I Built a Text File That Gives AI Assistants Persistent Memory

#ai #productivity #machinelearning #chatgpt

Post-launch update (2026-04-19)

A few readers flagged that this post overclaimed the file's contents. That's fair. The original post described modules that weren't in the actual v2.30 product, described the Options Engine as a "validated iron condor" when the real statistical result is a thin and not-significant edge (OLS p = 0.072), and listed an early-adopter price of $49 that is no longer accurate.

This version is the honest rewrite. The product is the same file; the description now matches it.

The problem

If you use AI assistants for anything beyond one-off questions, every new session is a blank slate. You re-brief the model every time. Vendor memory features (Claude Projects, ChatGPT Memory, Gemini saved info) help, but each one only works inside that one vendor. Switch to another model and your context doesn't follow.

What the file actually is

ContinuityFile is a ~25,000-word .txt file you paste (or upload) at the start of any AI session. It's portable because it's plain text — Claude, ChatGPT, Gemini, and local models all accept it.

It's a structure, not an algorithm. The insight that made it work (after three prior attempts that decayed into contradictory noise within a few months) is "modules + dated patch log." Every rule change is recorded as an explicit supersedence at the bottom of the file. Without that, any living context document rots.

The eleven modules (v2.30)

Core:

I. System Kernel — identity, operating rules, patch protocol
II. Execution Engine — task triage, planning, SOP framework
III. Professional Operations — meetings, decisions, follow-through

Optional (load when relevant):

IV. Health & Longevity OS
V. Behavioral Systems
VI. Personal Life OS

Core continued:

VII. Financial Framework — personal-operator scale, not advisor-grade (explicitly not a substitute for a CPA/CFP on complex tax or retirement work)
VIII. AI Install & Maintenance Manual — how to install and patch the file itself
IX. Module Selection Guide — which modules to activate for which session types

Optional:

X. Options Engine — an iron condor framework with explicit statistical honesty: historical OLS p = 0.072, Newey-West-adjusted p = 0.149, and 79.5% of historical CAGR came from a single five-month window (Aug–Dec 2020) on n = 30. The edge is not statistically significant. It's included because I use it; it's optional and the disclosure is part of the module.

Core:

XI. Template Library — meeting templates, decision log, SOP scaffold, patch entry

What it is not

Not a prompt template (prompt templates are per-task; this is persistent across tasks). Not a knowledge base (rules and frameworks, not facts). Not advisor-grade finance (the finance module is decision logic at personal-operator scale). Not a novel algorithm (the insight is that plain text plus explicit supersedence beats vendor-specific memory for portability).