I want them to succeed, but can't see how they can win in this agentic world, where people are barely opening their IDEs, and living in agents terminal whole day and from time to time opening a PR to get a review
The volume of code is only increasing. Problems that used to be solved with a few lines (or even removing lines) are now solved with changes across 10 different files.
My IDE is invaluable for me when trying to navigate this stuff. I am completely convinced that people reviewing LLM code only via PR diffs are consistently missing massive systemic issues. You *need* to review changes in context of the surrounding code.
And doing this without an IDE is just making your life harder for no reason.
That's a small subset of the functionality you would need if you're actually authoring code. So the IDE maybe be more valuable, but the feature set you need is actually much smaller and can probably be redesigned given your narrow workflow.
Is this accurate?
Just some important things off the top of my head that I use daily:
* Instantly jumping between definitions and usages
* Folding and unfolding code blocks
* Quick overview of structures in files
* Database access, so I can browse the data and the code at the same time
Also, sometimes I just want to quickly refactor/rename some variable across all its usages or whatever. It's just quicker to do it in the IDE with a simple keyboard shortcut, rather than prompting an agent "rename oldName to newName".
We have put a lot of thought into how we can add value for the workflow you just described. Here's the way I think about it: FSD for a car. Can you imagine trying to sell a car without a steering wheel? I still want to code, I still want the deterministic side, even in cases where I'm letting the machine do the work.
But the choice of AI providers was very limited, and I found no ways to add the ones I use (the only options for providers were Anthropic, OpenAI and Google).
Any plans of opening up for more providers?
Unless you are blindly vibe coding everything without an IDE to help navigate what is being generated to review the output, working would be hellish.
And if you are blindly vibe coding everything, I am just glad I don't have to work with you. While LLMs mostly speed me up, they make serious mistakes often enough that a review is necessary.
I have opened an IDE for very very few things in the last 9 months or so. Im still renewing a license at this point mostly because I like JetBrains the company and it's not a financial burden but they are in a bad place.
After using Air in PyCharm, it's some good stuff. It's re-imagined for this hybrid world. But my question for you, honest question: do you think you still want to look at code? How about write code? If so, what are some things you'd do, if you were in charge of our IDEs?
Note https://youtrack.jetbrains.com/projects/LLM/issues/LLM-17450... is still open although https://youtrack.jetbrains.com/issue/LLM-16479/Be-able-to-ac... appears to be closed as completed now
Yes, we're writing less code, but the job of "a programmer" has always been about so much more than _writing_ code. It's about code reviews, QA, debugging, discovery, and maintenance; much more than the "writing" part. Anyone with production code will agree, I think: the writing part is the easy part. That's where I still very much prefer a proper IDE over a text editor or terminal. I want easy code navigation, I want proper static analysis without relying or waiting on third party tools, I want a visual overview of changes that happened. I want to stay in charge of whatever is writing my code. On top of that, I want my agents to have access to the tools that have made me the most productive developer for years.
Curious about your thoughts?
Of course, but agents are also opening up new possibilities to do these jobs.
For example, you can do quick first code review with agent, or you can ask agent to add couple of print statements and debug the code until it figures out the issue, also for discovery, you can do discovery with agents and generate nice output with the details you want (although not accurate sometimes)
As a JetBrains user myself, whose organization cancelled the IntelliJ license recently, I can attest things I am doing inside IntelliJ is getting less and less. I am primarily using IntelliJ for code navigation at the moment, which is still the best one
AI can do all of that quite well. If you take your hands off of the wheel, lean into it, start building systems for it, start figuring out where it goes wrong on your codebase and building systems to prevent that... you might get 10x more done. I do, and the results are generally better than anything I'd have coded by hand, because there was never enough time.
I don't look at or try to grok all changes, though. I couldn't get near as much done, otherwise, and I'm not convinced that 1 hr of me looking over it is better than 4 hrs of AI looking over it, and the AI is far cheaper. I look over the code, occasionally, to look for code that's objectively terrible, but I don't fix it directly - I use it as a test-case for improving my AI stack, to make it able to find and fix or prevent that class of terrible.
I probably spend 9 out of 10 credits on review and refinement, not writing fresh code. In spite of that, I'm still producing things way faster than I ever did by hand, with only a single $200/mo Codex plan.
I still keep my JetBrains' subscription, so that agents can use the MCP. I probably spend <5h per month in an IDE myself, though.
I suspect we won't for several years yet - and it's likely going to be way more integrated/specialized for specific technology than our previous ides were.
But you're unlikely to go back once you've the full stack integrated with full observability, release pipeline, system integration tests etc pp
Still, it needs way more integration to work well and not be DOA, hence this is still far out I think.
I do think JetBrains has an advantage here with their background and reputation. I'm constantly reshaping my tools literally every month, so hopefully they build some type of staple development IDE to fit these type of workflows. I don't know what that is yet.
AIR looks interesting, but I can't say this is the "AI IDE moment" yet.
But there's still a group that needs to own the result, who still read code and write code. Folks who still believe in developers and development. This group might not be loud in SV, but they are big out in the world.
I'm a Python advocate at JetBrains and our team runs into folks like this all the time.
I've been enjoying using Air inside the IDE as a way to get the acceleration without giving up coding. I get the speed side from the agent and the trust side from the deterministic IDE.
Not to mention all the lies about the new UI adoption and when faced with backlash, they put the old UI into a plugin that "might" receive updates and support in the future.
I actually did try their latest IDE a few months ago to see if their AI integration was worthwhile. There appeared to be no functional way to use a local LLM. You can set up OpenCode as a provider, but any time you click anything it resets your model to one of the free API services, with no discernable way to change the default.
I don't have high confidence in Air. So far JetBrains has not shown they know what a good AI integration is like. Frankly I don't think they know who their customers are or what they want.
For my part, I'm still using 2024.1 versions on my perpetual fallback license. I had Claude slap together an MCP plugin so now I have my own AI integrated IDE that actually works and functions as a usable tool. All my buttons have text labels instead of hieroglyphs and padding.