ThinkRail from JetBrains

troupo

> a spec graph the agent actually reads Nope. The agent doesn't actually read, and never will. That is, it may pull it into context randomly, and then purge it from context randomly. But any tool/harness/wrapper that claims that agents actually do something is selling you a bridge. --- Also, the entire page somehow manages present a condensed example of the worst marketing + the worst LLM-adjacent speech in a few short paragraphs.

pelagicAustral

I don't get it, is this like an IDE for Claude Code + worktrees but proprietary and with added telemetry?

dude250711

Any company incapable of training a frontier model is trying to squeeze themselves in between you and the frontier model that you are using.

PippaSedai

I feel somewhat bad for JetBrains. Their entire world was flipped upside down. My first intro to a coding harness was through their IDE. Not even a month later and I stopped using their platform altogether. After the year is up I won’t be renewing any of my company’s subscriptions. They have to try to push this stuff out to survive but none of it is innovative. Not sure why anyone would pay for this when it can be one-shot with a single good prompt. Sure, it looks fancy but I never cared about that.

justauserid

I would like to use a harness that integerates with an IDE very easily. This will be helpful for me when coding in a language I largely don't know. I can kinda check things quickly in front of me. I know it just needs an editor and maybe a language server setup – but the cost of visual context switching is still costly heavy for us humans. Often things out of sight are also out of mind. I was aiming for VS Code but since I have used Intellij and android studio a lot maybe I'll give it a try.

julesrms

As usual when there's a harness release by a billion-dollar company, I like to ride its coat-tails with a little plug for mine ( https://github.com/juggler-ai/juggler ) since my marketing budget is $0. Juggler usually gets a warm response from HN people because it's a bit more hands-on than most of them. I'd be interested to hear whether the 'spec' stuff works or not. I've been pondering adding a plugin for juggler based on the fact that we're rubbish at writing specs: so instead, when you prompt it, it'd get the model to list the assumptions that it's making about the task. Then the user can spot and correct any incorrect assumptions, rather than expecting the user to provide correct specs. (Haven't actually tried this yet, just speculating)

brainless

A year ago, it was not sure if the core companies around programming would move to more agent enabled workflow. Companies like JetBrains. I love their products. But it's clear now that programming is now spec management and how do we control agents. I don't need JetBrains for that. And spec management will become less technical as models get better. What does JetBrains think is the long term goal for this product? I don't think it'll make any sense in a few months.

rbehrends

As somebody who likes and uses JetBrains IDEs, I find this fragmentation of their AI strategy (if they have one) frustrating. They now have both Air and ThinkRail in a feature-incomplete state, while the existing AI Assistant – for me far more useful than either Air or ThinkRail – seems to be getting neglected. There are numerous harnesses already, with or without GUIs and editing capabilites; good integration with existing IDE functionality is where there are very limited offerings and where JetBrains could differentiate itself. As a point of comparison, Zed does the agentic side panel overall better (though still not perfectly), but Zed is still by its nature more limited than e.g. PyCharm or CLion or RustRover or Rider.

jtwaleson

I was at a large tech conference this week in Amsterdam, it was < 1km from the JetBrains office. They set up the booth 3 hours too late and there was no one actually there when I walked by (twice). I think they have given up?

faangguyindia

Never used JetBrains products, but always used vi; after that, VS Code for its extensions. But due to agentic stuff, I moved to zed.

nullchan

oh wow this is nice!!

