 3-4 similar ideas/business appearing roughly simultaneously), in 100% cases required extraordinary efforts/insights/devotion to make it happen, often against all odds, often via a series of rather radical pivots, rather than via small tweaks.
Ie human stubbornness, crazy dedication, or some soft of unfair advantage (larger angel round, larger pre-existing audience, …) turned products into hits, not small random evolutionary steps or a brilliant initial idea. Ideas are easy to copy. A hundred of people running the same software factory - where’s the advantage?
The idea of automatically developing a portfolio of dozens of small business, while feeding their grows with AI-generated content on TikTok/Reddit/Twitter might sound viable, until you realize that thousands and thousands of people have already oversaturated all channels with AI slop, so unless you have the resources to overspend incumbents/competitors, you stand no chance.
You're not missing much, and distribution is the hardest part. I said as much in the piece: software engineers love to believe that if they build something clever enough, humanity will detect it through vibrations in the Earth's crust. Somebody has to distribute the bloody thing.
Where I'd push back is on hits. I'm not trying to build hits. A hit needs exactly the stubbornness, pivots and unfair advantage you describe, and I agree a factory won't manufacture that. The bet is at the other end: thousands of small, profitable niches that were never worth serving because the software cost too much to build. A product making €8,000 a month that costs almost nothing to build and run is a lovely little business. Nobody with a big angel round is fighting you for it.
And most of the factory's value isn't in the building, it's in the killing. A thousand possibilities might become fifty serious investigations, fifty might become five, and perhaps one gets built.
If a hundred people run the same factory, the advantage moves where it always does: to judgement about what's worth building, and to distribution. On the AI slop point, I think you're right. It's the strongest objection.
This makes a TON of sense to me. It is like in sports when the major league team has a pile of minor league ones feeding it - they can sign dozens of players and see which develop into stars (rare), or into scrubs (common) then only use them if they are star level. Signing kids was cheap decades ago for that, thus why in baseball they used to sign dozens upon dozens of kids from the Dominican hoping one or two would become stars - easily worth the minor fiscal risk.
With AI as it gets better and better the cost of risk will keep dropping - for example, you can test out 100+ different game ideas at once and have them go out to the market and if 1 hits then you win big time. No overhead beyond what the AI cost, no programmers, no board of directors, just yourself and that AI. Can easily see how it will blow up in a few years as too many do that, making it damn hard to find anything online. Early adaptors like yourself will be laughing and making a killing in the early days, then as others figure it out they'll jump in but it will be like mining crypto - early days you could get coins reasonably easily, today not so much.
The farm system is a good analogy and I hadn't thought of it that way round. The crypto point is the bit that nags at me. Once the cost of trying something falls to almost nothing, everyone tries everything, and the scarce thing stops being the building and becomes being found at all. I don't have a good answer to that one yet.
I get the overall idea of an automated software/business factory. What I’m rather sceptical about is that turning an average idea/business (or breaking out of the initial bunch of 3-4 similar ideas/business appearing roughly simultaneously), in 100% cases required extraordinary efforts/insights/devotion to make it happen, often against all odds, often via a series of rather radical pivots, rather than via small tweaks.
Ie human stubbornness, crazy dedication, or some soft of unfair advantage (larger angel round, larger pre-existing audience, …) turned products into hits, not small random evolutionary steps or a brilliant initial idea. Ideas are easy to copy. A hundred of people running the same software factory - where’s the advantage?
The idea of automatically developing a portfolio of dozens of small business, while feeding their grows with AI-generated content on TikTok/Reddit/Twitter might sound viable, until you realize that thousands and thousands of people have already oversaturated all channels with AI slop, so unless you have the resources to overspend incumbents/competitors, you stand no chance.
What am I missing?
You're not missing much, and distribution is the hardest part. I said as much in the piece: software engineers love to believe that if they build something clever enough, humanity will detect it through vibrations in the Earth's crust. Somebody has to distribute the bloody thing.
Where I'd push back is on hits. I'm not trying to build hits. A hit needs exactly the stubbornness, pivots and unfair advantage you describe, and I agree a factory won't manufacture that. The bet is at the other end: thousands of small, profitable niches that were never worth serving because the software cost too much to build. A product making €8,000 a month that costs almost nothing to build and run is a lovely little business. Nobody with a big angel round is fighting you for it.
And most of the factory's value isn't in the building, it's in the killing. A thousand possibilities might become fifty serious investigations, fifty might become five, and perhaps one gets built.
If a hundred people run the same factory, the advantage moves where it always does: to judgement about what's worth building, and to distribution. On the AI slop point, I think you're right. It's the strongest objection.
I read your 8000/mo businesses pitch. Human judgement and taste will become the differentiating factor for sure.
This makes a TON of sense to me. It is like in sports when the major league team has a pile of minor league ones feeding it - they can sign dozens of players and see which develop into stars (rare), or into scrubs (common) then only use them if they are star level. Signing kids was cheap decades ago for that, thus why in baseball they used to sign dozens upon dozens of kids from the Dominican hoping one or two would become stars - easily worth the minor fiscal risk.
With AI as it gets better and better the cost of risk will keep dropping - for example, you can test out 100+ different game ideas at once and have them go out to the market and if 1 hits then you win big time. No overhead beyond what the AI cost, no programmers, no board of directors, just yourself and that AI. Can easily see how it will blow up in a few years as too many do that, making it damn hard to find anything online. Early adaptors like yourself will be laughing and making a killing in the early days, then as others figure it out they'll jump in but it will be like mining crypto - early days you could get coins reasonably easily, today not so much.
The farm system is a good analogy and I hadn't thought of it that way round. The crypto point is the bit that nags at me. Once the cost of trying something falls to almost nothing, everyone tries everything, and the scarce thing stops being the building and becomes being found at all. I don't have a good answer to that one yet.