The Human Stack is a personal knowledge system that helps you think, create and work with AI. Five layers give your knowledge a home you control, with a way for AI to find the context it needs.
A useful idea sits in a meeting note. A project decision is buried in an old conversation. Your way of working lives mostly in your head. When you need that knowledge, you have to find it or explain it again.
The Human Stack gives that material a place to live. You can capture experience, connect ideas and develop knowledge you can use again. This is personal knowledge management, or a second brain, with AI helping you retrieve and apply what you know.
The same structure helps brief AI: who you are, how you work, what you are building and what matters now. You choose what becomes established knowledge and which information belongs in each task.
Identity forms the foundation. Methods, active work, current priorities and daily thinking build above it. The pyramid shows that relationship. Its widths don't measure improvements in AI output.
Identity | Who you are. Your voice, values, purpose and boundaries. This changes slowly, with your agreement. A short identity file helps AI follow what matters to you.
Systems | How you operate. Your frameworks, methodologies, intellectual property, standard processes. The accumulated knowledge of how you do what you do. When your AI knows your systems, it stops reinventing what you have already solved and starts building on it.
Projects | What you are actively building. Goals, decisions, constraints and current phase. Keep a briefing for each active project, and update it when the work changes.
Pulse | What matters now. This week's priorities, current decisions, capacity and blockers. Review it weekly, or when circumstances change. Keep detailed tasks in your task list.
Thinking | Your capture and creation stream. Journal entries, insights, meeting notes, learning and drafts. Search it when needed. Useful patterns can become methods in Systems. Changes to Identity need your agreement.
Your second brain holds more knowledge than AI needs for one task. The Interface tells AI where to start and how to find the relevant parts. You can use the knowledge system with or without AI. With AI, access and loading need to be set up and tested for your chosen tool.
Load. The agent reads AGENTS.md for working instructions, then ME.md, INDEX.md and NOW.md for identity, routing and current priorities.
Route. The Systems routing table tells the agent which frameworks, tools, and skills to load for the type of work being done. Writing? Load the voice constraints. Strategy? Load the operating system. The agent does not load everything. It loads what the task requires.
Focus. If the work is project-specific, the agent loads that project's briefing document. Now it knows not just who you are and how you work, but what you are building and where you are in the build.
Work. The session happens. Ideas surface. Decisions get made. Priorities shift. Items emerge that are not this week's concern.
Update. At session end, record agreed decisions and useful learning in their proper homes. Review changes that affect priorities, methods or identity. Keep current summaries consistent with their sources.
These files form the Interface: four shared entry files, plus CONTEXT.md for the project in hand. Thinking is searched on demand. A short CLAUDE.md adapter can point tools that need that filename to AGENTS.md, keeping the operating instructions in one place.
Written from the AI agent's perspective. This is what each layer gives the machine you are working with.
Identity gives me your voice. It helps me follow your values, voice and boundaries. I still need accurate facts and instructions for the task.
Systems gives me your methodology. The routing table tells me which frameworks to apply. When you say "plan this programme," I know which model to use, not guess. When you say "write this post," I know to check for your voice constraints. Your intellectual property becomes operational, not decorative.
Projects gives me your world. The briefing document for each project tells me what phase you are in, what decisions have been made, who the stakeholders are, and what constraints matter. Without this, I give you advice for a generic project. With it, I give you advice for yours.
Pulse gives me your week. The live state document tells me what is urgent, what is blocked, and where your energy is. It helps me account for your current priorities and capacity.
Thinking gives me your raw material. When I need to reference a past conversation, a meeting note, or an insight you captured three weeks ago, this is where I look. I do not load it every session. I mine it when the work demands it.
If you ask me to revise a workshop, I need its learning objective, your teaching standards and the current material. I may retrieve an earlier decision. Unrelated journals and client histories add material I do not need. A bounded task can succeed with a focused brief.
AI works with a limited amount of active context. More material doesn't automatically improve its answer. Read the shared instructions, load the relevant project and methods, then retrieve further sources as needed.
Persistent notes help later sessions recover decisions. Dates and source links help check what's current. Keep private material within its permissions, and verify the resulting work. Your knowledge can travel between tools, but each tool needs a tested way to access and use it.
