Stylo

Why we built Stylo

Spark at Berkeley. In a talk at UC Berkeley, Garry Tan hammered a simple truth: your prompt is your leverage. We felt that in our own work—same task, five different AI apps, five different ways to “say it right.” Vague prompt → vague output. We were wasting hours rewriting the same ask per tool.

The pain. Fragmented syntax: ChatGPT ≠ Claude ≠ Perplexity ≠ Midjourney ≠ Cursor/Notion AI. Vague-in, vague-out: most people don’t have time to craft the perfect instructions. No memory: tools forget your voice, constraints, and what “good” looks like. Result drift: outputs change across platforms and over time—hard to get repeatable quality.

Our insight. The bottleneck isn’t the models; it’s the request. If we can translate any messy ask into a platform-specific, preference-aware prompt—every time—we unlock consistent, high-quality results.

What Stylo does (one line). Stylo turns your messy ask into a platform-tuned prompt with memory—adapting to your taste to deliver the output you actually wanted.

How it works (fast): Parse intent from your raw request. Tune to the target (ChatGPT, Claude, Perplexity, Gemini, Midjourney, Cursor/Notion AI)—using the right patterns, tokens, model optimization, and guardrails per platform. Layer your memory (tone, format, do/don’ts, examples) and any context (docs, persona, constraints). Learn from acceptance/edits to refine future prompts automatically.

Why now. Models keep improving, but human time doesn’t. Teams need consistent, on-brand, cross-tool results without becoming prompt engineers.

Outcome. Less prompt thrash. Fewer retries. Consistent, “this is exactly it” outputs—everywhere you work.