Three is already a lot of tools to be doing more or less the same thing. Not many people are going to have informed, experience-driven opinions on Cursor, Windsurf, and Claude Code, let alone the other tools you mention.
As boring an answer as it may be, mindshare matters. Newcomers face outsized inertia in winning that mindshare over from the major incumbents.
I was surprised recently to see that gemini-cli[0] and codex[1] each have way more GitHub stars than Claude Code[3]. Currently 62k, 31k, and 25k respectively.
Stars are only a proxy for use, of course, but I'm not sure what a closer public indicator might be.
I use Sourcegraph Amp (plugin for vscode) and am very happy with it. I assume it works like all the others. I experimented a bit with CC and Cursor (some months ago) but for me Amp works more easy.
I do wonder why never ever anybody else mentions Amp though. Am I their only user?! It explains why it works so fast ;)
> As per another comment Gemini CLI and OpenAI Codex have more stars, and they're both much larger companies.
The GitHub repos for Claude Code and Cursor are basically just issue trackers and some examples. I'm actually surprised they have as many stars as they do.
I’ve built multiple RAG pipelines across Windsurf, Claude, and even Gemini-Codex hybrids, and I’ve learned this:
Most of the current devtools are competing at the UX/UI layer — not the semantic inference layer.
Claude dominates because it "feels smart" during code manipulation — but that’s not a model quality issue. It’s that Claude’s underlying attention bias aligns better with certain symbolic abstractions (e.g. loop repair or inline type assumptions). Cursor and Windsurf ride that perception well.
But if you inspect real semantic coherence across chained retrievals or ask them to operate across nonlinear logic breaks, most tools fall apart.
That’s why I stopped benchmarking tools by "stars" and started treating meaning-routing as a core design variable. I wrote a weird little engine to explore this problem:
You can’t underestimate the viral nature of vibe coding which was almost exclusively done using Cursor and Windsurf. There are tons of YouTube videos boosting both products.
Claude Code grabbed tons of attention because it uses the command line and enables developers to boost their productivity. Because it works via the command line, it doesn’t require using an editor/IDE you may not use, and it’s extensible via custom slash commands and hooks.
It’s only been out a few months and there’s already a cottage industry of websites, tutorials and videos supporting it.
It’s not perfect but it’s really good in ways that matter to developers.
I’ve never seen so many developers making videos and blogposts describing how happy they are to pay $200/month for the Max plan due to the productivity gains.
From my perspective, Gemini CLI is useless—it’s the only LLM that has repeatedly refused to work with me because of so-called "offensive language." And when it does work it's overly verbose and often just plainly wrong. Codex CLI just feels like a stripped-down version of Claude Code.
Claude Code wins for me for one simple reason: I can just pay for a MAX account and not worry about running out of credits or going through my credit card limit. That makes me use the tool twice as much as the others, where I’m constantly watching my credit balance.
Cursor wins because it was the first big one, not sure if it has any important advantages over claude code anymore.
Cursor lets me try all the models with my subscription. The IDE tooling is great. I really dont like 1 line boxes like claude code has. I like to instruct my llms and and read my input. The whole 1 line box claude code gives is super annoying.
Do you mean the text input box only allows one line? In that case, try pressing Alt+Enter for a new line. It's a little unintuitive; I would expect Shift+Enter for a new line.
woah its alt-enter, I always did shift+enter and thought these tools were terrible. They still are terrible if they use alt-enter, what a waste of time lol/
As a CC user I can say there there are at least two reasons why CC dominates.
One, it is very good except for its prise. Because of the cost, I tried Codex, but after a few simple questions I gave up. Nowhere close to CC when it comes to tool usage. Gemini cli asked for a Cloud project id, no idea what is meant with this. Cursor is not an option because I think long term they can’t compete with the integration that Anthropic can offer.
Second, CC offers capabilities you will learn as power user, in that sense it is like a game; addictive.
I’ve used Claude and Gemini and Claude works a lot better, but Gemini is a bit cheaper so I have both and use Gemini when I don’t want to waste my Claude availability.
However, Claude wastes less of my time than Gemini.
In my experience, Gemini is a lot more likely to have an empty tool call and get in a loop of using a tool that isn't working, or get confused and use some previous instruction from the context window instead of the task at hand and go on a totally crazy tangent. Claude is generally a lot more accurate.
There is a YouTuber, GosuCoder [0], who is benchmarking all of the agentic coder tools and in some of his recent videos you will see Claude Code is currently ahead of the field.
Very interesting to see Github Copilot so close to the top here. GHCP is weird because it basically "won" out the gate and then just disappeared from mindshare so hard that nobody's even mentioned it on this thread.
Why do worse products win out or have more mindshare over better products? It is an old question, and the answer is it depends. Use what you think is best, and tell others.
