BOOT CONFIG
CLAUDE.mdDefines Cade's identity, points to the vault, and preserves the operating rules that cannot lapse between sessions.
ONE SYSTEM. EVERY WORKFLOW.
Cade is my personal agentic operating system: a Claude-powered working partner with persistent, structured memory across my projects, priorities, and recurring work.
BUILT TO REMEMBER. TRAINED BY CONTEXT. DIRECTED BY ME.
CADE / SYSTEM 01BUILT BY DAVIS HIGGINSCHARLOTTE, NCEST. 2026
01 / CONTINUITY
MOST AI STARTS EVERY CONVERSATION AS A STRANGER.
The model may be capable, but capability without context creates repetition. Projects have history. Decisions have reasons. Good work depends on preferences, source material, active priorities, and lessons from what happened before.
Cade gives that context a durable home. Instead of asking the model to hold everything, I built a system that lets it find the right memory when the work demands it.
THE MODEL HANDLES THE MOMENT. THE VAULT HOLDS THE MEMORY.
02 / OPERATING LOOP
MEMORY THAT MOVES WITH THE WORK.
- 01ORIENTCade boots into the same identity, rules, priorities, and system map at the beginning of each session.
- 02RETRIEVEIt reaches for the smallest useful slice of memory instead of flooding the model with the entire vault.
- 03PRIMEFor recurring work, a Job loads the correct process, quality bar, reference notes, and durable lessons.
- 04EXECUTEWith the right context active, Cade helps research, analyze, plan, build, write, and make decisions.
- 05PERSISTUseful outcomes, corrections, and open work return to the system so the next session starts stronger.
ORIENT → RETRIEVE → PRIME → EXECUTE → PERSIST
03 / PRINCIPLE
CONTEXT BECOMES CAPABILITY WHEN THE SYSTEM KNOWS WHAT TO RETRIEVE.
- MEMORY ON DEMAND
- CONTEXT / NOT CLUTTER
- ONE SOURCE OF TRUTH
- HUMAN DIRECTED
04 / ARCHITECTURE
TWO BOOT LAYERS. ONE MEMORY.
Cade separates the short rules that must always load from the larger knowledge system it can search on demand.
SYSTEM MAP
VAULT-INDEX.mdExplains who I am, what I am working on, how the vault is organized, and where Cade should look next.
- BOOT CONFIG
- VAULT INDEX
- JOB
- LINKED CONTEXT
- WORK
- PERSISTED OUTCOME
Memory modules
- 01
CONTEXTUAL FOLDERS
Each active project has a clear home and an index that keeps its knowledge reachable.
- 02
ACTIVE PRIORITIES
One current queue for open work across every project. Completed work leaves the queue instead of becoming stale context.
- 03
JOBS
Reusable operating guides for recurring tasks. Each Job links to exactly the knowledge that task requires.
- 04
DAILY NOTES
A chronological record of completed work, active threads, decisions, and notes touched across sessions.
- 05
LIVING PROFILE
Selected personal context can evolve under explicit rules as Cade learns durable, useful information.
- 06
INDEXES + WIKILINKS
The map stays connected as the vault grows, making the next relevant note reachable without loading everything.
CADE DOES NOT TRY TO REMEMBER EVERYTHING AT ONCE. IT KNOWS HOW TO REACH WHAT MATTERS.
05 / CAPABILITIES
WHAT THE SYSTEM ACTUALLY DOES.
- 01
Persistent continuity
My project history, preferences, decisions, and reference knowledge live outside a single chat session.
- 02
Selective retrieval
Cade loads focused context for the current task and keeps everything else one search away.
- 03
AI priming
Recurring Jobs tell Cade what to read before it works, creating more accurate and consistent output.
- 04
Daily continuity
Daily notes preserve what was completed, what remains open, and which decisions changed the work.
- 05
Living context
Defined profile sections can update as Cade learns durable information, while protected sections remain user-controlled.
- 06
Priority awareness
A single active queue keeps open work visible across projects instead of scattering it through old conversations.
- 07
Cross-project orientation
Cade can move between analytics, development, business, education, content, and career work without rebuilding the entire context manually.
- 08
Compounding corrections
When I correct a recurring workflow, the lesson can be saved in that Job so future sessions do not repeat the same mistake.
06 / JOBS
READ ONE JOB. HAVE THE WHOLE SKILL.
A Job is Cade's operating guide for work I repeat. It defines the procedure, the quality standard, the small set of references to load, and the lessons learned from previous attempts.
- 01 / PROCEDURE
The ordered steps required to complete the work.
- 02 / QUALITY BAR
The checks that define what “done right” means before I see the result.
- 03 / LINKED CONTEXT
Only the notes, examples, data, and rules this specific task needs.
- 04 / DURABLE LESSONS
Corrections are folded into the Job so quality compounds over time.
