About this analysis. Prepared by Avisha NessAiver (Distilled Science). Every video the program's ~60 creators posted in July 2026 was collected from the program portal, direct account crawls, and per-platform stat lookups, then transcribed and run through an LLM analysis pipeline: content-quality scoring across nine dimensions, hook-framework classification, a topic taxonomy (including whether a video is AI-safety-specific), and audience profiling. Platform stats (views, likes, comments, shares, followers) come from the platforms themselves. Email me to discuss this analysis, add components to it, or commission the same analysis for other accounts or groups of accounts.
plzdontkillus — Program Impact
A one-month creator bootcamp: get participants making broadly good, varied content — and specifically good content about AI and AI safety.
Program Timeline
Posting volume and views in three-day steps from July 1 — a sustained month, not a burst that fizzled.
Content Breadth
Goal 1 was variety, not just AI content. Here's what people actually made.
Did the Mission-Aligned Content Hold Its Own?
Goal 2 was good AI-safety content specifically. Comparing it against general AI/tech content and everything else — same pipeline-scored quality metric, same engagement metrics.
Cross-Platform Reach
Most participants weren't single-platform — the same ideas, tested on multiple audiences.
How Broadly Did Success Spread?
Reach concentrates in any content ecosystem — the honest question is how many people still broke through.
Content Quality Over the Month
How quality is measured: every video with a usable transcript is rated by the analysis pipeline's LLM on nine content dimensions — demonstrated expertise, insight, educational value, myth-busting/fact-checking, education-vs-entertainment balance, context framing, content clarity, hook curiosity, and hook strength. Each dimension is normalised 0–1 across the program's videos, and the quality score is their average. It judges the content itself, deliberately not raw views — views are biased by how long a video has had to accumulate them (a video posted last week hasn't caught up to one posted five weeks ago regardless of quality).
Biggest individual gains
Biggest individual declines
What Worked, Pooled Across ~60 Creators
Hook-framework performance pooled across every participant — a sample size no single-creator dashboard has. "Creators" = how many different people this actually worked for, not just one person's lucky video.
Who Did This Actually Reach?
Pooled audience profiling across every scored video — did the program preach to the AI-literate choir, or reach people new to the topic?
How this is determined: for each scored video, an LLM reads the transcript and caption and predicts the audience the platform's algorithm would most likely serve it to — expertise level, age band, and viewing intent. The charts pool those per-video predictions. These are model estimates inferred from the content, not platform-reported viewer demographics (which platforms don't expose for other people's accounts).
Age distribution
Viewer intent
Highlight Reel
The videos worth watching to understand what this program produced.