← Back to context

Comment by csense

4 days ago

I'm willing to accept this claim:

- You can make money when your product is a text editor.

I am very skeptical of these claims:

- When your product is a text editor, $42 million in capital can be effectively deployed to make meaningful improvements to your product.

- When your product is a text editor, your lifetime inflation-adjusted profit will eventually exceed $42 million.

Sequoia is apparently not so skeptical, and willing to put the money on the table. That must have been a truly extraordinary pitch deck...

> Sequoia is apparently not so skeptical, and willing to put the money on the table. That must have been a truly extraordinary pitch deck...

They invested like $200 million in FTX and had a glowing article about SBF about mile long on their website.

The big VC firms are by no means immune from just doing plainly stupid things.

Right now all the hype around anything even at the periphery of "AI" is enabling a lot of similar stupidity.

Counterpoint, apparently cursor's revenue is in the 300-400 million range. So it's not wildly inconceivable that you'd do 40m profit (although I too am doubtful).

  • I strongly doubt that Cursor makes anywhere near 40m profit. All they're revenue is spent on tokens with the LLM vendors. I'd be surprised if they are even running at positive margin and not just subsidizing usage with the VC money.

    Unsure of what the end goal is, but I expect everything AI related to be a load-leader right now and then the goal being to figure out how to drive down costs or make even more money later.

    Maybe that's what Sequioa thinks too...

If the world is moving towards open models which are "good enough" – the main winner will be the one controlling the distribution. So if you're controlling the web browser, the editor or the OS – you are the winner.

Obviously, the risk here is very high from this perspective, since nothing guarantees anything.

Zed isn’t just a text editor. It is the only working platform for code assistants.

Neither Google nor Claude nor anyone can’t at the moment get right basic operations like file edits. Zed is flawless in co-operating with most LLM models. And not just that - also switching models during conversation and more.

I am at Zed Pro at $20 but when Zed offers $200 Max plan I will sign up right away.

  • I switched off of Zed's agent system onto the Claude Code CLI because I was blown away by how much better CC was.

    Even with "auto edit" turned on, Zed just kept asking me for confirmation. I'd be like "hey your code has this bug", and it'd be like "you're right, and this is why. here's how you can fix it" ??? just fix it man it's your code. Maybe this is fixed now, but Claude Code never has this issue and is very good at only stopping when truly stuck (and generally for good reason!)

    Changing the topic a bit, Zed's collaboration features seem really good but it's quite hard to use when nobody uses it in the first place. With VSCode, I can use the LiveShare extension and everyone on the team can just join with no fuss at all. LiveShare is likely not nearly as technically great, but the simple fact that people can use it easily makes it win hard.

    Honestly it would be cool if Zed can somehow become more popular thanks to this investment. As long as it can keep its speed and technical excellence. VSCode used to be super lean and cool, but now it's just another fat IDE with unlimited bells and whistles. It feels like how Eclipse felt back in the day.

  • How does this compare to the rust rover with Junie? it is definitely well integrated for code edits and iterations with the IDE

No....there is value in having many developers use your tool for daily code development. You should think if it as a lot more than a text editor as well, just like Github is a lot more than just hosted Git (also VC backed at first).

What they are really selling is an AI service subscription.

They are giving out the actual text editor for free.