undefined | Better HN

0 pointsJumpCrisscross2y ago0 comments

Author concludes the only moats for an AI start-up are captured compute and proprietary data. I'm disagreeing.

Good execution remains differentiated. It just requires continuous iteration, evolution and improvement.

If you build an MVP over a weekend and then pivot 100% of your efforts to fundraising and marketing, as has been the trend over the past decade, yes, you're screwed. You're building dollar apps for another App Store.

Most of the arguments the author levels would have worked against the first waves of computerization, digitisation and the emergence of the Internet, in some cases more powerfully. Yet the prediction didn't hold. Capex and IP weren't sole, or even strong, predictors of new-entrant success. For Exhibit A to the first part, see Softbank.

0 comments

6 comments · 3 top-level

Swizec2y ago· 2 in thread

> Good execution remains differentiated.

Levels is a good example of this.

While we're sitting here waxing poetic about moat or no moat and what to do with these AI things, he's made some 7 figures in cold hard cash revenue from building and shipping things people want.

https://twitter.com/levelsio/status/1669269424543793153

lumost2y ago

7 figures is pretty great for an individual, assuming that there is at least a 50% margin. For a company it's a good start in SF. Scaling a business with no moat will quickly bump into margin compression and a race to the bottom.

Not a problem if you are an individual with no intent to scale, but a big problem if you are investor looking to invest 8 figures.

Swizec2y ago

Distribution is a defendable moat however. Even for investors.

Once you have a few thousand users giving you cold hard cash to use your service, you also have access to way _way_ better product development and marketing information. Not to mention a lot of data you can use to fine-tune your AI in ways that a competitor starting from scratch couldn't dream to replicate.

This is a big part of why you see all these BigTech companies adding AI features. Their existing user-base is the moat.

simbolit2y ago· 1 in thread

> Good execution remains differentiated. It just requires continuous iteration, evolution and improvement.

But with the underlying basis for everything being open source, everything you learned with "continuous iteration, evolution and improvement" can be copied relatively easily.

JumpCrisscrossOP2y ago

> with the underlying basis for everything being open source

The models may be open (or available to anyone who can pay). But so was TCP/IP and the App Store.

Nobody is going to win on the tech alone because nobody has ever won on the tech alone. If your founding team has the relationships to get you ensconced at a beachhead of customers, and your technical team can move fast enough that inertia keeps them put, you have a solid shot at denying that space to other entrants.

gremlinsinc2y ago

An AGI system will eventually eclipse all software. Think about it - at first you can tell it -- Create for me an accounting software with an exact replica of Quickbooks ui/ux. First gen: Sure, give me a minute... 10th gen: Why don't i just do your taxes, no inteface needed at all.

This one AI will be able to do everything you ask it, or give you any interface you desire, should you need one - even creating a real world-like interface inside 3d/vr systems.

It's kinda depressing - the first to succeed in AI (assuming they don't kill us all), will become the top company in the world (if not the only company we need at all, because nobody can compete at all, by the time the next competitor gets to AGI - they'll be 50 generations improved, and it's just a runaway loop of self-improvement.

j / k navigate · click thread line to collapse