I really appreciate this post, Kent. After trying CoPilot and then switching to JetBrains paid version, I just gave up on JetBrains in favor of relying on Gemini because Gemini could at least comprehend my files while JetBrains required me to keep pointing it at specific code fragments. Your post allowed me to better understand the problem.
That said, this post seems like one big advertisement. Is Ona unique in its offering?
Update: The post's nature aside, I just used Ona for the first time and am awestruck. Asking it to perform a task and then watching it work through the various steps, running into problems, addressing the problems, retrying, etc. was amazing. It made Kent's 2023 tweet "The value of 90% of my skills just dropped to $0. The leverage for the remaining 10% went up 1000x. I need to recalibrate." hit home in a way it hadn't until now.
JetBrains Junie is awesome, liked it even better than Claude. What was the problem with it? Please elaborate on this, I'm sure it depends on your workflow and how you're trying to use them.
Are you trying to baby step or are you vibe coding?
>my files while JetBrains required me to keep pointing it at specific code fragments
I was using the JetBrains "AI Assistant" for which I paid a monthly fee. When I wanted it to refactor some file, it often said it "couldn't load that file" and required me to point to the specific functions I wanted modified.
That’s my intention, to add context to what seem like impossibly rapid changes. I mentioned Ona because they helped allow me to keep writing. I’m transparent about sponsorship because I want you to have a sense of how much to trust me.
I agree the content is not about Ona in particular, but in addition to the opening "sponsored by" comment, which is completely understandable, the last section starts with "Ona represents an important step beyond the current status quo." That explicit tie-in back to the sponsoring product isn't something I've seen in other posts.
At the beginning there was the command line (cit.), and then again I found myself working with terminal wrappers infused with AI. (Fantastic reference to the long gone world of ST!)
My workflow is pretty much identical to what you described, command line to prompt the agents, IDE to browse and compare.
I’m very happy with Warp, index a project root, proceed with small incremental steps, generate unit tests and documentation, validate and refine intentions within the IDE, repeat. When done, agents append to the change log, commit, CI-CD and deploy.
Besides the “warp” speed, my (?) code looks better, works better, and is so much easier to read and understand than it ever was.
Brilliant post, this is what I always felt, a friction of what I want and what the LLM gives me, but couldn’t really tell what it was or what was creating that friction and after reading your post, it made me realize that IDE weren’t built for AI-based coding - review workflows. We need new tools for that.
Ona was previously GitPod, which provided cloud-based development environments you could start easily from a GitHub repo online. I used it (in the free version) for some of my OSS projects but gradually stopped using it as I preferred a local VS Code instance, but it was pretty slick even in that "cloud IDE" form.
I really appreciate this post, Kent. After trying CoPilot and then switching to JetBrains paid version, I just gave up on JetBrains in favor of relying on Gemini because Gemini could at least comprehend my files while JetBrains required me to keep pointing it at specific code fragments. Your post allowed me to better understand the problem.
That said, this post seems like one big advertisement. Is Ona unique in its offering?
Update: The post's nature aside, I just used Ona for the first time and am awestruck. Asking it to perform a task and then watching it work through the various steps, running into problems, addressing the problems, retrying, etc. was amazing. It made Kent's 2023 tweet "The value of 90% of my skills just dropped to $0. The leverage for the remaining 10% went up 1000x. I need to recalibrate." hit home in a way it hadn't until now.
JetBrains Junie is awesome, liked it even better than Claude. What was the problem with it? Please elaborate on this, I'm sure it depends on your workflow and how you're trying to use them.
Are you trying to baby step or are you vibe coding?
>my files while JetBrains required me to keep pointing it at specific code fragments
I was using the JetBrains "AI Assistant" for which I paid a monthly fee. When I wanted it to refactor some file, it often said it "couldn't load that file" and required me to point to the specific functions I wanted modified.
Yea Junie is totally different, under the hood it uses AI Assistant but i haven't had problems with Junie.
It's more agent based / vibe coding (give it a task). It's like Claude. Yea you're going in small steps sounds like.
check this out https://www.notion.so/happyotter/AI-7feaa705793a444b9fb9e954bafaacf4
Thanks for the link to your website, Dave. I appreciate the tools overview!
Likewise if you wanna share what you learn, I modified this notion page's permissions so that anyone can add comments!
I see it as less about Ona in particular, and more a general thought about how tooling might change to accommodate new ways of programming.
That’s my intention, to add context to what seem like impossibly rapid changes. I mentioned Ona because they helped allow me to keep writing. I’m transparent about sponsorship because I want you to have a sense of how much to trust me.
I agree the content is not about Ona in particular, but in addition to the opening "sponsored by" comment, which is completely understandable, the last section starts with "Ona represents an important step beyond the current status quo." That explicit tie-in back to the sponsoring product isn't something I've seen in other posts.
At the beginning there was the command line (cit.), and then again I found myself working with terminal wrappers infused with AI. (Fantastic reference to the long gone world of ST!)
My workflow is pretty much identical to what you described, command line to prompt the agents, IDE to browse and compare.
I’m very happy with Warp, index a project root, proceed with small incremental steps, generate unit tests and documentation, validate and refine intentions within the IDE, repeat. When done, agents append to the change log, commit, CI-CD and deploy.
Besides the “warp” speed, my (?) code looks better, works better, and is so much easier to read and understand than it ever was.
Thank you Kent!
Thanks, Kent! Ona sounds fascinating,
There's something missing in your opening paragraph: "I have a hard to augmented coding without it."
Brilliant post, this is what I always felt, a friction of what I want and what the LLM gives me, but couldn’t really tell what it was or what was creating that friction and after reading your post, it made me realize that IDE weren’t built for AI-based coding - review workflows. We need new tools for that.
Ona was previously GitPod, which provided cloud-based development environments you could start easily from a GitHub repo online. I used it (in the free version) for some of my OSS projects but gradually stopped using it as I preferred a local VS Code instance, but it was pretty slick even in that "cloud IDE" form.