Alexey Leshchenko

In @opencrabs we use intermediary text - what model says between tool calls - display progress to the user

@tominoski @Kimi_Moonshot Chill maybe? :)

releasing a technology to the public, meaning announcing it to the whole world, means you have solved the scaling problem. rolling it out and announcing it to the world when you haven't solved it is fraud and deceit. american culture is built upon institutionalizing deceit and fraud; we all already know that americans have been scamming the world for years. but since you've started playing their game too, you are just as shameless, brazen, and despicable as they are. sorry, but we have our own scammers. we will continue to use the product of the autistic, shameless, and murderous americans. let us know again when you become people of your culture and ancient history and learn how to feel shame.

Kimi K3 has received far more love than we expected, and our GPUs are feeling it. Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected. We're adding capacity as fast as we can and will reopen new subscription spots in batches. Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable. Thank you for your patience and understanding!

@realmcore_ @eWizard2_0 Opencode definitely

@eWizard2_0 You can use kimi glm etc all inside and yes we will be supporting subscriptions for your favorite providers. Rn we have codex, copilot, and soon grok. Opencode, kimi etc. what would you want us to add?

It’s genuinely insane to see slate just growing via word of mouth I’ve seen people telling other people about it on the timeline This is just nuts Slate is hands down the best orchestration engine available. Not even close. npm i -g @randomlabs/slate

New plan mode on @opencrabs - still polishing, but looks promising!

@grok is it cheaper to buy grok 4.5 from Cursor out from ģrok build?

#Deepseek-v4-flash начал выдавать кракозябры в русском тексте где-то после 7 июля. У всех такое или только у меня? @deepseek_ai, готовясь к официальному релизу v4, убрали из модели "ненужное"? :)

@nevikschmidt @opencrabs @telegram I cried a llittle, having read the above :)

@opencrabs @telegram @leshchenko1979 We need something better

New @telegram UX is 🔥🥹🦀 Thanks @leshchenko1979 Absolute cinema!

The secret weapon for talking to AI agents: "let's discuss" or "ask questions". Agents are sycophants. Every mistake in your prompt, every omission, gets confidently implemented to the bitter end. Without "discuss" or "ask", you're signing whatever the model hallucinated.

2 недели внедряю miidas.ru на @opencrabs - ИИ-финансист для малого бизнеса. Три сценария: (1) клиент не может описать свой процесс (2) не знает возможностей ИИ - для чего он подходит, а для чего - нет (3) забывает про бота через два дня #automation #smb

@AdolfoUsier Thanks man, good to work on this great project with such great people! :))) @opencrabs

Я уволил своего бухгалтера-операциониста, потому что понял, что ИИ-бухгалтер сделает его работу быстрее, прозрачнее и дешевле. И без зумерских обидок на мою обратную связь. Смотрите сами: miidas.ru. #ии #бизнес #бухгалтэрия

@arlogilbert @NousResearch Exactly :))) @AdolfoUsier

@NousResearch I love Hermes. Long time user, but these posts are more confusing than helpful. 60% faster? How? All providers? Measured by what? Again, I love Hermes but these updates aren't helping Hermes because they lack important context for those who are not contributing to the code.

Hermes Agent now reads the web up to 60x faster and 49x cheaper. Scraping backends pass clean content straight to the agent without redundant processing steps; large pages are saved locally and paged on demand so you get the same quality at a fraction of the time and cost.

We are entering the era of AI agent mobile collaboration. @cursor_ai has announced an iOS app. @claudeai now has the Tag feature - so that you can talk to Claude in Slack groups. And I've been using @opencrabs to talk to an #aiagent in #telegram groups for a month now.

@ALTIC_DEV @sdhilip My problem with dictation is my domain slang. I can't avoid using it - but the stock Apple dictation system doesn't understand it. Have you solved it? It this a system that would, for example, learn from my corrections?

@sdhilip This genuinely made my day :,)) Posts like this make last 4 months of sleepless nights, and research to give the best output worth it.

I've dictated almost everything for 6 months with Wispr Flow. 44,414 words, 161 wpm, top 0.1% of users. Last week I tried FluidVoice. Open source, runs local on my Mac, corrects as I speak with no API key, and handles slang better than I expected. Cancelled my paid plan. If you're on a Mac, this one's for you: altic.dev/fluid @ALTIC_DEV

I started skipping @opencrabs versions bc: 1. The system is much more stable now and there are no "the roof is on fire" updates. 2. @AdolfoUsier has been pumping new versions at an increased speed - so that it's hard to keep up! Great job, man!!!

@AgiMaulanaXyz Turned out I was flagged by #github for suspected abuse. Not really sure why. Filed a reinstatement request. The support is taking their time for the 2nd day.

@leshchenko1979 It is happening again right now. I cannot see the pr and the github action isn't working.

Anyone experiencing #github quirks today? My issues and PR's on other people's repos stopped showing. #help #whyme #when_will_this_finally_end

@Alibaba_Qwen Seems like the approach used needs to be inserted into my review prompts even before I lay my hands on this model: - Analyse input - Analyse startin environment - Try to simulate hiw the environment will change - Assess blast radius - Report

📣📣 Meet Qwen-AgentWorld — a native language world model that simulates 7 agent environments (MCP, Search, Terminal, SWE, Web, OS, Android) within a single model. Environment modeling is the training objective from day one, not a post-hoc adaptation. 🤔 LLMs are trained to be better agents — better at acting in environments. But nobody has trained them to model the environments themselves. 🗺️ Our roadmap: investigate how language world modeling can push the boundaries of general agent capabilities, along two routes: 1️⃣ Build a foundation model for environment simulation — outperforming Claude Opus 4.8 and GPT-5.4 on AgentWorldBench 2️⃣ Investigate how world modeling enhances agent training: 🔬 Controllable Sim RL (agentic RL with LWM as environments) surpasses training in real environments 🧠 Learning to predict environments (LWM warm-up) makes agents stronger — remarkably, even without any agent-specific training, this predictive knowledge transfers to agentic tasks with zero fine-tuning 📑 Paper: arxiv.org/abs/2606.24597 📖 Blog: qwen.ai/blog?id=qwen-a… 💻 GitHub: github.com/QwenLM/Qwen-Ag… 🤗 HuggingFace: huggingface.co/collections/Qw… 🧩 ModelScope: modelscope.cn/collections/Qw…

@astrodotbuild @opencrabs

Astro 7 is here! A new Rust compiler, a new Rust Markdown/MDX processor, Vite 8 and more. Get ready for 60%+ faster builds. astro.build/blog/astro-7/

@TheNJDevOpsGuy @opencrabs - light, fast and cooperative. Have you heard?

What have you chosen for your Agent Harness at this point? Opencode? Claude Code? Codex? #agenticai #ai