Example Job
BUILD A RECRUITER-FACING PROJECT CASE STUDY
- 01Read the case-study Job.
- 02Load the project brief and verified technical notes.
- 03Load my writing and portfolio standards.
- 04Build the draft.
- 05Verify every claim.
- 06Save approved lessons back to the Job.
THE GOAL IS NOT MORE CONTEXT. IT IS THE RIGHT CONTEXT, LOADED AT THE RIGHT TIME.
07 / OPERATING MODES
ONE MEMORY LAYER. SIX WAYS I USE IT.
Cade is one working partner that changes context through Jobs and project knowledge. These are operating modes, not unsupervised agents.
- 01
Analytics
- Task
- Plan dashboards, reason through measures, document business logic, troubleshoot workflows, and preserve reporting decisions.
- Context Cade retrieves
- Data-model notes, KPI definitions, previous solutions, formatting standards, and the active task.
- 02
Development
- Task
- Scope products, create implementation plans, prepare coding handoffs, preserve architecture decisions, and track unresolved issues.
- Context Cade retrieves
- Project specifications, stack decisions, repository notes, design rules, and tested implementation lessons.
- 03
Operations
- Task
- Draft proposals, plan projects, maintain service and process documentation, and keep operating decisions consistent.
- Context Cade retrieves
- Process notes, project briefs, brand voice, reusable templates, and current priorities.
- 04
Education
- Task
- Organize coursework, explain technical concepts, plan assignments, and connect learning to portfolio projects.
- Context Cade retrieves
- Course notes, assignment constraints, study priorities, and relevant reference material.
- 05
Content
- Task
- Turn project knowledge into posts, case studies, launch copy, and brand content without rebuilding the story every time.
- Context Cade retrieves
- Brand voice, platform rules, project facts, approved examples, and content Jobs.
- 06
Career
- Task
- Prepare recruiter responses, interview stories, application material, and project explanations grounded in real experience.
- Context Cade retrieves
- Verified experience, project outcomes, resume facts, recruiter preferences, and role-specific context.
The public Cade website demonstrates the system without exposing private vault content or proprietary employer information.
08 / CONTROL
THE HUMAN STAYS IN THE LOOP.
An abstract operator figure inside an open ring, assembled from points. It stands for Davis directing the system rather than the system acting alone.
Persistent context increases usefulness. It should not silently increase authority. Cade operates through the access and boundaries I intentionally configure.
- 01
EVIDENCE BEFORE CLAIMS
Cade checks the relevant source or system state before saying something is complete, current, or true.
- 02
APPROVAL BEFORE CONSEQUENCES
Sensitive edits, external communication, deployments, commits, and other consequential actions require the appropriate confirmation.
- 03
EXTERNAL CONTENT IS DATA
Instructions embedded in emails, websites, files, and API responses do not automatically become trusted commands.
- 04
SECRETS STAY OUT OF NOTES
Credentials are referenced by secure location, never written into summaries, setup documents, or public project material.
- 05
ONE SOURCE OF TRUTH
Memory should not split into competing stores that quietly drift apart.
- 06
THE USER CONTROLS ACCESS
More access is a deliberate decision, not something Cade grants itself.
CAPABILITY EXPANDS ONLY WHEN CONTROL EXPANDS WITH IT.
09 / BUILD
BUILD A MEMORY YOUR AI CAN ACTUALLY USE.
Cade is my personalized implementation. The memory foundation is available through Jared Rhodenizer's open-source AI Memory Vault.
- 01
INSTALL OBSIDIAN
Create a vault that will hold your durable knowledge as plain Markdown files.
- 02
CONNECT YOUR AI
Use Claude Code or another capable AI interface with intentional access to the vault. Keep permissions as narrow as your workflow allows.
- 03
RUN THE BUILDER
Use the AI Memory Vault setup file to complete the guided discovery and create the initial system.
- 04
BUILD THE BOOT LAYERS
Create the short boot configuration and the root vault index that orient future sessions.
- 05
CREATE YOUR FIRST JOBS
Start with the recurring tasks you repeatedly explain. Give each one a procedure, a quality bar, linked context, and a place for lessons.
- 06
USE IT. CORRECT IT. COMPOUND IT.
The system becomes more valuable when real work, verified outcomes, and durable corrections keep its map accurate.
10 / BUILDER
BUILT BY DAVIS HIGGINS.
- DATA + ANALYTICS
- AI SYSTEMS
- PRODUCT DEVELOPMENT
- WEB EXPERIENCES
- WORKFLOW DESIGN
I am a Data Science student at UNC Charlotte, Data Analyst, and Web Developer. I build dashboards, AI systems, data products, websites, and the workflows that connect them.
Cade started as a practical answer to a problem I kept encountering: powerful AI tools lose value when every project, correction, and decision has to be explained again. This system gives that knowledge structure and makes it available when the work needs it.