This represents what may be the first deep integration of a human cognitive framework into a Large Language Model.
What makes this different:
Most AI development focuses on making models more capable—better answers, faster responses, more knowledge. This integration takes a different approach: teaching AI to understand human cognitive states.
The Assess-Decide-Do framework isn't information about productivity. It's a model of how humans move through thinking processes. By integrating this model into Claude, we've created something new: an AI that detects and responds to where you are mentally, not just what you're asking.
The prevailing AI paradigm:
- Make AI smarter
- Automate human work
- Replace cognitive labor
- Measure by output volume
This integration's paradigm:
- Understand human cognition
- Augment human thinking
- Support cognitive processes
- Measure by flow quality
The difference matters. When AI understands your mental state—are you exploring, committing, or executing—it can provide cognitive support instead of just task completion.
Users report something we didn't anticipate: Claude feels "weirdly empathic" with ADD integration.
This isn't anthropomorphizing. It's cognitive alignment creating the experience of being understood.
What's actually happening:
- ADD-aware Claude detects you're stuck in analysis paralysis
- It gently acknowledges the pattern without judgment
- It supports transition to decision-making when you're ready
- You experience this as "being understood"
The insight: When tools match human cognitive patterns instead of fighting them, interactions shift from transactional to relational. The barrier between human and tool softens.
This is early evidence of what human-centered AI can feel like.
Large Language Models excel at pattern recognition in text. But understanding human mental states requires something more: models of how humans think.
This integration is a small step toward that future:
Current state: LLMs process requests This integration: LLM recognizes cognitive states (exploring vs. deciding vs. executing) Future possibility: AI that understands human mental models deeply
The Assess-Decide-Do framework is simple—three realms. But it demonstrates a principle: AI can learn to recognize human cognitive patterns when given an appropriate model to work from.
As frameworks become more sophisticated, AI's ability to understand and support human thinking will deepen. This is one early experiment in that direction.
The ADD framework emerged partially from managing ADHD-related challenges. Realm separation reduces cognitive load:
- In Assess: No pressure to decide or execute—just explore
- In Decide: No content editing, no execution—just commit
- In Do: No re-evaluation—just complete
Each realm has one job. This reduces overwhelm.
When Claude operates with ADD awareness, it respects these cognitive boundaries. It doesn't push you to decide when you're assessing. It doesn't pull you back to research when you're executing.
For ADHD minds (and many neurotypical ones), this reduction in cognitive friction is significant. It's not just "better AI"—it's more cognitively humane AI.
Most discussions about AI center on capability: What can AI do? How fast? How well?
This integration asks different questions: How can AI enhance human capability? How can it support rather than replace? How can it augment cognition instead of automating it?
The difference in practice:
Task automation approach:
- User: "I need to write a blog post"
- AI: [Writes the blog post]
- Result: AI did the work
Cognitive augmentation approach:
- User: "I'm exploring blog topics" (Assess)
- AI: Supports exploration without decision pressure
- User: "I'm committing to this angle" (Decide)
- AI: Supports decision-making and structure
- User: "I'm writing now" (Do)
- AI: Supports execution and completion
- Result: Human did the work with cognitive support
The second approach keeps humans in the creative loop while reducing cognitive friction. The human remains the thinker, decider, and creator. AI becomes the cognitive support system.
For AI development:
- Demonstrates value of integrating human cognitive models
- Shows path beyond pure capability enhancement
- Suggests framework for human-AI cognitive alignment
For productivity thinking:
- Challenges pure task-completion metrics
- Emphasizes balanced cognitive flow
- Reframes AI as cognitive support, not replacement
For human-computer interaction:
- Shows tools can understand mental states
- Demonstrates relational quality from alignment
- Points toward more empathic technology
For accessibility:
- Provides ADHD-friendly cognitive support
- Reduces overwhelm through realm separation
- Makes AI collaboration more humane
Not claiming:
- AI has consciousness or real empathy
- This solves all AI alignment problems
- Cognitive frameworks are the only path forward
- ADD is the "best" or "only" framework
Actually claiming:
- Cognitive framework integration is possible and valuable
- AI can recognize human mental states when given appropriate models
- Cognitive alignment creates better human-AI collaboration
- This approach enhances humans, not just AI
This integration is an early experiment, not a finished solution.
What we're learning:
- How to teach AI about human cognitive states
- What cognitive alignment feels like in practice
- Where framework integration adds value
- What users experience as "empathic" AI
What remains unknown:
- How this scales to more complex cognitive models
- What other frameworks could integrate similarly
- Whether this generalizes across different AI systems
- Long-term effects on human cognitive patterns
The repository is open source. The framework is documented. The experiment is public.
We're discovering together what happens when AI learns to understand how humans think.
This matters because: For the first time, we're not just making AI more capable. We're teaching AI to understand human cognition. That's a fundamentally different kind of progress.
And the early evidence suggests: When AI understands where you are mentally, collaboration transforms from transactional to relational.
That transformation might be more important than any single AI capability improvement.