Turn a single topic into platform-native, voice-preserving social posts — with trend discovery and closed-loop learning.
AI content tools have a dirty secret: they normalize your voice toward a generic, polished register — even when you explicitly tell them not to. A landmark April 2026 arXiv study ("Voice Under Revision," analyzing 300 personal narratives rewritten by three frontier LLMs) confirmed what creators have been reporting for years: LLM rewriting produces consistent stylistic normalization regardless of prompt. Function words, contractions, and first-person pronouns decrease. Vocabulary diversity and punctuation elaboration increase. Even "voice-preserving" instructions only reduce, don't eliminate, this directional pull.
Meanwhile, creators spend 4-8 hours per long-form piece adapting content manually across 7+ platforms — reformatting, retoning, resizing — with each platform demanding different character limits, tone expectations, and format conventions. Multi-channel network (MCN) data shows ~30% of accounts get restricted for batch-duplicated, un-adapted content. The tools exist in pieces (Publora for publishing, Claude for generation, n8n for orchestration), but no one has packaged them into a single agent that preserves the creator's voice across every platform.
Cited research:
- Voice Under Revision: Large Language Models and the Normalization of Personal Narrative (arXiv, April 2026)
- Publora — #1 Product Hunt, June 10, 2026 (producthunt.com)
- SocialEcho 2.0 — Product Hunt Launch of the Day, June 1, 2026 (producthunt.com)
ResoNote replaces aspirational prompts ("write like me") with structural voice constraints. Instead of hoping the LLM preserves your voice, it builds an "engram" — a quantified voice profile from your actual prose samples — then generates content that must satisfy verifiable constraints:
Your writing samples → Voice Engram (tone + structure + anti-patterns)
↓
Topic brief → Trend Scout → Adaptation Engine (constraint-based generation)
↓
Review Queue (human approves/edits)
↓
Publish across platforms
↓
Performance Oracle (engagement data → refine engram)
The engram contains:
- Tone calibration: formality register, emotional valence, humor type, authority stance
- Structural patterns: sentence length, post length, paragraph density, emoji frequency, hook style
- Anti-pattern library: banned phrases, banned formats, voice-diluting constructions — things you would never say on each platform
The difference: telling a self-driving car "drive safely" vs. programming its lane-keeping algorithm. One is aspirational. The other is operational.
Drop SKILL.md into your Claude skills directory (typically ~/.claude/skills/ in Claude Code, or install via the Claude.ai skills panel). ResoNote activates whenever you mention cross-platform publishing, content scheduling, or turning a topic into social posts.
Tell ResoNote what platforms you use and paste your best posts:
Prompt: "ResoNote: I want to publish across Twitter, LinkedIn, and Bluesky. Here are 3 of my best tweets and 2 LinkedIn posts. Build my voice engram." Paste your samples. ResoNote analyzes them and builds a structured voice profile. You review and refine it.
Tell ResoNote what you want to post about:
Prompt: "ResoNote: I want to post about the new arXiv study on LLM voice normalization. It proves what I've been saying for a year. Give me angles for Twitter, LinkedIn, and Bluesky." ResoNote searches for trending angles, presents a briefing, generates platform-adapted drafts in your voice, and queues them for your approval. Approve → published.
| Phase | What happens | Your role |
|---|---|---|
| Trend Scout | Searches Reddit, HN, Product Hunt, news for trending angles on your topic | Pick which angles to pursue |
| Voice Forge | Builds/loads your per-platform voice engram from your writing samples | Provide samples; confirm the engram captures your voice |
| Adaptation Engine | Generates platform-native drafts using constraint-based generation + anti-pattern verification | Review, approve, edit, or reject each draft |
| Performance Oracle | Publishes approved drafts; feeds engagement data back into engram refinement | Check back in 24-48h for learning insights |
| Platform | Character Limit | Key Constraints |
|---|---|---|
| Twitter/X | 280 | Thread-compatible expansion; no link previews without context |
| 3,000 (1,200-1,800 recommended) | 3-5 line-break-separated paragraphs; clear CTA | |
| Bluesky | 300 (URLs count in full) | Lighter tone; conversational; app password auth |
| Threads | 500 | Personal, not overly polished; Meta App Dashboard OAuth |
| 2,200 | Visual-first caption structure; emoji-friendly if in engram | |
| TikTok | N/A (script) | Hook in first 1.5s; pattern-interrupt structure |
| Mastodon | 500 | CW if needed; alt-text for media; instance-dependent |
| Feature | ResoNote | Buffer/Hootsuite | Publora API | SocialEcho 2.0 | n8n DIY Pipeline |
|---|---|---|---|---|---|
| Voice preservation | ✅ Engram-based (constraint system) | ❌ None | ❌ None | ||
| Trend discovery | ✅ Autonomous (multi-source) | ❌ None | ❌ None | ✅ Social listening | |
| Per-platform adaptation | ✅ Constraint-enforced | ❌ Identical cross-post | ❌ Identical via API | ✅ AI rewriting | |
| Closed-loop learning | ✅ Engagement → engram refinement | ❌ Analytics only | ❌ None | ❌ Manual | |
| Anti-pattern enforcement | ✅ Structured library | ❌ None | ❌ None | ❌ None | ❌ None |
| Human review queue | ✅ Built-in | ✅ Dashboard | ❌ Code-only | ✅ Dashboard | |
| Setup time | 5 min (samples) | 30 min (connect accounts) | 60 min (API wiring) | 30 min (dashboard) | 2-4 hours (build pipeline) |
| Cost | Free (skill) + Claude subscription ($20/mo+). Publora/SocialEcho API costs may apply for direct publishing. | $6-120/mo (varies by platform) | $2.99-5.99/acct/mo | $10-15/mo | Free (self-hosted) or $20/mo+ (cloud) |
- Voice Under Revision: Large Language Models and the Normalization of Personal Narrative — arXiv, April 2026
- How AI-Powered Crossposting Is Replacing Manual Social Media Workflows in 2026 — The AI Journal
- How I Automated Multi-Platform Social Posting With Claude and n8n — SitePoint
- Brand Consistency in Generative AI Content — Markup AI, 2026
- Your AI Agent Needs Communication Modes, Not a Voice Clone — Dev.to, 2026
resonate/
├── SKILL.md # The skill — ResoNote's full instruction set
├── README.md # You are here
├── LICENSE # MIT
├── CONTRIBUTING.md # How to contribute
├── .gitignore # Standard ignores
└── engrams/ # Voice profil