Here is a list of agents I've tried:
- Good: Gemini CLI, Opencode
- Okay: Jules (hampered by its limited sandboxed env)
There is basically no reason to use a tool that is "almost as good" for the same price as the best one. So people concentrate on the best ones out there.
I recently tried Replit. After it chewed through all the available free agent credits. It gave up. I then received an email that my account was banned for violating the terms of service (which includes all sorts of squishy language).
Replit is a cash grab. It couldn't build a basic app with Three.js without stepping on itself.
Claude Code is the only AI tool that can grok my C++ codebase.
Cursor feels right so long as you have proper .md's
Waiting for access to Kiro to try it.
Currently running GGUF models on my own with an OpenAI mock server I wrote so Continue.dev works in VS Code as an agent. Meh. I don't want to spend a bunch of money for a project that I can't use in the end so I don't waste my money on tokens.
Three is already a lot of tools to be doing more or less the same thing. Not many people are going to have informed, experience-driven opinions on Cursor, Windsurf, and Claude Code, let alone the other tools you mention.
As boring an answer as it may be, mindshare matters. Newcomers face outsized inertia in winning that mindshare over from the major incumbents.
I was surprised recently to see that gemini-cli[0] and codex[1] each have way more GitHub stars than Claude Code[3]. Currently 62k, 31k, and 25k respectively.
Stars are only a proxy for use, of course, but I'm not sure what a closer public indicator might be.
1. https://github.com/google-gemini/gemini-cli
2. https://github.com/openai/codex
3. https://github.com/anthropics/claude-code
Both Gemini CLI and Codex are open source, with the actual code in the repo, so that could be the difference.
Yup. Claude Code's repo is basically just an issue tracker.
That indeed is very surprising. I guess they're larger companies (even getting their own engineers to star in Google's case is a pretty big deal).
From HN conversations I'd have said Claude Code was 90%+ of the market by what I remember seeing.
It is on YouTube and Reddit too, having done a bit of a dive into it last weekend.
Clause code massively dominates the AI agentic coding space at the moment.
I use Sourcegraph Amp (plugin for vscode) and am very happy with it. I assume it works like all the others. I experimented a bit with CC and Cursor (some months ago) but for me Amp works more easy.
I do wonder why never ever anybody else mentions Amp though. Am I their only user?! It explains why it works so fast ;)
Amp has only been out for a couple of months, and it's growing fast. Cursor and Claude Code have been around for longer, but obviously that's not always a good thing since everything is constantly changing. Check on X, like at https://x.com/sawyerhood/status/1945865702159925734 or https://x.com/search?q=%40AmpCode&src=typed_query&f=top.
I used Amp a lot when it was named Cody
Why do Pepsi and Coke dominate the cola conversation?
Because they are the most popular.
Is that even true though? As per another comment Gemini CLI and OpenAI Codex have more stars, and they're both much larger companies.
Then Replit, V0, Lovable are all extremely popular.
I don't have a horse in this race just find it interesting. Your Coke vs Pepsi analogy is a good one though.
> As per another comment Gemini CLI and OpenAI Codex have more stars, and they're both much larger companies.
The GitHub repos for Claude Code and Cursor are basically just issue trackers and some examples. I'm actually surprised they have as many stars as they do.
That's true, but not informative. The underlying question is why they are the most popular.
I’ve built multiple RAG pipelines across Windsurf, Claude, and even Gemini-Codex hybrids, and I’ve learned this:
Most of the current devtools are competing at the UX/UI layer — not the semantic inference layer.
Claude dominates because it "feels smart" during code manipulation — but that’s not a model quality issue. It’s that Claude’s underlying attention bias aligns better with certain symbolic abstractions (e.g. loop repair or inline type assumptions). Cursor and Windsurf ride that perception well.
But if you inspect real semantic coherence across chained retrievals or ask them to operate across nonlinear logic breaks, most tools fall apart.
That’s why I stopped benchmarking tools by "stars" and started treating meaning-routing as a core design variable. I wrote a weird little engine to explore this problem:
https://github.com/onestardao/WFGY
It’s more a semantic firewall than an IDE — but it solves the exact thing these tools fail at: continuity of symbolic inference.
tl;dr: The tools that win attention don’t always win in recursive reasoning. And eventually, reasoning is what devs will benchmark.
You can’t underestimate the viral nature of vibe coding which was almost exclusively done using Cursor and Windsurf. There are tons of YouTube videos boosting both products.
Claude Code grabbed tons of attention because it uses the command line and enables developers to boost their productivity. Because it works via the command line, it doesn’t require using an editor/IDE you may not use, and it’s extensible via custom slash commands and hooks.
It’s only been out a few months and there’s already a cottage industry of websites, tutorials and videos supporting it.
