LinkPost V2 ships with a new algorithm. More precise, and most importantly personalized for each user.
Progress since V1:
• 300+ factors analyzed (vs 49 before)
• 27 personal factors learned from your style (vs 0 before)
• 1 user out of 3 sees their actual best post correctly identified (vs 1 out of 4)
• +71% posts that break through (vs +67% before)
• −75% flops avoided (vs −48% before)
To get these results, we didn't write "one magic prompt".
We built an analysis system that interconnects:
• Several AI models (regressor, ranker, calibration)
• Several specialized algorithms (novelty detection, audience modeling, topical trend analysis)
• A proprietary scoring engine
Every post is analyzed from different angles (structure, emotion, controversy, clarity, expected engagement, novelty vs your usual style, fit with your audience, topic maturity on LinkedIn), then cross-referenced with over 300 factors from real data, including 27 personal factors learned from your history.
This system was trained and validated on more than 50,000 posts from 3,500 creators, compared to their real post-publish performance.
The algorithm predicts the probability that a post outperforms your usual average.
Experimental results on 49,908 posts:
• "Break through" = do 1.5× or more your average likes
• "Flop" = do less than 0.5× your average
• Natural baseline: ~24% of posts break through naturally
Scenario A, without selection (100 random posts):
• ~24 break through | ~63 average | ~14 flops
Scenario B, with scoring selection (top predicted scores):
• ~40 break through | ~56 average | ~3 flops
Observed gain:
• Posts that break through: +71%
• Flops avoided: −75%
And it's only the beginning. The algorithm keeps learning — the more data we accumulate, the more precise it becomes.
