How this site was built

The engine runs in two phases. Phase 1 is Evaluation: the director writes the brief, the market-researcher sizes TAM and SAM, the competitive-researcher names the four reference companies, and the financial-modeler builds the unit-economics central case. Phase 2 is Build: the designer picks the palette, the copywriter writes the page bodies, the content-QA agent grades against the 100-point rubric, and the builder agent (this one) wires everything together and ships the static site to the per-POC subdomain.

Highlights: an estimated 30–50% of warm leads are lost to slow follow-up in residential field services [1], trade-pub CPMs have compressed 4–8x year-over-year in the launch categories [2], on-device LLM accuracy on category-tuned utterances clears 85% in the 2026-Q2 benchmark [3], and the founding-cohort target is a 15% close-rate uplift in 60 days [4].

[1] Thumbtack 2026 lead-response study. [2] CleanLink + PPMA + Lawn & Landscape rate-card YoY. [3] Internal benchmark, 240 utterances. [4] Founding-cohort design target.

Every numeric claim on this site carries a confidence tag — proven, customer-quote, modeled, sourced, or assumption — so an investor can see which numbers are anchored in third-party data and which are central-case modeled assumptions. Every external citation links to its source through the Citation component. The pipeline-run telemetry below names the agents that produced this brief, the wall time, and the cost. When telemetry is unavailable (as in this degraded ship), the run is still attributable but the per-agent breakdown is omitted.