Sander Saar, founder of AgentCDN, making videos work inside AI
Creating
Shorts
Your next customer is an AI agent
Cars in China are next level 🚙🇨🇳📈
Top 10 Tech Experiences in China - Future of Tech
AI Is Now The Interface: 3 Human Skills for the Agent Era #aiagents
3 levels to your "oh sh*t” moment with AI
92% goals fail - here’s a system that might help, inspired by Daniel Pink
My top 5 most-used AI tools you’ve probably never heard of
The best gift? Something people actually use.
Reviews
Related videos · 8 videos
Shorts · 8
Your next customer is an AI agent
Cars in China are next level 🚙🇨🇳📈
Top 10 Tech Experiences in China - Future of Tech
AI Is Now The Interface: 3 Human Skills for the Agent Era #aiagents
3 levels to your "oh sh*t” moment with AI
92% goals fail - here’s a system that might help, inspired by Daniel Pink
My top 5 most-used AI tools you’ve probably never heard of
The best gift? Something people actually use.
Apps
Analysis
Articles
3 levels to your "oh sh*t” moment with AI
Loading YouTube player.
3 levels to your "oh sh*t” moment with AI
Feb 13, 2026 · 1:41
The video outlines three progressive levels of AI adoption in the modern workplace. It covers using tools like Claude, ChatGPT, and Gemini for everyday browser and spreadsheet tasks, collaborating with systems like Claude Code and Codex to generate finished reports or software, and deploying autonomous agents like OpenClaw on dedicated servers as a proactive workforce. The presenter also reflects on the rapid pace of AI evolution over six months and its impact on future jobs.
Chapters
- 0:00
Level 1: AI as a tool in browsers and spreadsheets
The presenter introduces Level 1 of AI adoption, demonstrating how tools like Claude can navigate browsers, book reservations, and build spreadsheet analyses autonomously.
- 0:18
Level 2: AI as a coworker for code and knowledge work
The speaker explains Level 2 AI tools like Claude Code, Codex, and Claude Cowork, illustrating their ability to generate dashboards, apps, and codebase commits.
- 0:46
Level 3: AI as an autonomous workforce
The host highlights Level 3 autonomous agent setups using OpenClaw on a dedicated server to manage tasks, review schedules, and create workflows proactively.
- 1:17
Reflecting on the future of work and AI adoption
Reflecting on the rapid progression across all three levels in six months, the host discusses the changing nature of jobs and asks viewers which level they currently use.
Transcript
- 0:00
There are 3 levels to your oh shit moment with AI. Level 1 is when Claude is working inside your browser. No, not to chat with, but as an agent that navigates, clicks, fills, and completes tasks across your tabs. Or inside your spreadsheets. It creates the formulas, graphs, analysis. It reads the page, understands the context, and does the work. Level 2: Claude Code and Codex for coding, and Claude Cowork for knowledge work. Working on your actual files. You describe what you want, it creates a plan, uses all the tools needed like web, files, code, research.
- 0:29
You walk away and come back to a finished product. Features, build presentations, spreadsheets, reports created from nothing. I built a lot of apps recently, and 1 in 20 of public code commits on GitHub are now written by Claude Code. Cowork itself was built by Claude Code in under 2 weeks, and it triggered a $285 billion sell-off in software stocks. Level 3 is fully autonomous and proactive. I run Open Claude on Mac mini server. It connects to my email, calendar, notes, maps, GitHub, storage, web, everything.
- 0:54
But here's what's different. It doesn't just follow instructions. It observes my daily patterns, writes its own skills, and builds its own tools to solve them and executes. Completely proactively. It's the fastest-growing AI project on GitHub ever, and it's open source. I wake up to a morning briefing, analytics check, news scan 3 times a day, my emails and calendar reviewed, day planned, in addition to work completed overnight while I slept. All 3 of those levels happened in the last 6 months, and honestly, it's made me question a lot what I thought was my job.
- 1:23
If something can do it faster, more consistently, and cheaper, a lot of what I thought was my work just lost its meaning pretty fast. The question isn't whether AI can do your work. It's now that it can, what are you actually going to do with it? Let me know which level you're at. And if you'd like to deep dive in any of the setups.
