zavg

Idea: Instead of pure “vibe coding” (just letting ideas flow into code), what if we first vibe code a bunch of reusable components with well-defined, predictable APIs between them? Then we ask AI to vibe code using only these components.

The hope: by constraining the building blocks, we reduce unpredictable hallucinations that often happen when vibe coding an entire monolithic codebase from scratch.

Has anyone tried this? Does it work in practice?

I am thinking about making videos explaining how popular technologies were built starting from the early commits.

Usually the very first versions are quite easy to follow and see how the core was built and how it was developed.

For example, the initial commits of Ethereum or Vagrant or even Linux are quite easy to comprehend.

Then it is easy to follow how the tool was evolving looking through the history of commits.

Do you think that it could be a decent content?

One friend of mine said that the best concept that he learnt is the idea of "doing something X times".

X could equal to 10 or 100, e.g.:

- send 100 emails to check the conversion

- do 100 cold outreach connections to evaluate the effectiveness

- launch 10 projects to see which one runs better.

- try 10 different marketing channels to figure out the best one

- etc.

I found the idea quite useful. Do you use the same technique?

What do you think about the new kind of model to build startups together (disclaimer: I am just thinking loudly)

It is a DAO on blockchain. - When investing money directly you receive some shares of equity - When doing commits to repo you receive some shares as well - Other activities (like marketing, copyrighting, etc.) could be measured and converted to shares as well

Is it possible to become a one-person unicorn? (e.g., to build the company that is worth $1B)

Let's do the figures.

1) Let's count for simplicity that project valuation equals to 10 Annual Revenues. That means that the annual income should be $100M, and monthly income is ~$8M

2) $8M means 400'000 clients paying $20/month.

3) 400'000 paying customers means 20'000'000 total users. (since average conversion is about 2%).

4) There are 5.5 billion internet users worldwide. So 20M users are 0.35% of the total audience.

So, probably, it is achievable to find a specific niche product for these particular 0.35% of the users that is possible to build solo. AI-powered solutions fits even better to this concept.

1. Both are about texts

2. Both types of texts are processed by some interpreter (either CPU or human mind). In the first case we get the sequence of actions as a result, in the second case we get a sequence of the images.

3. In both cases the creator is like a God who creates a kind of universe "from scratch". No any limits for creativity.