ALPHA BUILD v0.1
Read Memo Launch Workspace
INTRODUCING INNERSCRIPT

Your writing remembers
more than you do.

InnerScript is a private diary and AI memory companion that turns years of journals, notes, relationships, ideas, and personal writing into source-backed self-understanding.

Begin Reflection Verify Cryptography
ARCHITECTURE Local-first SQLite
ENCRYPTION AES-GCM-256
AI ENGINE Source-Backed RAG
Pondering Socrates Statue with digital matrix overlay representing memory retrieval

INTERACTIVE WORKSPACE

Experience how streams of thought connect semantically to your past.

journal / 2026-06-05.md
EDITING
|
Is this related to team dynamics or product speed?
Self-Reflective Mirror
AI RETRIEVAL ACTIVE
GROUNDED REFLECTION QUESTION

Waiting for thoughts to be recorded...

InnerScript analyzes active entries in real-time without sending data to public clouds.

SOURCE MEMORIES FOUND (RAG)
Write in the journal to trigger semantic recall.

LOCAL MEMORY GRAPH (ACTIVE SEMANTIC RELATIONS)

Nodes: 124 | Links: 341 | Sparsity: 0.12
LOCAL STORAGE • AES-256 SYMMETRIC • ZERO CORPORATE CLOUD • P2P ENCRYPTED SYNC
LOCAL-FIRST
AES-GCM-256
EVALUATED

NOTICE: THIS WORKSPACE IS CRYPTOGRAPHICALLY SEIZED BY THE SOVEREIGN USER

All journals, notes, relations, and personal logs stored within this instance have been encrypted under absolute zero-knowledge security. Under the laws of local data preservation, no commercial platform, LLM web scraper, corporate server, or central hosting node has permission to access, store, or run processing pipelines on this database.

USER DATA SOVEREIGNTY

Data exists entirely on your machine. Databases are stored as simple, standard SQLite or plaintext Markdown, leaving no proprietary footprints.

SOURCE-BACKED AI GUARANTEES

Retrieval-Augmented Generation operates inside your privacy sandbox. When InnerScript claims a pattern exists, it points to a cryptographically validated local source file.

SECURITY PROTOCOL: LOCALHOST-ONLY SYNC | AES-256-GCM SYMMETRIC SESSION KEYS | VERIFIED COMPILE: SHA-256 VALIDATED

The Tragedy of the Silent Archive

You write for years. In paper notebooks, notes apps, private documents, random chats, and scratch drafts.

That writing contains your emotional curves, recurring fears, repeated habits, relationship histories, buried ambitions, contradictions, and solutions to problems you solved years ago but have now forgotten.

But because notes are stored in linear lists or disconnected piles, 99% of your journals become a silent cemetery. You almost never go back to read them. You repeat the same mistakes, ask the same unresolved questions, and suffer from the same alignment loops because you cannot converse with your own long-range memory.

the_problem_statement.md
Your memory is scattered Across DayOne, Obsidian, paper, and Apple Notes
No central semantic layer to link ideas over time.
AI wrappers want your data They send text to public APIs.
They lack a localized privacy boundary.
They generate generic, shallow advice ("therapist-speak").
No source verification AI assertions about your patterns cannot be verified.
If AI says: "You tend to feel anxious in winter," it must link the exact December entries to prove it.

Two Loops of Self-Understanding

InnerScript is built on a tight feedback loop that keeps software simple, safe, and actionable.

LOOP_01

The Immediate Feedback Loop

Daily writing that returns active, grounded reflections without long-term setup.

  • Write safely: A localized, distraction-free markdown outliner.
  • Save reliably: Instantly commit entries to localized encrypted database.
  • Immediate Reflection: As you write, receive exactly one grounded reflection question based *only* on the current stream.
LOOP_02

The Long-Term Semantic Loop

Turning years of archives into a source-backed mirror of the self over decades.

  • Bulk Ingestion: Import years of DayOne, Obsidian, or Apple Notes archives.
  • Local Chunking & Embedding: Compute semantic nodes locally in the background.
  • Semantic Search & Cite: Query your past in natural language and receive answers with precise sentence citations.
  • Pattern Detection: Discover emotional patterns, unresolved arguments, and habit triggers.

Tailored for a Serious Objective

InnerScript speaks transparently to builders, employers, and journalers without marketing fluff.

OPEN TO COLLABORATION

Let's build emotionally serious software.

InnerScript is looking for software engineers, product thinkers, and cryptographers who care about personal data systems, sovereign AI, and tools for thought.

We are building local-first sync systems, custom SQLite semantic search index structures, and evaluation harnesses to test AI grounding against highly subjective personal documents. If you care about building systems that will last a lifetime, we should talk.

Get in Touch
INTERESTS: CRDTs, SQLite, Local AI, Vector DBs
PHILOSOPHY: No ads, no data mining, sovereign files
LICENSING: Open source core (AGPL-3.0)
TECHNICAL PROWESS

An engineering depth review.

For recruiters and engineering leaders: InnerScript shows how we handle complex system engineering problems.

The application architecture implements custom chunking strategies (handling nested outliner lists), hybrid retrieval (BM25 token lookup merged with cosine similarity on local ONNX embeddings), strict data-leakage sanitization, and automated grounding validation.

Explore Repository
LANGUAGES: TypeScript, Rust, SQL
RAG STACK: SQLite-Vec, Minhash Deduplication
SECURITY: WASM Sandbox, Safe Tokenizers
HUMAN FOCUS

For those who write to think.

If you have folders of markdown notes or decades of physical diaries, InnerScript is built to give your words a voice.

Unlike therapist replacement apps, we don't diagnose you, give generic advice, or send notification badges to hook your attention. InnerScript acts as a quiet, offline-first study. It helps you see what you wrote so you can make decisions with clarity.

Try Workspace Sandbox
DESIGN: Minimalist, dark-mode, calming
UTILITIES: Export Markdown/JSON anytime
GUARANTEE: Works offline, forever

System Specifications

Technical blueprint of the InnerScript local-first memory companion.

01 / ENCRYPTION LAYER

AES-GCM-256 Envelope Encryption

Database pages are cryptographically protected on disk. Decryption keys reside inside your operating system's native keychain (Keychain Access / Secret Service), never in memory or configuration files.

02 / LOCAL RETRIEVAL

On-Device Vector Search

Runs lightweight, localized Transformer model tokenizers via WASM. Vector computation is 100% offline, keeping sensitive memory files off corporate GPU clusters.

03 / ARCHIVE INGESTION

Multi-source Importer

Ingests DayOne JSON exports, Apple Notes database copies, and directories of markdown files. Automatically deduplicates entries and parses historical timestamps.

04 / ALIGNMENT EVALUATION

Grounding Verification Harness

InnerScript features a local verification framework that scans generated reflections. If a reflection makes an assertion about you that cannot be backed by a highlighted sentence citation in your database, it is automatically flagged and filtered.

PROJECTMEMO // INNERSCRIPT.IN

Building emotionally serious software.

To whoever is reading this,

Most modern software is built to exploit your attention. Apps flash notifications, gamify basic habits, and mine your deepest memories to sell ads or train corporate base models.

I believe there is room for software that behaves like a quiet study. An archive that is secure, local, completely owned by you, and designed to help you think rather than keep you scrolling.

InnerScript is an exploration of this idea. I am looking for collaborators, system engineers, designers, and testers who want to work on a privacy-sovereign memory system.

If this resonates with you, let's build this together.

Joel
Creator, InnerScript

Submit a Collaboration Inquiry