A mode-based system for thinking with AI — a concept
A personal concept for using AI to assist deeper thinking — structured into modes so ideas output.
Jul 11, 20265 min read
Alex M. Chong – Director of Product Design & UX Career Coach
The problem
Projects fail before they begin
Most projects don't fail in execution — they fail in the thinking that precedes it. Incomplete thoughts, mode-switching mid-session, and ideas that never land anywhere produce a weak foundation. And a weak foundation means the project shows it from the start. The three patterns that cause this:
The drift
Modes bleed into each other. Ideating flips into executing. Each mode needs different rules — mixing them collapses the thinking before it completes.
The scatter
Jumping between threads without completing any of them. The thinking stays shallow — broad but never deep enough to produce something worth building from.
The loss
Good thinking buried in a sea of chats. No artifact, no trail, nothing to hand off. The project starts from fragments instead of a foundation — and shows it throughout.
The system
A deep thinking infrastructure
My system puts guardrails around thinking so it can be done with intent and depth. A master file reads every message, assembles the right context, and routes to the right mode — silently.
Modes
How I'm working
Generative
Ideating Conceptualizing Discussing
Structuring
Planning Specifying Re-aligning
Execution
Building Project managing
Refinement
Reviewing Iterating Editing
Coaches
Who I'm talking to
Life coach
Priorities, intentions, north star
Learning coach
Skills, growth, deliberate practice
Underlying Principles (always active)
How to work with me. The ground everything operates on.
Communication styleWorking preferencesSession rules
How a session assembles:
Chat begins
"idea..."
↓
_master.md assembles context
alex-principles.md
_principles.md
_context.md
loads every session
↓
generative.md routes the session
Ideating →
Conceptualizing
Discussing
Planning
routes the session
↓
Chat session
think together
↓
idea output
01_idea.md
artifact produced
The experience
I arrive. The system takes care of the rest.
Chat is the mechanism. The system asks questions, challenges assumptions, and shapes the extraction — until something worth keeping emerges.
① Start with a prompt
"I'm trying to work on an idea — I keep struggling with how I work with AI, every session I have to re-explain myself and I always end up somewhere I didn't intend to go, I think I need some kind of system but I'm not sure ..."
② Think & discuss: within a mode
Mode of thinking
Ideating →
Conceptualizing
Planning
Specifying
↓
Thinking & Discussing
"I keep losing the thread mid-session..."
What does "losing the thread" cost you?
"thinking never completes, it just drifts"
What if each thinking type had its own rules?
"Oh, maybe we can base it on modes of thinking?"
↓
The Output: An Idea
01_idea.md ↓
③ The output: 01_idea.md
01_idea.md
A file-based thinking infrastructure that separates different types of thinking into distinct modes — each with its own rules, goals, and guardrails. A master routing file reads every message and assembles only what is needed. The system ensures that thinking actually completes: generative thinking stays generative, structuring stays structuring, and nothing bleeds into execution before the foundation is ready.
The insight: the problem isn't the thinking — it's that different types of thinking need different containers.
↓
Ready to take this idea →
Conceptualize mode
The handoff
Thinking assembles. Then completes.
Each session produces an artifact. As the artifacts accumulate the thinking deepens. When all pieces are in place, they package into a single project brief — the foundation a project can actually be built from.
Thinking completed
✅ 01_idea.md
✅ 02_concept.md
✅ 03_plan.md
✅ 04_spec.md
↓
project-brief.md
idea · concept · plan · spec
→
↓
thinking complete
Project begins
alex-projects/
project-brief.md
intent · concept · plan · spec
No system files. No master. Just the brief and what gets built from it.
Could be anything
📱 Apps & tools
✍️ Articles & writing
🔧 Products & systems
The structure
Two folders. One purpose.
The thinking system and the projects folder are intentionally separate. Thinking happens on the left. Building happens on the right. The brief is the bridge.
🗂️ alex-thinking-system/
brief
→
📁 alex-projects/
The file system is the system.
Why not just use skills or projects?
Claude's projects and skills features are useful — they evoke upon trigger, similar to the system shown here. The challenge comes from what happens after.
Sessions accumulate without structure, and output lands in erratic formats: sometimes it's a thought held in your head, sometimes a long response to scroll back through, and even when you're organised enough to export a markdown file, now you have to manage that file yourself. Where does the concept go? Where does the plan live? What connects them?
This thinking system sidesteps that entirely by owning the file structure from the start — working within something like VSCode so that the interface outputs directly into your own organised workspace, in a structure that maps to how the thinking actually happened.
To close
A different relationship with AI
This isn't a productivity system. It's a thinking system built around amplifying the depth of thinking I'm already doing.
The default relationship with LLMs is you directing it — prompt in, it outputs. This thinking system reverses that direction: the LLM holds the conditions that push the thinking deeper. The output isn't what the model produced, instead it just packages where I arrived at.
The 0-to-1 space is where everything consequential about a project gets decided. When that thinking scatters or drifts, everything downstream inherits its weakness. This thinking system exists to protect that generative space — so that deep thought amounts to a solid foundation to build a project on. Specifications that guide. Decisions that hold. Work that knows why it exists. Without that rigour, we're just vibing.
