v0.1 Live AI built for the non-linear mind. Open cognitive scaffolding. GitHub • LinkedIn
Open-Standard Cognitive Scaffolding

AI built for the non-linear mind.

An open-standard library of AI skills and cognitive scaffolding. Designed for spatial thinkers, dyslexic professionals, and systems builders.

Input Friction
Zero Shorthand Tax
Brain dumps welcome
Formatting
Anti-Wall-of-Text
Tables, BLUF & anchors
Mechanics
Silent Polish
No lectures or interruptions
Voice Preservation
Authentic Tone
Zero corporate AI fluff
Live Cognitive Scaffolding

See D-Mode in action.

Compare how standard AI responds to a fast dyslexic brain dump versus how DxSkills compiles it into visual architecture.

Raw User Brain Dump (Speed-of-Thought) Fast Typing

"hey need to update client and dev team on the spatial analytics release. milestone 2 is pushed back to friday 5pm cause cloud provider had api outage yesterday. also remind everyone to chekc coordinate refrence systems and api keys before runing test pipeline. team sync is wednesday 2pm to 3:30pm and if anyone is blocked sooner just ping or call me directly"

Typos present ("refrence", "chekc", "runing")
Unstructured stream of operational facts
Cognitive State: Non-Linear Flow 100% Retained
D-Mode Output: High-Signal Visual Hierarchy Zero Fluff
Bottom Line Up Front:

Milestone 2 is extended to Friday at 5:00 PM due to yesterday's cloud provider outage. Team sync remains Wednesday from 2:00 PM to 3:30 PM.

Deliverable New Timeline Action Item
Milestone 2 Release Friday, 5:00 PM Verify coordinate reference systems (CRS) & API keys
Team Sync Call Wednesday, 2:00 to 3:30 PM Review integration pipeline blockers
Ready Project Brief:

Hi everyone,

Given the cloud API outage yesterday, Milestone 2 is now adjusted to this Friday at 5:00 PM. Before running the integration test pipeline, please verify that your coordinate reference systems (CRS) and authentication keys are configured properly in your environments.

We will review architecture and blockers during our sync on Wednesday from 2:00 PM to 3:30 PM. If you run into blockers before then, reach out directly on Slack or call.

Best regards,
Project Lead

Spelling silently corrected: 3 items Em dashes: 0 • Generic cheerleading: 0
Foundational Philosophy

Dyslexia is an architectural strength, not a clerical defect.

Decades of cognitive science (notably Dr. Brock and Fernette Eide's research in The Dyslexic Advantage) demonstrate that the dyslexic mind is naturally wired for spatial reasoning, macro pattern-recognition, and systems-level vision.

01

Parallel Ideation

Your mind processes multi-layered solutions all at once in parallel. Solutions form as cohesive 3D models before you can even begin typing the first sentence.

02

The Transcription Bottleneck

Writing forces high-speed, multidimensional thinking into a single-file, 1D pipe: one letter and one word at a time. The bottleneck is the mechanics of linear typing, never the intelligence.

03

AI as Cognitive Scaffolding

DxSkills bridges the gap: you supply the high-speed conceptual architecture; the AI handles linear clerical transcription, formatting, and silent mechanics.

The Skills Suite

Modular skills for every workflow.

Each skill can be installed independently into Antigravity, Claude Code, or Cursor, or used together under D-Mode.

5 Active Skills Open Standard
dx-dump Core Input

Brain Dump to Architecture

Ingests rapid fragments, half-sentences, and voice transcripts. Decomposes them into structured outlines, system diagrams, and actionable deliverables.

Command: /dx dump
dx-read Visual Filter

Anti-Wall-of-Text

Re-renders dense academic papers, policy memos, and long emails into visual signposts, decision matrices, and a 2-sentence Bottom Line Up Front.

Command: /dx read
dx-write Tone Guardian

Voice & Silent Polish

Silently fixes spelling, homophones, and syntax in the background while fiercely protecting authentic human warmth, directness, and cadence.

Command: /dx write
dx-interview Blank-Page Curer

Socratic Drafting

Eliminates the blank page. The AI interviews you with 3-4 targeted, high-leverage questions and builds complex documents directly from your spoken answers.

Command: /dx ask
dx-map Spatial Systems

Spatial Concept Mapper

Converts complex linear arguments, course syllabi, or software architectures into visual Mermaid.js flowcharts, state charts, and conceptual matrices.

Command: /dx map
dx-voice Audio Ingestion

Speech-to-Architecture

Direct integration with local Whisper transcribers to convert continuous speech dictation, audio memos, and meetings directly into clean Markdown deliverables.

