Загрузка видео...
Не удалось загрузить видео
Claude Desktop is becoming a full operating system. Every day Claude is releasing a new feature that makes it easier control AI agents to get work done. Here's every feature they've released in the past 3 months: TIMESTAMPS: 00:00 Intro 01:17 Evolution from Chat to General Agents 07:48 Cowork... show more
57,965 просмотров • 6 месяцев назад •via X (Twitter)
Комментарии: 26
The scheduled tasks + dispatch features are underrated. I run Claude Code on a Mac Mini 24/7 with cron jobs, heartbeat polling, and 50+ custom skills. It handles email triage, deploys code, monitors 5 products, and posts social content autonomously. The "full operating system" framing is right. Once you add persistent context (AGENTS.md, SOUL.md, memory files), it stops being a chatbot and starts being infrastructure. Anthropic is quietly building the runtime layer that makes everything else work.
the convergence of desktop app and AI agent is wild. once you have accessibility API access instead of just screenshots, your agent can actually understand UI structure natively. way more reliable than pixel matching
How are you handling the integration complexity? I'm building AI features into my SaaS and finding that each new capability creates 3 new edge cases to debug.
The goal was always to build an AI OS, when you think about it am operating system is basically a browser file explorer and apps, but right now these things are all still separate which is why I built KEB OS to unify these things correctly
is temporarily offering 50% off Opus 4.7 and also gives 5$ credits per month. Its supposed to be websites only tho, but i asked it to make a discord bot and just downloaded the zip :fire: Anyway heres my totally not referral link
Already there. Running a full AI OS on top of Claude right now - skills, memory, automation, scheduling. It's not becoming an OS. It already is one if you build the layer.
still waiting for claudebook
yep they are shipping like crazy!! looking forward to this one
post to yt
Claude did become extremely agentic over the last couple months. It’s actually been so fun because it has access to my notion and my computer and now it does so much more moving of things around to the appropriate places in my computer freeing up my time.
Sounds ambitious, but turning something like Claude into a full OS raises questions about usability and security. What’s the endgame here?
Chat to agents to plugins to computer control. That's not a feature list — it's a platform eating everything around it. The remaining question: does the default agent surface end up on the desktop or in messaging? We're building for messaging at
clown operating system go brrr
Nice video!
Looks like they're betting big on transforming the way we interact with AI. I keep wondering if these new features actually solve real user pain points or if it’s just flashy updates.
I'm curious which feature changed your workflow most. For me it's dispatch + computer use.
The cost finding is the part nobody's talking about Code policy outscored GPT-5.2-High. GPT-5.2 experiments cost ~$640. Code policy cost basically zero. how many enterprise AI use cases are actually just follow rules reliably at high volume Because if that's most of them, the economics of frontier models just got a lot harder to justify.
Not sure a feature flood means it's a true OS. Sounds more like they're throwing everything at the wall to see what sticks. Can’t help but wonder about usability amid all this “evolution.”
But only if you don't work for more than 10 minutes at a time. Let alone during working hours!
dms :)
Good summary if you have been living under a rock the last couple months. Wish I watched another video with my time that taught me something other than news.
been running this exact stack via OpenClaw for months lol. cron jobs, persistent memory, skills, the whole thing. claude desktop is catching up to what the terminal-first crowd already built. the scheduled tasks + dispatch combo is where it gets real though, thats when it stops being a chatbot
i love the edits on the intro - specially the little crowns lol
They know what's coming, OpenAI's version will be better at everything
Nice video that perfectly sums up Claude’s latest developments! It’s incredible to see how capable Claude is becoming and how insanely fast Anthropic is shipping. The progress is unreal! 🚀
the dispatch + scheduled tasks combo is quietly the most important thing here — that's when Claude stops being a chatbot and starts being infrastructure you actually rely on
