fairestvalue

how do you differ from tinygrad?

Uber and Lyft are literally the textbook examples of (local) network effects.

OpenAI is not.

Except, in practice (not "traditionally"), the cost of a sophisticated market maker to acquire these constituents is usually much less than if you or I were to trade on the market in our brokerage account. SPY's spread is only 2 pennies wide (3 bps), for example.

Ummm, have y'all thought about spread costs?

If you look at the spread of any of these ETF's mentioned (spread = ask px - bid px), you will notice that the spread is much smaller than if you were to sum up the spreads of each component stock.

That's possible because of a mature ecosystem of ETF market makers and arbitrageurs (like Jane Street).

If you buy all of the stocks individually, as it sounds like y'all's solution does, you will pay the spread cost for every. single. stock. The magnitude of these costs are not huge, but if we're comparing them against VOO's 17 bps/yr expense ratio, it's worth quantifying them.

I imagine eventually you can hope that market makers will be able to quote a tight spread on whatever the basket of stocks a client wants, but in the meantime, users would be bleeding money to these costs.

(Source: I work in market making and think about spreads more than I would like to admit.)