Perhaps I’m missing something, but isn’t Cursor simply a VSCode fork with some AI extensions integrated into it? Hate aside, what is the actual technical value of the product?
They don’t care at all about the vs code fork (probably his colossus data center with 89% vacancy can vibe code a clone in a few hours)
They care about the subscribed users and how they can enshittify their experience by removing access to Claude opus and introducing grok the Nazi vibe coder
Early on in the AI game, before Vibe Coding was even a term, Cursor was the only tool that could connect VsCode to an AI model. Since then, there are now dozens of tools to do this for you. Cursor basically has no unique features, the only value it has as a company is an existing userbase and an established dev team. The worst part is that they are built on top of VSCode, which means their existence hinges on Microsoft benevolently sharing one of their most valuable products, and Microsoft has made several moves to try to kill the Cursor userbase and bring them back to VSCode.
Cursor is one of the worst investments you could make right now. It’s surrounded on all sides by rivals trying to eat their lunch, and they have basically zero IP to protect their market advantage.
Even back then, copilot already existed if you didn’t mind choosing one of their models. Cursor was at most 2 months ahead in features, but already charging twice as much as copilot.
While it is nowhere near the value Elon paid. It is a tool for easy coding with ai. Like the others thousands that out there. I mean spacex buying companies left and right at prices that make no sense is weird. Spacex being valued that much when all they do is burn money at a rate of several states is confusing.
Elon getting richer by the second with all that nonsense is suspicious…
Haven’t used cursor but I assume it’s like other harnesses where the special sauce is in how it manages context, schedules sub agents, feeds them context, enforces standards, uses tools and skills etc. that make it better then just directly prompting the model.
For example you prompt opus directly with “refactor the auth flow” and it’s going to try and “one shot” it and produce the code from that prompt. Whereas a harness has instructions to say
Research the current implementation
Search the web for standards
Ask the user questions on how they want to do it
…
Claude code’s harness isn’t the best, but it’s still better than just promoting the model directly. One thing cursor has going for it is the vector embeddings of your codebase to make the harness more token efficient, but I still switched from cursor to managing parallel Claudes in cmux a few months ago and have had little reason to go back.
They have their own model called Composer that’s specialized for coding.
Personally, I think generic Opus 4.8 beats it, but it is quite cheap and fast. I used Composer via Cursor for a few months in late 2025. I hear they have improved the model somewhat since.
They are also claiming to be training their own new model on SpaceX’s infrastructure. I don’t particularly trust them, but I’ll give the citation anyway just for additional info.
Perhaps I’m missing something, but isn’t Cursor simply a VSCode fork with some AI extensions integrated into it? Hate aside, what is the actual technical value of the product?
They don’t care at all about the vs code fork (probably his colossus data center with 89% vacancy can vibe code a clone in a few hours)
They care about the subscribed users and how they can enshittify their experience by removing access to Claude opus and introducing grok the Nazi vibe coder
Early on in the AI game, before Vibe Coding was even a term, Cursor was the only tool that could connect VsCode to an AI model. Since then, there are now dozens of tools to do this for you. Cursor basically has no unique features, the only value it has as a company is an existing userbase and an established dev team. The worst part is that they are built on top of VSCode, which means their existence hinges on Microsoft benevolently sharing one of their most valuable products, and Microsoft has made several moves to try to kill the Cursor userbase and bring them back to VSCode.
Cursor is one of the worst investments you could make right now. It’s surrounded on all sides by rivals trying to eat their lunch, and they have basically zero IP to protect their market advantage.
Even back then, copilot already existed if you didn’t mind choosing one of their models. Cursor was at most 2 months ahead in features, but already charging twice as much as copilot.
While it is nowhere near the value Elon paid. It is a tool for easy coding with ai. Like the others thousands that out there. I mean spacex buying companies left and right at prices that make no sense is weird. Spacex being valued that much when all they do is burn money at a rate of several states is confusing.
Elon getting richer by the second with all that nonsense is suspicious…
Maybe he’s buying with overvalued useless SpaceX stock. SMH
Ding ding
It’s probably gonna be fine, Elon is just doing what Grok tells him to at this point
I mean even grok says that he is trying to manipulate the stock market.
https://www.youtube.com/watch?v=j8bEzOjFdYQ
/s*
He is buying the users that cursor had. But these users can switch to another AI harness any day they wanted to.
Haven’t used cursor but I assume it’s like other harnesses where the special sauce is in how it manages context, schedules sub agents, feeds them context, enforces standards, uses tools and skills etc. that make it better then just directly prompting the model.
For example you prompt opus directly with “refactor the auth flow” and it’s going to try and “one shot” it and produce the code from that prompt. Whereas a harness has instructions to say
Which produces way better results
Claude code’s harness isn’t the best, but it’s still better than just promoting the model directly. One thing cursor has going for it is the vector embeddings of your codebase to make the harness more token efficient, but I still switched from cursor to managing parallel Claudes in cmux a few months ago and have had little reason to go back.
They have their own model called Composer that’s specialized for coding.
Personally, I think generic Opus 4.8 beats it, but it is quite cheap and fast. I used Composer via Cursor for a few months in late 2025. I hear they have improved the model somewhat since.
It’s fine-tuned Kimi 2.5
They didn’t train this up for themselves.
It’s an OK model that is fast and cheap. It outperforms other models of the same price class, but it’s nowhere near Frontier model capability.
They are also claiming to be training their own new model on SpaceX’s infrastructure. I don’t particularly trust them, but I’ll give the citation anyway just for additional info.