Founder & CEO
We ran automated accessibility testing against 285 homepages sampled across the full Tranco top-1M range. 89.5% failed a WCAG A/AA check, robust under every counting convention we preregistered. Our second hypothesis did not survive its own sensitivity checks, and we report that as a divergence.
Your presence inside AI answers is invisible by default. Here is how to measure your AI search visibility with four methods, manual prompt testing, visibility tools, server logs, and GA4, and the traps that make people misread the data.
We audited a four-lens retrieval system for a subtle paraphrase-preference bias — and caught the LLM in our own analysis loop writing the conclusion it expected before reading the results, three times. The discipline that fixed it, and four confounds that each shrank our effect.
We built a four-lens fusion retrieval stack, then ran the full 15-way ablation. The cheapest two-lens pair — one dense, one sparse — beat the expensive four-lens stack on every ranking metric, and a 27B model was one of the weakest lenses.
We discovered that HuggingFace Text Embeddings Inference and standard PyTorch produce vectors with cosine similarity as low as 0.30 for the same model.
