The Problem
AI-generated text has an uncanny valley problem. It's grammatically correct, well-structured, and completely devoid of personality. Every LLM output sounds the same -- polished, hedging, vaguely enthusiastic. When you use AI to draft emails, project descriptions, social posts, or documentation, the result reads like it was written by a committee. Kenify exists to solve that gap by capturing what makes a specific person's writing sound like them.
How It Works
Kenify analyzes a corpus of someone's existing writing -- emails, Slack messages, project docs, social posts, anything with authentic voice -- and extracts a detailed style profile. This isn't just tone detection. The system maps vocabulary preferences (which words you reach for, which you avoid), sentence structure patterns (short and punchy vs. long and nested), rhetorical habits (do you use analogies? rhetorical questions? self-deprecation?), punctuation style (em dashes, parenthetical asides, ellipses), and personality markers (humor type, directness level, formality gradient).
The resulting profile is a structured document that can be injected into any LLM prompt as a style constraint. Instead of asking Claude to "write in a casual tone," you give it a precise specification of what casual means for this specific person.
Why It Matters
As AI writing tools become standard across every workflow -- from portfolio copy to client emails to internal documentation -- the people who use them risk losing the thing that made their communication distinctive. Kenify preserves voice at scale. Write something once in your own voice, and every AI-generated piece that follows maintains that consistency.
The Technical Approach
The analysis pipeline uses Claude's language understanding to perform multi-dimensional style extraction. It processes writing samples across several axes: lexical analysis (word frequency, vocabulary richness, jargon usage), syntactic patterns (sentence length distribution, clause complexity, paragraph structure), pragmatic features (hedging frequency, assertion strength, humor markers), and interpersonal style (formality, directness, warmth). The output is a reusable JSON style profile that acts as a voice fingerprint.
Applications
Portfolio sites (like this one), email drafts, social media content, project documentation, client-facing communications, and any context where AI-generated text needs to sound like a specific person rather than a generic language model.
Tech Stack
Claude API (style analysis and profile generation), NLP analysis pipeline, JSON style profiles, prompt engineering framework.