Silence is
Intelligence

An autonomous AI writing partner that earns the right to interrupt.
Built with Gemini 3.

The Inspiration

Real collaboration isn't loud.
It's intentional.

Most AI writing tools feel like an overexcited intern—constantly interrupting, flooding the screen with suggestions, and breaking flow.

The inspiration behind Co-Author came from a simple question: What if AI behaved like a disciplined, thoughtful co-author instead of a reactive chatbot?

"I wanted silence to mean intelligence, and intervention to mean value."

> Analyzing intent...
> High relevance (0.92)
> Suggesting edit...

Autonomous Partner

Observes writing in real-time, contributing only when genuinely useful using a Notion-style research editor.

Context Engine

Treats the entire document as living context, allowing for deep analysis and paraphrasing without hijacking.

Real-time Voice

Talk through ideas while the AI quietly handles background research using WebSockets & AudioWorklets.

Hand over the pen.
Step In Mode.

Delegated authorship with complete transparency. For the first time, hand over full editorial control to an AI that thinks, plans, executes, and self-critiques—just like a human co-author.

1

Planning

Reads your instructions and formulates a deep editorial strategy.

2

Execution

Rewrites, continues, or polishes your work directly in the editor.

3

Evaluation

Self-evaluates the output against your goals and finds weak points.

4

Refinement

Makes final surgical adjustments to ensure absolute perfection.

The 7-Layer Maturity Stack

Most AI writes too much. This stack is designed to teach the AI when not to speak.

Observe

The AI silently watches keystrokes, building context without acting.

Wait

Enforces a mandatory cooldown to ensure it doesn't block the writer's flow.

Evaluate Intent

Analyzes the text: Is the user stuck? Or just pausing for thought?

Draft Solution

Generates a potential contribution in a hidden buffer.

Score Relevance

Is this genuinely useful? If the score is low, discard immediately.

Justify Interruption

Calculates the cost of breaking the user's focus.

Present

Subtly renders the suggestion in the side margin, never the main text.

Enforces Cooldowns Validates Intent

Built for Performance

Real-time systems demand real engineering. Shortcuts in audio, scheduling, or buffering always surface as UX problems.

Next.js

App Router Framework

Tiptap

Headless Editor

Zustand

State Management

Supabase

Persistence & Auth

Google Gemini 3 Flash

Reasoning & Context

AudioWorklets

Real-time Voice Playback

Security First

Preventing prompt injection was critical. I implemented a security layer that actively blocks jailbreak attempts while allowing "Shadow Prompts" for intent scanning.

Audio Engineering

To solve jitter and buffering, I moved playback logic off the main thread entirely using AudioWorklets, achieving smooth, uninterrupted audio.

Reflection

What I Learned

"I learned that good AI UX is more about restraint than raw intelligence. Latency, silence, and timing matter just as much as model quality. Users trust AI more when it behaves like a collaborator, not a performer."

What's Next?

  • Multi-document reasoning & cross-paper citation mapping.
  • Refining the voice agent into a true thinking partner.
  • Collaborative modes for teams.

Long-term, Co-Author aims to become the standard environment where humans and AI co-write — calmly, deliberately, and at scale.