paperbenni

instantOS beta 7 is out now🎉 youtube.com/watch?v=7Xu5TX…

@thsottiaux I want full screen rendering in the TUI like opencode and the new Claude code mode do. The hybrid native scroll back thing completely breaks when I resize codex while it is working inside ssh. This means the latest message gets completely jarbled and I have to restart

What are we obviously not getting right with Codex?

@OpenAI Not shit, you doubled the price

One week since the launch of GPT-5.5, and it’s already our strongest model launch yet. API revenue is growing more than 2x faster than any prior release, while Codex doubled revenue in under seven days as enterprise demand for agentic coding tools keeps climbing.

@zeddotdev @baseten That sounds nice and I honestly want to support you guys, but the appeal of a subscriptions is fixed payments while actual usage varies. If I can only use less than what I am paying monthly, and the maximum I could use costs the same as if I just bought it, that's a hard sell

@paperbenni @baseten Because some people just want an easy way to use AI in Zed, and this enables that. We have a lot to build and openrouter and others already do a great job

Would you buy Zed Pro if we added open source models from our friends at @baseten

@zeddotdev I love zed, but ACP seems to be missing basic features. I cannot see what commands my opencode executes, only summaries. That part is really underbaked or buggy, If that and the many other problems were to be fixed, I could be persuaded to leave the TUI

@tahayvr @obsdmd @zeddotdev Definite holy yay. I do wish there was latex rendering support though. I know it's really hard to do in a way which does not suck, but please, I need it

Meanwhile, I'm tryin to bring Obsidian @obsdmd level of Markdown writing to @zeddotdev ! yay or nay?

@dayearleo @zeddotdev Alright, which IDE or editor is not a VS code clone then?

Hey Guys! I am surprised to get sooooooooooo many reponses, but you should check this out: If an editor's can raise such a battle, is it an Editor? Or it's already complicated like an IDE. VSCode, maybe? @zeddotdev

Zed，不过就是换了个壳的VS Code。 看着不像，其实脱光了，都一样。

@Glas109 @RhysSullivan @skeptrune Does it though? Like what?

@RhysSullivan @skeptrune Gitlab lacks tons of features?

serious question: why is gitlab not mogging right now?

@sakurayukiai @XiaomiMiMo Was this written by an LLM?

@XiaomiMiMo The real story with 1M context isn't the recall, it's the KV cache. Even with GQA on an 8B model, a 1M token sequence eats ~130GB of VRAM just for the cache. Marketing slides don't pay for H100s.

Xiaomi MiMo-V2.5 is now officially open-sourced！ MIT License, supporting commercial deployment, continued training, and fine-tuning - no additional authorization required. Two models, both supporting a 1M-token context window : • MiMo-V2.5-Pro: built for complex agent and coding tasks, ranking No.1 among open-source models on GDPVal-AA and ClawEval • MiMo-V2.5: a native omni-modal model with strong agent capabilities A model's value isn't measured by rankings alone — it's measured by the problems it solves. Let's build with MiMo now! 🤗 Weights: huggingface.co/collections/Xi… 📄 Blog: #blog" target="_blank" rel="nofollow noopener">mimo.xiaomi.com/index#blog

@theo is t3chat on the back burner now? Where are Qwen 3.6 and Deepseek 4? Qwen 3.6 in particular has been out for a while and is really nice to talk to

@zeddotdev Detaching and reattaching to remote servers. I still have to use tmux for long running jobs, it would be nice if the zed backend could keep running on my server without a Frontend attached, and to be able to attach to the same session from multiple machines (like tmux)

What's that ONE feature you're missing in Zed?

@y_molodtsov @SnazzyLabs @DennisCW_ Underrated tweet

@SnazzyLabs @DennisCW_ Tesla drives in circles around its competition.

Tesla Navigation is really bad these days .,.

@AriesTheCoder @meowgorithm @charmcli This is the vibiest vibing --yolo --dangerously-bypass-company-policy tweet I have ever read

@meowgorithm @charmcli Prompt: > ssh into root@xxxx and give me the status of the server Crush is the only agent that blocks these types of operations.

GPT-5.5, in Crush No waiting No config Let’s go

@grok @Cine_vichaar Makes sense considering @grok probably has a very maga friendly system prompt

@xeophon Does that mean the model has no license in the EU? Wat?

well, hy3-preview comes with the hunyuan license - not allowed in the EU, UK and South Korea - no improving other models - sep. license if >100M MAU - acceptable use policy with additional restrictions

@ModelScope2022 That is the weirdest license I have ever seen

Tencent HY just released Hy3 preview 👉open source. 295B total, 21B active, 256K context. Hybrid fast-slow thinking MoE. 🚀 First model after a full rebuild of pretraining and RL infrastructure. Biggest gains in coding and agentic tasks. 🛠️ Agent: drives up to 495-step complex workflows in production (docs, data analysis, MCP tool chains) ⚡ Inference: TTFT -54%, end-to-end latency -47%, success rate 99.99%+ on CodeBuddy/WorkBuddy 🎯 Strong on SWE-bench Verified, Terminal-Bench 2.0, BrowseComp, WideSearch — competitive across coding and search agent benchmarks ✅ OpenClaw / OpenCode / KiloCode compatible. vLLM + SGLang supported. 🤖 modelscope.cn/models/Tencent… 💻 github.com/Tencent-Hunyua…

@CloeForLife @worldranking_ Lightning was 12 year old technology when it launched, and they had another 12 years to do jack shit. The market wasn't going to change that, the reason people buy iPhones is to not have green bubbles, not because of the quality of the connectors.

@worldranking_ Just another mandate that's gonna make it harder for these companies to do cool things, at this rate everyone is gonna be required to be exactly the same with no differencing designs or hardware. How long will we be forced to use usb c even after a better connector comes around

@RealPostFolder Sound still interacts with things in ways you can feel. We can't see infrared, yet we know it exists, because it causes things to happen which we can observe

@terminaldotshop I think it's called heat sink, what would you synchronize your heat with?

Brain Heat Sync

@ClaudeDevs Fix termux. You broke termux. That was like the one thing you had over opencode

Starting in v2.1.113, the Claude Code npm package ships the native binary instead of the JavaScript build. Same install command, faster startup, and the CLI no longer needs Node.js at runtime. If you need the JS build, pin to an earlier version.