Suresh G

@JoshCaughtFire @JamesWard Here's an old benchmark that shows something different at scale - pkolaczk.github.io/memory-consump… . I'm sure virtual thread footprint and scalability have improved a lot since then

@sur3shg @JamesWard The reason it’s 2KB is because go pre allocates them and recycles them, which is faster and cheaper than having to resize. Most of that 2kb is empty on init

Generally developers think of Go as being great for concurrency. Its not. JVM approaches are vastly superior. And even some of the best in the whole industry when you include virtual threads, structured concurrency & Effects.

@JamesWard Just curious, is there any language/runtime that provides a simpler concurrency model (the old, clean, synchronous blocking code we used to write) than what can be achieved on the JVM with virtual threads + structured concurrency + scoped values (ugh, no more context passing)?

@JoshCaughtFire @JamesWard >and importantly is very memory efficient in that model Just curious, how? IIUC, each coroutine stack starts at 2 KB, whereas a virtual thread is anywhere from 300 bytes to 1 KB. The only footprint advantage for Go is value types, which are still WIP on the JVM.

I think it’s important to remember the time when go was created, Java had none of that and its concurrency model was a nightmare because of shared memory model and having to manually reason with locks and mutexes, etc and resource management And Go does excels at concurrency within its model, and importantly is very memory efficient in that model. But many of the things people would do in Java is an anti-pattern in go, rejecting shared memory patterns entirely That said, Go by design is a bit locked in time as far as semantics and does not have the ergonomics of recent PL changes, but also doesn’t have the baggage of supporting multiple paradigms at the same time

@headius @Cufe_Haco >but it is not the default behavior yet openjdk.org/jeps/534 - I think it will become default in JDK 27. For Lilliput 2 (4 byte), this is the response from Ron - reddit.com/r/java/comment…

@Cufe_Haco Sounds like an interesting idea. With Lilliput, JRuby objects do get down to 16 bytes baseline, but it is not the default behavior yet. I assume your 64 bits would need to be added onto that? Maybe you can show me an example repository that explains it.

If we can reduce object size to 16 bytes, my Dual-Encoded Bit-Mapping could potentially allow us to store relational numeric values directly in the header, eliminating the need for heap allocation entirely for complex math. @headius the builtins work is loosely based on these experiments ive kept to myself. I've been prototyping a method for relational arithmetic that treats a 64-bit word as a phase-rotated vector. It’s showing sub-microsecond stability on an ARM quad-core for deep recursive operations with Cruby. I haven't given it a shot on Jruby....yet @yukihiro_matz, I'm looking at a way to implement 'Field-Based Arithmetic' in the Ruby core. By utilizing bi-symmetric bit-encoding (where 0 carries relational value), we can perform 16-bit operations with a near-zero instruction carry. It effectively turns a recursive loop into a 3 phase-rotation of the CPU clock.

@getsome_air Any plans to support real-time follow-ups (live prompting) like in Junie CLI? I don't want to queue most of the time - #real-time-follow-ups" target="_blank" rel="nofollow noopener">junie.jetbrains.com/docs/junie-cli…

Introducing Air’s latest feature, the prompt queue. No more waiting just to type the next instruction – queue your next prompts, move on, and Air will deliver them in order when the agent is ready.

@deusaquilus Hi, I know you're busy with your new projects , would you be open to a new ExoQuery release with support for 2.3.20 and 2.4.0-Beta1? Happy to submit the changes if you're ok merging and publishing.

@sur3shg Will do very soon! I'm about to announce something new that I've been working on for the last few months.

@deusaquilus Hi, Kotlin 2.3.20 was released a few days ago. Any plans for ExoQuery to support it? I've created a GH issue to track it with the details.

@why_oleg Python, Go both have log packages in their stdlib, and iiuc those are the canonical way to do logging in their ecosystems. IMHO, any non-trivial app (CLI or service) needs some kind of logging, and having something out of the box is helpful.

@sur3shg I’m curious, why do you think that logging should be first party? Not every popular/modern language/ecosystem has first party logging API, or, even could have one, but no one will use it (hello Java)

Why does everyone wants to create a logging library? I mean, I’ve seen more than 50 Kotlin libraries, most of them doing roughly the same… There is even four libraries with ~+1000 stars! And still, every month in a while one more is created And still, there is no default choice

@thdxr Clickstack, works out of the box for traces, metrics, and logs - clickhouse.com/docs/use-cases…

why has no one vibed up really good local otel tooling i want to be able to look at all these traces during local dev

@KonradSalkowski @ThePrimeagen @dillon_mulroy A better scalable, performant, long-term maintainable application with less headache?

@ThePrimeagen @dillon_mulroy maybe it is, but dude, at what cost?

if you're trying to break into software dev, learn java and spring then apply to every company you've never heard of on the fortune 500 list

@ThePrimeagen @dillon_mulroy JVM is a great platform for building performant applications. We are running many services written in Kotlin and Java 25, and it's been far better than anything we've used. Solid platform, good performance, monitoring and observability, huge ecosystem.

@dillon_mulroy i have literally several shorts and several times on live stream saying this java is so seriously slept on as means to employment its absurd

Wow, the latest @JetBrains Junie EAP CLI supports remote sessions, you can resume a session from the browser junie.jetbrains.com/remote/

@pavel_sergeev11 @ajitid @vdtankov Are there any plans for Compose Multiplatform to eventually use Kotlin Desktop Toolkit as its window toolkit and move away from AWT?

@ajitid @vdtankov Yes, you're right. Air uses Kotlin Desktop Toolkit for window management, Skia for rendering, and Compose Multiplatform on top of it to build UI components.

We released air.dev — a new Agentic Development Environment by JetBrains. Claude, Gemini, Codex, and Junie side-by-side with you — powered by code insight, beautiful UI, and vast development tooling. But underneath there's a technological iceberg. Let me walk you through it. 🧵

@vdtankov Nice...does this make Air the most popular production ready Compose Desktop app? Not sure if JetBrains Toolbox has migrated to Compose UI yet

On Desktop, Air renders via a Skia-backed canvas with declarative UI powered by Compose Multiplatform. The result: writing desktop UI actually feels good for our developers. That happens rarely in the wild.

@getsome_air The same API key is also working fine on Junie CLI also.

@sur3shg You should not put the token on that screen, rather you should click the link from the #1 where it will generate the code to paste here. Please clarify which subscription do you use?

Air now supports Gemini CLI and Junie – alongside Codex and Claude Agent. Choose the agent that fits your task. All agents are included with JetBrains AI Pro or AI Ultimate. Prefer your own API keys? That works too. Enjoy maximum flexibility, all in one place. Download Air for free: jb.gg/7auq36

@getsome_air I'm using a personal API key from Anthropic, provided by my enterprise, logging in through the Anthropic Console doesn't work in my case. That said, the personal API key works fine with third-party AI providers in IntelliJ

@getsome_air @pashtet_qa Thanks... somehow the same token which works fine on IntelliJ third-party providers shows as an invalid code on Air

@sur3shg @pashtet_qa You can click Claude Agent on the activation screen and log in using your API token. Or do the same from Settings - AI Providers - Anthropic - Connect

@pashtet_qa @getsome_air I see. I'd like to see an option for setting API tokens, similar to IntelliJ's approach. This would be really helpful for enterprise users who have API keys. While Anthropic's Claude API token works fine in IntelliJ, it's not possible to use it in Air.

@sur3shg @getsome_air Have you already logged in with any of the options here? Connect your agents here is BYOK actually.