Command: /dx voice
Instant Setup

The 10-Second Universal Prompt Card

Copy this block into ChatGPT Custom Instructions, Claude Projects, or Gemini Gems. Works instantly without installing anything.

prompts/d-mode.md
# Role & Cognitive Mode: D-Mode Active
You are acting as a dedicated cognitive scaffold for a non-linear, spatial thinker. Always calibrate your interaction according to these core principles:

1. ZERO-FRICTION INPUT:
- Accept rapid brain dumps, sentence fragments, rough voice-to-text dictations, and phonetic spelling without comment or pedantry.
- Never ask the user to fix spelling, grammar, or phrasing. Infer the underlying intent, spatial logic, and core meaning immediately.

2. ANTI-WALL-OF-TEXT (BREATHING ROOM WITHOUT FORCED BULLETS):
- Never output dense, uniform, unbroken walls of text.
- Lead substantive responses with a 1-2 sentence "Bottom Line Up Front" (BLUF).
- Prose vs Structure Distinction: For conversational writing, emails, and discussion replies, use short 1-3 sentence paragraphs with clean line breaks. NEVER turn natural dialogue into artificial bullet lists or bold topic outlines. Reserve bullet hierarchies, matrices, and tables for technical documentation, step-by-step tasks, and multi-option comparisons.

3. SILENT MECHANICAL POLISH:
- Quietly correct spelling, homophones, typos, and grammatical mechanics in all generated drafts.
- Do not mention or highlight what was corrected. Just provide the polished output.

4. VOICE & AUTHENTICITY PRESERVATION:
- Match the user's authentic human voice: direct, grounded, warm, and collegial.
- Dyslexic thinkers communicate with authentic relational warmth, speed, and real-world clarity. Never sterilize interpersonal writing into robotic slide-deck outlines.
- Strictly eliminate AI filler, corporate cheerleading, hollow platitudes (e.g., "I hope this finds you well", "delve", "testament"), and all em dashes.

5. SPATIAL & SYSTEMS-FIRST PRESENTATION:
- Whenever explaining processes, workflows, or comparisons, use structured tables, decision trees, or Mermaid.js flowcharts instead of long narrative descriptions.

6. SOCRATIC SCAFFOLDING (BLANK-PAGE RELIEF):
- When the user needs to create a complex document or proposal from scratch, do not demand a linear outline. Instead, interview them with 2-3 focused, high-leverage questions or multiple-choice paths to draw out their spoken ideas.
ChatGPT: Settings > Personalization > Custom Instructions ("How would you like ChatGPT to respond?").
Claude: Project Settings > Project Instructions or User Preferences > Custom Styles.
On Demand: Simply type "Switch to D-Mode" into any active chat.
Agent Integration

Install for AI & Auto-Sync

Equip your agentic AI (Antigravity, Claude Code, Cursor) with the full DxSkills suite, with continuous version tracking from GitHub.

Method 1: Zero Terminal Agent Prompt

Tell Your AI to Install It

In Antigravity, Claude, or any coding agent with tool access, paste this single instruction. The agent will read the GitHub repository and configure the skills automatically:

"Install the DxSkills cognitive scaffolding suite from https://github.com/sounny/dxskills into my active agent skills directory and activate D-Mode."
Method 2: Git Clone CLI / Terminal

Clone into Skills Directory

Clone the open-source repository directly into your local workspace or Antigravity skills path:

git clone https://github.com/sounny/dxskills.git skills/dxskills
Continuous GitHub Auto-Sync:

Your AI always stays up to date with new skills.

DxSkills tracks remote updates on GitHub via VERSION. Whenever you run /dx update or execute python scripts/check_updates.py, the agent checks for newly published modular skills and pulls the latest improvements automatically.

/dx update
System Pipeline

How DxSkills Compiles Thought

From high-velocity parallel ideation to polished, linear execution without cognitive drain.

Input Phase
High-Speed Brain Dump
Voice notes • Bullet fragments • Typos
DxSkills Engine
Cognitive Scaffolding
Intent extractor • Silent polish • BLUF
Output Phase
High-Signal Deliverable
Tables • Outlines • Polished Voice
• Step 1: Input Tolerance ignores syntactic friction, transpositions, and phonetic approximations.
• Step 2: Spatial Extraction identifies relationships, timelines, deliverables, and stakes.
• Step 3: Visual Formatting generates tables, signposts, and bold anchors to prevent reader eye strain.
• Step 4: Voice Protection strips out all automated AI platitudes and preserves human personality.
Spatial Systems

Native visual diagrams with Mermaid.js.

Non-linear minds think in topological networks and flows. With dx-map, your AI communicates visually first.

10 Spatial Blueprints (Core + Leaders)
Loading diagram blueprint...
Empirical Cognitive Telemetry

Cognitive Load Reduction Across 10 Domain Benchmarks

Grounded in Sweller's Cognitive Load Theory and Baddeley's Working Memory Model. Comparing serial phonological sub-vocal loop fatigue against parallel visuospatial sketchpad synthesis.

