Анвар Бакиев

Все говорят: учи вайбкодинг, иначе останешься за бортом. Это неправда. И я объясню почему. Вайбкодинг — инструмент для одного типа людей: разработчика, которому нужно быстро прототипировать. Для всех остальных он решает не ту проблему, хоть сейчас это и не кажется очевидным. Настоящий разрыв в навыках 2026 года — не "умеешь ли ты промптить код", а знаешь ли ты, что уже существует? Например, на GitHub и HuggingFace прямо сейчас лежат бесплатные инструменты, которые заменяют дорогие агентства, $50K консалтинг и целые команды специалистов. Большинство людей о них никогда не слышали. Несколько примеров — только для маркетологов, это один из многих случаев: 1. Хочешь знать что думают 50,000 клиентов? cardiffnlp/twitter-roberta-base-sentiment — анализ тональности тысяч отзывов автоматически. На русском, казахском, английском одновременно. 2. Хочешь знать какой рекламный канал реально принёс деньги? Google и Meta открыли свои модели Media Mix Modeling. lightweight_mmm и Robyn. Бесплатно. Раньше это стоило сотни тысяч в консалтинге. 3. Хочешь доказать что кампания сработала — а не просто совпала с сезоном? causalimpact даёт байесовский ответ. Разница между "кажется сработало" и "точно сработало". 4. Хочешь сгруппировать 100,000 ключевых слов по смыслу? sentence-transformers/all-MiniLM-L6-v2. Один импорт. Это не программирование, не вайбкодинг. Это грамотность. Не нужно писать код. Нужно знать что инструмент существует, понимать зачем он, и уметь его запустить. И ещё одно — о чём почему-то никто не говорит: Читайте. Настоящие книги. Длинные статьи. Исследования. Понять почему люди покупают нельзя с помощью промпта. Это всё ещё человеческая работа. Цифровая грамотность в 2026 — это не "пиши код на Claude". Это знать что инструмент, который раньше стоил миллионы, лежит на GitHub бесплатно. И быть достаточно умным чтобы им воспользоваться.

Ребята, я делаю новый AI проект Extella - Multi agent platform & Corporate AI OS Строю отдел продаж с нуля. 5 человек. Что предлагаю: - Бери💰столько сколько можешь забрать - Enterprise kickers - Фаундер, который не будет мешать, а будет помогать продавать - Сильная команда с международным опытом Что ждем: голод, мозги, умение вести сложные B2B-сделки. 1 вопрос: ты хочешь продавать продукт, в который веришь? Пишите в ТГ @jj_zhans Алматы

@jcasti93sx @claudeai We can do it for you for free. Get me a message with your inquiry, I will try to build the needed integration for you

@claudeai Hey can we get one for AutoCAD and Revit?

Claude now connects to the tools creative professionals already use. With the new Blender connector, you can debug a scene, build new tools, or batch-apply changes across every object, directly from Claude.

@givros Ask me for detailed instructions, if you wish to test it someday. I will be happy to assist

@Anvario Wow it’s amazing !!

The new 2026 AI workflow 🤖⚡ GPT Image 2.0 → assets → Codex + GPT-5.5 → playable browser game 🎮 I turned my AI image into a web game. 🖼️ GPT Image 2 created the visual. 🎨 Codex generated the sprites and animations. 🕹️ GPT-5.5 built the prototype.

@cocktailpeanut Thanks for post! Want to found out more about it. We have made ease installation and tuning of AI4Animation with Extella x.com/Anvario/status…

AI4AnimationPy 1-Click Launcher --- Every frame is generated, not authored. An animation engine that uses AI to GENERATE every next pose in realtime, instead of human animators having to build frames manually. And it's FAST. Wrote a UI that runs on ALL machines, 1click launch.

@oliviscusAI We took it further — launched it with a single command, no terminal, no setup. Our AI agent wrote the code, fixed the shaders for Apple Silicon, and recorded this demo automatically. This is Extella running AI4Animation. One command. Zero code.

Meta just open-sources a tool that animates complete 3d characters in pure python. no heavy game engines. no hassle. just pytorch, numpy, and pure ai motion.

Here's what nobody talks about with AI tools: they're great at solving problems. They're terrible at remembering they solved them. Vibe coding tools, Claude, ChatGPT — they'll help you solve a problem once. You get the answer, close the tab, and next week you're starting from zero. Re-explaining, re-prompting, re-doing. Extella works differently. Every problem you solve becomes a saved expert — a permanent, executable system that runs the exact same logic on demand. Not a suggestion. Not a draft. A working system. A team of architects built a tool in a couple of hours that detects clashes between engineering layers in CAD drawings — the kind of thing a lead architect used to check manually to make sure no one placed a lamp over a ventilation shaft. The architects were shocked at how fast it happened. But here's the real part — that solution didn't disappear. No .py files scattered on a desktop. No deployment. No GitHub repo to maintain. It lives inside Extella — shareable with the whole team, ready to run again and again with one click. And the memory and rules that helped solve it in the first place? Still there. Still compounding. That's the difference. Other tools give you an answer. Extella gives you a capability. Other tools reset. Extella compounds.

