Everyone is becoming a one-person CEO. The game is leverage, and AI agents are becoming the new operating layer. OpenClaw felt exciting because it lets me run a 24x7 assistant on a VPS that keeps working in the background while I sleep!
I am building this assistant to run my personal life and support a wide variety of tasks. Right now, I have an army of mini-assistants doing the following jobs:
- Morning Digest: Curated AI news and key articles, delivered at 8 AM with my coffee. It can also send a voice note when I want to listen on the go, and optionally generate an explainer image on request.
- Learning Companion: Keeps my learning from drifting. It tracks what I am studying, surfaces relevant notes, and keeps my deep-dives active.
- Blog Assistant: Helps me brainstorm and outline blog posts. I still finish the writing, add images/videos, and publish. I don’t want fully AI-written posts; I prefer keeping a human touch.
- Task Assistant: Connected to my Notion workspace to capture tasks and remind me about what matters for the week.
- Knowledge Assistant: Has read access to my Notion knowledge base so I can ask questions quickly. It is basic right now, but I plan to push it toward deeper research.
Buildathon upgrades I added recently
In the recent buildathon, I added two useful upgrades to the assistant:
- Audio generation for morning briefings, so updates are easier to consume while moving.
- Explainer image generation, so dense topics can be visualized quickly.
How the stack looks today
- Infra: 2GB RAM machine on DigitalOcean for $10/month.
- Core model: Minimax 2.5 via OpenRouter for most knowledge-work tasks.
- Usage and cost: Burned through 30M tokens in the last 2 weeks for around $4.
- Media capabilities: OpenAI TTS for voice notes and Gemini 2.5 Flash (Nano Banana) for image creation.
What is coming next!
- Deep research on any topic (leveraging OpenAI and Claude Deep Research APIs where useful)
- Reminders over phone call
- Video generation with voiceover from prompts and screenshots
- Idea Factory: research product ideas and store them
- App Development Robot: Pitck prioritized ideas from Idea Factory, build web or Android prototypes overnight (on Linux), and deliver morning updates with screenshots, videos, and APKs
Core capabilities I still need to invest in
- Better video creation pipelines
- Voice-instruction interface
- Agent having its own GitHub and deploy environment access
The future is really exciting. It is time to build!!