It’s not perfect but it’s really good in ways that matter to developers.
I’ve never seen so many developers making videos and blogposts describing how happy they are to pay $200/month for the Max plan due to the productivity gains.
From my perspective, Gemini CLI is useless—it’s the only LLM that has repeatedly refused to work with me because of so-called "offensive language." And when it does work it's overly verbose and often just plainly wrong. Codex CLI just feels like a stripped-down version of Claude Code.
Claude Code wins for me for one simple reason: I can just pay for a MAX account and not worry about running out of credits or going through my credit card limit. That makes me use the tool twice as much as the others, where I’m constantly watching my credit balance.
Cursor wins because it was the first big one, not sure if it has any important advantages over claude code anymore.
Cursor lets me try all the models with my subscription. The IDE tooling is great. I really dont like 1 line boxes like claude code has. I like to instruct my llms and and read my input. The whole 1 line box claude code gives is super annoying.
Do you mean the text input box only allows one line? In that case, try pressing Alt+Enter for a new line. It's a little unintuitive; I would expect Shift+Enter for a new line.
woah its alt-enter, I always did shift+enter and thought these tools were terrible. They still are terrible if they use alt-enter, what a waste of time lol/
As a CC user I can say there there are at least two reasons why CC dominates.
One, it is very good except for its prise. Because of the cost, I tried Codex, but after a few simple questions I gave up. Nowhere close to CC when it comes to tool usage. Gemini cli asked for a Cloud project id, no idea what is meant with this. Cursor is not an option because I think long term they can’t compete with the integration that Anthropic can offer.
Second, CC offers capabilities you will learn as power user, in that sense it is like a game; addictive.
I’ve used Claude and Gemini and Claude works a lot better, but Gemini is a bit cheaper so I have both and use Gemini when I don’t want to waste my Claude availability.
However, Claude wastes less of my time than Gemini.
In my experience, Gemini is a lot more likely to have an empty tool call and get in a loop of using a tool that isn't working, or get confused and use some previous instruction from the context window instead of the task at hand and go on a totally crazy tangent. Claude is generally a lot more accurate.
Perhaps a combination of the Matthew and the network effect. https://en.wikipedia.org/wiki/Matthew_effect https://en.wikipedia.org/wiki/Network_effect
There is a YouTuber, GosuCoder [0], who is benchmarking all of the agentic coder tools and in some of his recent videos you will see Claude Code is currently ahead of the field.
[0]: https://youtu.be/Lon0oRRqB6A?t=334
I watched a recent video of his where he's presenting other models but he was constantly mentioning Claude Code almost absent mindedly.
Even if he did't say it clearly, it felt like he feels everyone else is a distant second to Claude Code.
Very interesting to see Github Copilot so close to the top here. GHCP is weird because it basically "won" out the gate and then just disappeared from mindshare so hard that nobody's even mentioned it on this thread.
I'm curious, did you delete this post? I can't see it in the "ask" tab.
No, guessing it just fell off the algorithm (unless for some reason YC felt it was harming Cursor or whatever)
Yeah this concerns me.. Did they delist it?
emailed the admins and got it unflagged
there was nothing nefarious
For Replit, bolt, lovable and similar tools are very frustrating to work as a software engineer.
V0 is talked about a bit more but they did a good job specializing it more on frontend (not trying to do everything)
Maybe that is why in circles dominated by people closer to the code, those tools aren’t discussed much.
None of them are good. Figma just launched Figma Make recently which is a) decent compared to those, b) included in a Figma license already.
Why do worse products win out or have more mindshare over better products? It is an old question, and the answer is it depends. Use what you think is best, and tell others.
Here is a list of agents I've tried:
- Good: Gemini CLI, Opencode
- Okay: Jules (hampered by its limited sandboxed env)
- Bad: Junie (buggy), GitHub Copilot (dumb)
Here is a nice leaderboard: https://huggingface.co/spaces/galileo-ai/agent-leaderboard
It’s very simple. They’re the only good products.
There is basically no reason to use a tool that is "almost as good" for the same price as the best one. So people concentrate on the best ones out there.
I recently tried Replit. After it chewed through all the available free agent credits. It gave up. I then received an email that my account was banned for violating the terms of service (which includes all sorts of squishy language).
Replit is a cash grab. It couldn't build a basic app with Three.js without stepping on itself.
Claude Code is the only AI tool that can grok my C++ codebase.
Cursor feels right so long as you have proper .md's
Waiting for access to Kiro to try it.
Currently running GGUF models on my own with an OpenAI mock server I wrote so Continue.dev works in VS Code as an agent. Meh. I don't want to spend a bunch of money for a project that I can't use in the end so I don't waste my money on tokens.
[flagged]
roo code