10 Voice Notes. Zero Editing. 10,373 Impressions.

I built SparkVox because I had a LinkedIn problem.

As the host of We Live To Build with 300+ founder interviews, I talk to operators and investors every week. I had things to say. The ideas were constantly there. What I didn't have was time to write. So my LinkedIn feed was inconsistent at best, nonexistent at worst.

I tried the generic AI tools. You know what happened. Every post came out sounding like a McKinsey intern trying to be inspirational. Grammatically perfect. Completely forgettable.

So I built something different.

The Setup

Before running this test, I did one thing: I fed SparkVox my ten best-performing LinkedIn posts from the past six months. Posts that got real engagement. Posts that people shared without being prompted. Posts where the comments were specific, not just "great insights!" filler.

SparkVox read them and built a voice model calibrated to my sentence rhythm, my vocabulary, my way of building to a point. The training took about two minutes.

The Test

Over five days, I recorded ten voice notes. Nothing polished. Brain dumps while walking to a meeting, shower thoughts captured in the hallway, observations I had after a podcast recording. The average note was 45 to 90 seconds.

I hit "Feed Sparky Now" after each one. Read the draft. Published it without editing.

That last part is important. I made a deliberate decision not to touch the drafts. If they were good, that was the product. If they weren't, I needed to know.

The Results

After seven days:

• 10,373 impressions across ten posts
• 3,652 people reached
• +55.8% week-over-week growth in impressions
• 34% of the audience were founders
• 26% were CXOs

I have 2,135 followers. That's not a massive audience. But the reach-to-follower ratio was higher than anything I'd seen from posts I'd spent 45 minutes writing myself. And the audience quality was exactly what I was targeting.

What Made It Work

A few things stood out after reviewing the posts:

The hooks were mine

Every post opened with something that sounded like me at my most direct. Not "I'm excited to share..." Not a list of five things. An opening line that read like the first sentence of a story, because that's how I actually talk.

The structure held up

The best-performing posts shared a pattern: short sentences up front, a pivot in the middle, a specific insight at the end. SparkVox had extracted that pattern from my training posts and applied it consistently without me having to think about it.

The first comment did its job

SparkVox generates a first comment alongside every draft. It's a short follow-up that adds context or asks a question. I published each one within the first few minutes. LinkedIn's algorithm treats early engagement as a quality signal, and that first comment consistently primed the thread for real replies.

The Honest Caveat

Two of the ten posts were average. Not bad, but not the kind of posts that make you stop scrolling. In hindsight, those were the voice notes where my thinking wasn't sharp yet. Garbage in, slightly better garbage out.

The lesson isn't that SparkVox fixes bad ideas. It's that it removes every other obstacle between a good idea and a published post. The idea still has to be yours.

What I Do Differently Now

I don't write LinkedIn posts anymore. I record voice notes. Specifically:

• One observation from a recent podcast conversation
• One thing a founder told me that surprised me
• One mistake I made or watched someone else make this week

Three posts per week. Total time investment: under ten minutes, including approval. My feed hasn't gone cold since I started, and the audience quality in my analytics looks like my CRM, not a random cross-section of LinkedIn.

Try It

If you're a founder or operator who has things to say but no time to write, SparkVox is worth testing. Top up $20 at signup and get $30 in credits, enough for 15 posts. Run your own version of this test and check your analytics at day seven.

The product is the proof. If Sparky doesn't sound like you by post five, something's wrong and you shouldn't keep using it. But in my experience, and in the experience of the founders I've shared it with, it sounds right from the very first post.