Not a chatbot - orchestrator!

I’m not sure the “ladder” metaphor is fully accurate. It suggests that we move up and leave the lower steps behind. But in reality, we don’t remove the lower layers — we stack on top of them. You can’t really operate only at the top level. The lower layers still define constraints, performance, correctness, and ultimately what is possible. AI might raise the level of semantic expression, but it doesn’t eliminate the need to understand and work with the underlying layers — it actually makes that understanding more critical when things go wrong. So it feels less like a ladder, and more like a layered system where every level still matters.

@unclebobmartin interesting take but i think there's a difference between tools that follow explicit instructions and ones that actually reason about intent. the ladder metaphor works until you hit ambiguity

AIs are just another step up the semantic expression ladder. We initially expressed our semantics in binary, then assembler, then Fortran, then C, then Java, then Python, etc. AI is just the next step up that same old ladder. And when you take that step, nothing else changes. You are still expressing behavioral semantics. You still need to express structural semantics. All the old principles still apply. You still have to be concerned about design and architecture. And even though the syntax allows informal statement, you cannot abandon formalism. When you express behavior you need a formal way to enforce the behavior you want. I use Gherkin for this. It seems to work pretty well. Consider that Gherkin is written in triplets of Given/When/Then. Each of those GWT triplets is a transition of a state machine. A full suite of Gherkin triplets is a formal description of the finite state machine that represents the behavior of the application. Other formalisms that matter are things like module dependency graphs, testing constraints, complexity constraints, and many others. This step up the semantic expression ladder provides you with an enormous amount of options. But you'd better choose those options wisely!

@billtheinvestor Here how Extella can take control over applications. Just some minutes to educate it and Voilà

Claude 刚刚获得了真正的浏览器控制权……这将改变一切。不是截图。不是脆弱的选择器。不是缓慢的 MCP 循环。而是真实的浏览器代码。实时运行。它被称为 dev-browser。 它让 AI Agent 能够像开发者一样控制 Chrome。以下是大多数人忽略的范式转移： 不再是发明新的“Agent 语法”…… AI 直接编写真实的 Playwright 代码： goto() click() fill() evaluate() scrape() screenshot() 没有抽象层。没有奇怪的封装。只有代码。但真正的突破在于架构： • 在 QuickJS 沙箱内运行 • 对你的系统零访问权限 • 无文件系统访问权限 • 无宿主执行权限 拥有完整能力，且完全隔离。因此，现在 Agent 可以： • 打开网站 · 一次性登录 • 保持会话活跃 • 像人类一样导航 • 提取结构化数据 • 运行多步骤工作流 全部在一次运行中完成。最疯狂的地方在于： 现在一个 Agent 可以： → 打开 X (Twitter) → 滚动动态 → 提取推文 → 返回干净的 JSON 无需插件。 无需扩展程序。无需编排。那么性能如何？dev-browser： • 3分53秒 • $0.88 • 100% 成功率 碾压： • MCP 配置 • Chrome 扩展程序 • “浏览器技能”技术栈 核心思想： AI 使用浏览器最快的方式， 就是让它自己编写浏览器代码。我们正在进入一个新阶段： 能够“阅读”网页的 AI 能够“操作”网页的 AI 一旦 AI 能够可靠地控制浏览器…… 它就可以使用任何带有 UI 的软件。无需 API。无需集成。 无需权限。只需：打开 → 行动 → 完成。AI 同事不再仅仅是“思考”。它们开始“点击”。

@DanilPashkevic4 @billtheinvestor correct. Powerful tool for computer control and any job with the system

@billtheinvestor Extella has had this since the very release. I can do everything what i need: Scarping, parcing a huge amount of information in a matter of seconds A tool for a new era. Download extella.ai and enjoy.

Claude said it best himself: "Your pipeline with birefnet-general would produce significantly better results if run in an unrestricted environment." That unrestricted environment exists. extella.ai

Two real problems. One solution. Cowork: cloud sandbox — safe but restricted. Local without isolation: free but fragile. Extella: each Expert runs in its own isolated proprietary container. Full network access. Full filesystem. Zero conflict between Experts. ML models download once. Run forever. Your Mac stays clean.

I asked Claude Cowork to do a real task. Halfway through, it stopped and explained exactly why it couldn't finish. I didn't write this. Claude did. 🧵