InsidiousWombat

AS Jetbrains user, tf is this shit, I just want AI integrated better in workflow I already use. If I wanted to use VSCode wrapper I wouldn't pay them for it... Their plan is so misguided. First they wanted to pay us for separate tool and THEIR AI, which is worse than what people already pay for, few months back they finally figured out nobody wants that and just integrated it into mainline editor and now this

exceptione

JetBrains should focus on the artisanal engineers and the 'vibing' engineers who need to code by hand from time to time. They have a big moat already. I imagine they try to keep the user base from fleeing their platform, but I think they have already covered prompt integration. There is imho no value in yet another LLM wrapper that seems to be totally unrelated to the existing offering, at least no monetary value. I would urge JetBrains to put all their LLM power and human brains at work to slash their IDEs backlog and deliver the GOAT of all IDEs. Building IDEs is their strength and there will always be a market for that category.

charlie0

They're fighting an uphill battle and it's a race against time. I think they've already lost to Claude Code. I can do everything they offer with Claude. Where is their moat?

lemagedurage

I like what they did with the website UI. I hope they can replace the text with something less AI heavy and more sincere/professional.

solatic

I really don't understand how this is so hard for JetBrains. Build a polyglot IDE that blends the agentic workflow (terminal chat sessions) with the power of an IDE. Index specs and code so that I can jump from specs to implementation and vice versa. Keep indexing all problems found in the project, with a simple "Ask agent to fix" dialog. Expose debuggers with Markdown output to agents via skills so that they stop trying to debug via logging. Keep the nice UI for build/debug/run configurations, which are still hugely important for reviewing PRs and acceptance testing what the agents are giving you. Upsell by charging to host JetBrains-cloud-isolated-agent-workspaces, like what it looks like Herdr is going to do, so that you're not limited by what you can develop solely on your machine, and development continues after you close the lid, and you can review the output with your teammates in multiplayer, like where Zed is going. Is it really that hard?? It seems so obvious to me.

bronlund

One would expect better from a company which makes some of the finest IDEs on the planet. An AI slop webpage trying to sell me VS Code bullshit. Really?

zihotki

> Specs are the ground truth. Oh, I would like to disagree. Specs are transitional and have low value in the long term. The project/product documentation which one should make based on the specs and the actual implementation and found gotcha's - are the truth.

piro0919

For me it's case by case. When I notice a rule in CLAUDE.md has gone stale, I just tell Claude Code to delete it.

brabel

I've tried ThinkRail now. I am a paying user of IntelliJ IDEA Ultimate, but lately I've been using mostly Claude Code Desktop App. This is a cool product. Clearly, they made the agent chat pane the primary feature, while things like the Project Files and the editor itself is secondary (you are not even allowed to change files manually!). The chat is very well done (I think it's just Pi?). People are saying how they use the IDE mostly for reviewing code, and ThinkRail makes that a primary feature as well: every change the agent makes gives you a "Review Turn" button that lets you look at exactly what was done. And the diff viewer is very good (not the same as IntelliJ, seems more like the VS Code viewer?). It stores everything the agent does as a "Turn" so you can review every change at any time, like a git history. It also has the git log similar to IntelliJ, so you can still click on each commit in the history as well. In fact, the two things seem very similar, since each "Turn" normally results on a commit as well. It's stil unclear to me how "Turn"s are useful beyond the commit history. It also has a fully functional terminal so you can try out things easily without leaving the tool. JetBrains seems to have figured out how people are using their tools nowadays, and decided to go all in with this "Innovation Lab" tool. I would still expect to be allowed to edit code/text myself in it, and maybe navigate around like in IntelliJ, but I guess I'm in the minority :P and most people just need something like this?! It's definitely much lighter than IntelliJ (it shows way down in my process list by memory, with around 140MB after quite a bit of usage, and it's highly responsive), so perhaps that's the main advantage, and bringing code navigation/editing/highlighting for all languages (it highlights the most common ones, but not LISP for example) would just result in a heavy tool again. Its way of working with git worktrees exclusively is something I was already doing in the Claude Code App, so I think they got this right. However, CC App can also use a SSH remote connection (like TRAMP in emacs) to a devcontainer, and I've been using that a lot, so would be nice to have that option as well. In conclusion: I like it and I think they may be going the right way with how things are going: the "IDE" becomes primarily a means to interact with agents. But at least for me, it's a bit too early to leave the full IDE completely, and I guess they could've made this just a "mode" inside IntelliJ without losing anything except for being very lightweight, and perhaps that's what they're going to do eventually?