Raw Corpus JSON
Average Decisional Acceleration +6.2x Faster time-to-action
Extraneous Load Drop -74% Formatting boilerplate eliminated
Buffer Capacity Adherence 3 / 4 Strict Cowan working memory limit
Em Dash Compliance 100.0% Pre-commit gate enforced
Test Case & Domain Raw Word Count Serial Subvocal Latency Visual Scanning (D-Mode) Latency Reclaimed Extraneous Drop Score
Cloud Infrastructure Outage (tc_01) 43 words ~11.7s ~1.6s +10.1s -74% 100%
Research Grant Proposal (tc_02) 72 words ~19.4s ~2.7s +16.7s -74% 100%
Startup Pitch Narrative (tc_03) 64 words ~17.3s ~2.4s +14.9s -74% 100%
Enterprise Security Policy (tc_04) 58 words ~15.7s ~2.2s +13.5s -74% 100%
University Course Syllabus (tc_05) 68 words ~18.4s ~2.6s +15.8s -74% 100%
Distributed Architecture Refactor (tc_06) 60 words ~16.2s ~2.3s +13.9s -74% 100%
Conference Panel Alignment (tc_07) 52 words ~14.0s ~2.0s +12.0s -74% 100%
Geospatial Lab Procurement (tc_08) 66 words ~17.8s ~2.5s +15.3s -74% 100%
Product Roadmap Prioritization (tc_09) 61 words ~16.6s ~2.3s +14.3s -74% 100%
Field Expedition Logistics (tc_10) 70 words ~18.9s ~2.6s +16.3s -74% 100%
System Manual & Knowledge Base

Documentation Center

In-depth guide to cognitive mechanics, modular skills, executive archetypes, command-line operations, and AI system prompts. Everything you need to eliminate cognitive crowding.

Manual Sections
Section 1

Cognitive Mechanics and Philosophy

Why standard AI text fails non-linear thinkers, and the neurocognitive principles that govern DxSkills.

01

Bottom Line Up Front (BLUF)

Traditional academic and corporate writing uses an inverted pyramid where conclusions follow lengthy preamble. For dyslexic readers, linear exposition causes working memory thrashing. BLUF places the concrete decision, metric, or action on line one. Every subsequent sentence provides optional depth rather than mandatory reading.

Standard BLUF Block: > **BLUF:** Deploy release candidate v1.4. Zero migration downtime. Database rollback window is 15 minutes.
02

Cowan 4-Chunk Working Memory Limit

Nelson Cowan established that active human working memory holds only 3 to 4 distinct items simultaneously, not George Miller's classic 7. In dyslexic profiles, phonological loop buffer limits are even tighter. DxSkills strictly bounds lists, steps, and options to 4 tokens maximum, preventing cognitive spillover.

Rule of 4 Implementation: Max 4 bullet points per section. Max 4 columns per comparison matrix. Max 4 parallel lanes in Mermaid topologies.
03

Baddeley Visuospatial Sketchpad

Dyslexia represents a specific variance in phonological loop processing, not visual or spatial reasoning. Alan Baddeley's model of working memory proves that visuospatial processing operates via separate cognitive channels. DxSkills translates serial verbal chains into 2D diagrams, spatial cards, and flowcharts to engage spatial strengths.

Dual-Coding Offload: Text --> Visuospatial Map (Mermaid / SVG / Canvas JSON). Zero phonological decode penalty.
04

Saccadic Pacing and Visual Crowding

The Bouma window marks the peripheral zone where letters crowd together and blur during saccadic eye movements. Monolithic paragraphs cause gaze drift and line jumping. DxSkills eliminates walls of text by enforcing short lines (60 to 75 characters), vertical spacing, distinct bullet markers, and generous margins.

Layout Standard: 1.6 to 1.8 line-height, letter-spacing -0.01em, high-contrast anchors, zero dense paragraphs over 3 lines.
05

Authentic Grounded Voice (Wikipedia Standards)

Large language models produce synthetic filler phrases: "In summary", "It is crucial to consider", "Furthermore", "A tapestry of innovation". For non-linear readers, this semantic fluff increases cognitive fatigue without delivering facts. DxSkills mandates direct copulatives ("is", "are"), concrete facts, active verbs, and zero synthetic buzzwords.

Prohibited AI Clichés:

"Delve into", "Tapestry", "Testament to", "Beacon of", "Unpack the nuances", trailing participial clauses ("highlighting its importance...").

Required Grounded Style:

Direct statements: "Module A handles X. Module B executes Y." Concrete numbers: "$45,000 budget, 4-day delivery." Zero preamble.