You bring the context. Your goals, patterns, psychology. The human side of the symbiosis.
Works with ChatGPT, Claude, Gemini, and Claude Code
"You said 'let's reorganize' 3 times this month. Nothing shipped between any of them. What are you actually avoiding?"
One short walkthrough on good and bad ways to use AI, plus the Ilya Sutskever clip that points at the bar: the best AI should feel more like a great teacher than a slot machine.
Foundational explainer
A quick framing for the line between using AI to think better versus using it to drift, outsource judgment, or stay entertained.
Signal clip
A 16 second cut where Ilya points to a more demanding use case: AI that helps you see yourself more clearly, not just answer faster.
"The best meditation teacher" is a much stronger product standard than "a useful chatbot."
That is the direction this project argues for: reflective, pattern-aware AI that can challenge you, not just comply with you.
Same person. Same AI. The only difference is context.
I'll record the video tomorrow. I just need to prepare a bit more.
Sounds good! Take your time to prepare. When you're ready, I can help you with an outline or talking points. Let me know!
Let's reorganize my project structure first.
Sure! Here's a suggested file structure for your project...
The video was never recorded. The project was reorganized 3 times.
I'll record the video tomorrow. I just need to prepare a bit more.
You've said "tomorrow" 12 times across sessions. Here's the data:
Jan 6: "tomorrow I will post it" -- posted 4 days later
Jan 9: "I will record tomorrow" -- never recorded
Jan 26: "tomorrow I will ship" -- shipped 3 days later
Feb 14: "tomorrow I will record" -- still not recorded
The script is done. Camera's there. Why not in the next 30 minutes?
...you're right. Recording now.
The video was recorded that day. It got 29K views.
Real data from 878 sessions. Read the full analysis
Real scenarios from real sessions. Click through each one.
Founders
"It caught me building features instead of talking to users. Again."
Creators
"8 weeks consistent. My longest streak. The agent noticed I batch-create better."
Developers
"Claude Code used to forget everything. Now it knows my patterns across sessions."
Your agent reads these at the start of every session. Personality stays stable. Tasks update in real-time. Just markdown.
SOUL and USER rarely change. NOW changes every session. Separation means the agent never forgets your personality even as your tasks evolve.
Pick a starting point for each file. Customize it, make it yours. The agent adapts from day one.
Real people using this daily. Open source and growing.
Proposed stripping the system to core accountability: track commitments, surface patterns, stop projects from fading.
Added dependency tracking between tasks. Integrated as [blocked by:task] syntax with warnings.
What actually happens when you run this system for months. Real data, real patterns, no hype.
86 days. 40,417 messages. Queried the database. Found loops, "tomorrow" patterns, and the 30% productivity tax.
Not your data. Your mind. Goals, fears, patterns of self-sabotage. What happens when it remembers everything?
The AI compiled a list of psychological bugs from watching me repeat patterns. I didn't write them. It noticed.
Core system, skills, and commands. All open source.
Works everywhere. Pick your platform.
Full 4-file system with skills and commands
curl -fsSL https://raw.githubusercontent.com/lout33/symbiotic-ai/main/install.sh | bash No install needed. 2 minutes.
Paste instructions into ChatGPT, Claude web, or Gemini. Same philosophy, zero setup.
View all platform guides →Pick an archetype, chat for 2 minutes, and get a usable workspace.