Talk 15

Building Monologue iOS

iosmonologue

Summary

Overview

Naveen Naidu, solo developer of Monologue at Every, demonstrated the voice-to-text Mac app and previewed the upcoming iOS release (targeting February 9th). The session focused on advanced features like modes, the notes feature, and how Naveen uses vibe coding with Claude Code and Codex to build competitive products as a solo developer against VC-backed teams.

Key Points

Product Features Demonstrated

  • Shortcut triggers: Hold-to-record for quick (5-10 second) recordings, tap-to-toggle for longer sessions (1+ minute brain dumps)
  • Modes: App-specific custom instructions that auto-activate based on which app is in focus (e.g., Claude Code mode activates when in Ghostty terminal)
  • Auto-enter: Automatically sends the transcribed text without manual confirmation
  • Paste last transcript: Quick shortcut to paste previous transcript anywhere
  • Notes feature: New capability for longer-form voice capture with transcription and summary

Development Approach

  • Uses vibe coding extensively: prototyped entire Notes feature in one hour using brain dumps via Monologue to Claude Code
  • Codex vs Claude Code split: Uses Codex for bug fixes and multi-file changes requiring deep codebase understanding; uses Claude Code for creative prototyping and new features
  • "Codex is that one senior engineer where it understands all the code... But when I'm vibe coding, I don't usually do Codex because Codex is not that creative"
  • Philosophy: "In the age that right now we are living in, implementing features is not really that important. It's knowing what to implement and what actually gets people excited"

Usage Statistics

  • P50 average: ~48 words per monologue
  • P90 users: 400-1000 words per session
  • Power users: 300 times per day
  • Two users crossed 1 million words total
  • Product growth: From 1 million words/month at launch to 1.5 million words/day

Competitive Landscape

  • Space is crowded with VC-backed competitors ($10-80M raised)
  • Reframed competition as positive: competitors educate the market, making adoption easier
  • "Apple dictation is also there, but which is really bad"

Roadmap Highlights

  • iOS launch: February 9th (modes and scores sync across devices)
  • Auto-improving modes: Monologue learns from user edits and updates modes automatically
  • Spiral integration: Connect notes to Every's writing tool for blog post creation
  • Windows support: Future consideration
  • Custom skins: Fun personalization feature being considered
  • Hardware: Naveen's "crazy idea" (Brandon: "I'm going to take this conversation off the live stream")

Key Insight

The session demonstrated how a solo developer leveraging AI tools (Claude Code, Codex) can build and maintain a product competitive with well-funded teams. The combination of Every's distribution, active Discord feedback loops, and rapid vibe-coded prototyping creates a sustainable development model.

Actionable Takeaway

For voice-to-text workflows: Set up app-specific modes with custom instructions and enable auto-enter. Use hold-to-record for quick inputs and tap-to-toggle for brain dumps. The more context you give (400-1000 words at P90), the better the output.

Key Concepts

Hold-to-Record vs Tap-to-Toggle

Two distinct interaction patterns for voice input: hold-to-record for quick 5-10 second captures using a modifier key, and tap-to-toggle for longer sessions lasting minutes. The choice depends on content length and context.

Modes (App-Specific Custom Instructions)

Context-aware profiles that automatically activate based on which application is in focus. Each mode can have its own system prompt, formatting rules, and behaviors like auto-enter.

Auto-Enter Feature

Eliminates the review step by automatically sending transcribed text upon recording completion. Enables truly hands-free operation when combined with modes.

Codex for Precision, Claude Code for Creativity

Strategic tool selection: use Codex when working with large codebases requiring multi-file understanding and precise edits; use Claude Code when prototyping new features or exploring creative solutions.

Brain Dump Prototyping

Using voice-to-text for extended feature descriptions (400-1000 words), then sending directly to AI coding assistants. Enables rapid prototyping - Naveen built the Notes feature prototype in one hour.

Notable Quotes

"If you want to live in the future you cannot be typing anything. You have to be using your voice. -- Dan Shipper (host)"
"I just love building products. That's it. -- Naveen Naidu"
"In the age that right now we are living in, implementing features is not really that important. It's knowing what to implement and what actually gets people excited. That's the most important part. -- Naveen Naidu"
"Always try to implement the feature and see if you're using it personally or not. If you're not using it, better to throw it away and start from scratch again. -- Naveen Naidu"

Tools Mentioned

MonologueSuperWhisperWhisperflowApple DictationCodex (OpenAI)Claude CodeGhosttyClaude (Desktop App)SpiralGranolaEveryDiscord

Transcript

15 of 18