Based on Cursorās new pricing policy, this is a considerable risk.
Iām already on the waiting list. At least it will be powered by Claude. The price seems much better than Cursorās. I guess, at this point, Cursor will lose clients as fast as it gained them.
3 days ago I canceld my cursor account, atm Iām using windsurf. When i get the invite from kiro, Iāll give it a try.
doing the same joined the waitlist at kiro because cursor became just too greedy and they want to just make money off the users thats itā¦ so if its money only focus without quality then its time for a good bye. with KIRO launch i too might cancel cursor at this point because the new pricing system is just robbing people nothing else.
I am using Kiro now, and I can 1,000% tell you it is a beast! I am loving everything about it. I still have Cursor, but I hardly ever use it. It is strictly CC and Kiro.
Kiro is most certainly eating the costā¦and wonāt be able to keep that up, and maybe they are hoping costs will come down by the time they do a full release. The volatility in the āAI-powered-IDEā space is very high because the speed of development is so fast, and these are new tools and we have not really figured out how to optimally use them yet. Pricing and features should shake out over time. Anysphere has raised 1.08 billion dollars, they arenāt going anywhere yet. Having said that, as a user, I do think it is wise to try multiple AI powered IDEs including Kiro, but I think it is way too early to say āX is the new thing and is Y killerā. I think feature will be easily replicated, so the āsecret-sauceā will be velocity (developing the best features first so that you can gain and keep customers), and then any proprietary data and algorithms (tuning etcā¦). If one of these companies finds a way to develop an high quality coding model, with an architecture that is smaller so that it is faster and cheaper, and then run that on specialized hardware (like what groq is doing),that would be great. There is a lot that a coding model does not need to know - like who the king of France was in 1650. Which I think can still be answered by this models without doing we web search, which means that is embedded in the model parameters. That is useless for a coding model.
You can tell by my horrible writing that this post was not generated by AI.
If Gemini and Grok donāt appear in Kiro, then it will be worse than Cursor for complex projects.
I donāt believe that. Gemini and Grok are worst, i donāt maybe you are writing html tags with these garbage LLMs.
Claude starts drowning in context after 80k tokens and might strip out critical functionality just to fix a bug. Brilliant as a QA engineer, but when it comes to building systems ā Gemini and Grok are the better picks.
And itās just too expensive in the current Cursor.
I tried it out yesterday. It is still pretty buggy and a lot of the requests donāt go through. The diff view compared to cursor is also really bad. Right now you need to click on a āFollowā button to find out what files it is editing then there is no Accept/Reject per line or per file - there is only Accept all or Reject all. Hopefully they improve that but I am glad there is another player I can consider moving to.
From the usual UX annoyances in Kiro 0.1.15:
- where are the global User Rules?
- why canāt I edit and resend a prompt?
- why is there no one-click copy for the Agentās reply in chat?
- why doesnāt the user terminal auto-activate the local Python environment?
Honestly, from a practical point of view, AWSās pricing just doesnāt make sense. If they were any more aggressive, itād basically be āfree to useā like trae. Whatās funny is, weāve all been through this phase with Cursor before, and we even saw the consequences of this kind of pricing comingā¦ Maybe Iām just being a bit too pessimistic.
Amazon is a big enough player to afford dumping. Not to mention that Claude is deployed on AWS.
I agree on this. github copilot still costs only 10 bucks but they are lazy to improvise. so aws can afford to lose and in future everything gonna be less expensive as new GPUās and more good model takes over.
KIRO spec mode looks very interesting !
i cant wait to see how much amzn spent on kiro in last month.. hope next earnings call announces how much they gave anthropicā¦ theyāre fully shut down now / locked beta.. cant get in after app update.. haha.. signed up for closed beta list now.. i went wild for 2 days with claude 4 on it.. was super fun.. but their frontend still stinkz compared to cursor with rollbacks, reverts, and how it actually doesnt stop.. and how cursor can run terminal cmds in bg.. kiro canāt.
@Ali_Sufiyan , I was very angry with Cursor after the changes. It was serious miscommunication and CEO apologies for it. But hey - to be honest saying that āthey want to just make money off the usersā by you is very wrong. We all are making money of our work. I am looking at my work - with Cursor I can complete at least twice much work than doing the same work manually. Is this increase of productivity? Surely it is. $20 price tag was unsustainable going forward. I dont like to give away my hard earned $ but 100% increase of productivity is worth more than $20 to me.
I no longer trust cursor this company, they are fickle, rather itās the student subscription that ruined cursor
I also have kiro, which is very nice, but still too young. A few more versions would be great! Iām currently testing trae.ai, which is exceptional !
Cursor has unlimited frontier model use in back ground agent mode, no reason to lose that for 1 dollar.