This approach aligns with Anthropic's context engineering guidance on focused context, retrieval and persistent notes. That supports the mechanism, rather than proving this particular five-layer model.
Within Amplify OS, short for The Amplified Operating System, the Human Stack holds knowledge and context. The Capability Stack includes AI, technology, tools and workflows that put relevant knowledge to work. The Interface connects them.
What you organise. Five folders of you: Identity, Systems, Projects, Pulse, Thinking. Lives in your existing knowledge tools. Notion, Drive, Obsidian, files anywhere.
How AI reads it. A small protocol of system files (AGENTS.md, ME.md, INDEX.md, NOW.md, plus CONTEXT.md per project). Instructions for loading and using your knowledge in a connected AI tool.
The AI, technology, tools and workflows used to do the work. Models, apps, integrations and tested workflows put relevant knowledge to use. Prompts, skills, plugins and schedules are forms within it.
Keep a source you own and can export. Plain Markdown files make the knowledge easy to inspect. Connected services have their own access and export limits.
Thinking becomes active, not reactive. Ideas you had years ago resurface because the system remembers what you forgot. Client calls feel different because you know everything is captured, prioritised, and waiting for you when you get back. Projects stop living in your head. The cognitive load drops. And the AI stops being a tool you manage and becomes a partner that operates inside your world.
The pyramid shows how identity can inform your methods and work. The funnel shows how ongoing capture and creation keep that knowledge useful. Turn the pyramid over: Thinking is at the wide top, Identity at the point.
What you capture and create each day lands at the top, in Thinking. Most of it stays there. Useful ideas can inform this week's priorities, a project or the way you work. A note can connect directly to where it belongs without passing through every layer. Very little reaches Identity, and it only gets there when you decide it should. What changes fast sits at the top. What lasts settles at the point.
The system stays useful through capture, review and correction. Here is the maintenance rhythm configured for my own Human Stack. Schedules help with recurring work, and still need checks that they ran successfully.
Every weekday morning. Open tasks carry forward to today's daily note. In-progress work keeps its state. I choose what belongs in Today, and task lines keep their wording and working-note links.
Three times each weekday. Meeting notes are imported into Thinking with project links. Private material stays within the appropriate access boundaries.
Monday morning. A check-in brings together the week's commitments and asks about capacity. I confirm the priorities before NOW.md changes.
Friday morning. Insight candidates surface from the week's notes. Old journal entries are sampled for buried insight. Unlinked meetings get auto-linked to projects and frameworks.
First Monday of the month. A three-pass diagnostic sweeps the stack. Indexing health, mining surface, memory hygiene. Findings only. The findings identify what needs attention. Corrections still need to be made.
Capture gives you material. Review helps you decide what deserves to last. Check your current summaries against their sources, especially after a decision changes.
Use a short identity file and a brief for one active task. Check whether AI followed your boundaries, used the right facts and produced something useful. Correct what it missed, then add context where the result shows a need.
Find Your Starting Point →Personal knowledge management helps you capture, connect and reuse what you know. The Human Stack brings that purpose together with a way to brief AI from the same sources.
Save ideas, notes and learning where you can find them again. Connect related material and develop useful patterns into methods. Keep your sources available.
Use the five layers to separate lasting knowledge from changing work. Brief AI with relevant sources, retrieve more when needed and record agreed changes.
Amplify OS is short for The Amplified Operating System. Human-First and the two circles lead: Why and Who, with How in the overlap. Human Stack and Capability Stack make that direction practical.
The Human Stack gives your knowledge a practical structure. Five folders hold the material, and the Interface tells AI what to read. You retain ownership of the sources and judgment over what changes.
“I built the Human Stack because I was tired of watching AI produce average work for above-average people. The technology was never the bottleneck. The context was. Five layers of who you are, organised so any AI can read them. That is all it took to go from generic output to genuine partnership.”Colin Scotland
Start with a short Identity file and one real task. Check the result, correct what AI missed and build from there. The Human Stack Builder writes your first three files in the browser. In the Accelerator, we build all five layers together.