Hah, I love the term 'Poppycock.' You’re 100% right -- Disney’s 10-K is PR noise. I don't use it for 'Intent'; I use it to derive a baseline for Systemic Friction.
My Nexus engine doesn't just scrape; it scores. Cold outbound starts with a 'Trust' score of 3.0. I use a Trust Exponent to penalize that low trust so severely that a high-intent '10-K signal' still results in a REJECT if there’s no structural fit.
Here is the core logic loop I’m running on my workers right now to prevent that exact 'vaporware' problem:
python
def calculate_score(self) -> int:
# 1. Evaluate Hard Gates First (Policy Check)
if self.fit < self.config.min_fit_gate or self.intent < self.config.min_intent_gate:
return 0 # Disqualified - No 'PR' news can bypass this.
# 2. Base Opportunity Score (Weighted 4-Pillar Model)
base_opp = (self.intent * 0.3) + (self.fit * 0.3) + (self.urgency * 0.2) + (self.trust * 0.2)
# 3. Apply the Trust Exponent
# Normalize trust, then shape it to penalize cold outbound baseline (3.0) severely.
t_multiplier = (self.trust / 10.0) ** self.config.trust_exponent # trust_exponent = 1.5
# Final 100-point composite signal index
return int(round((base_opp * 10) * t_multiplier))The '10-K news' is just one input into self.intent. If self.trust is at 3.0, the math literally kills the lead before it ever hits an outbox. Are you building your CRM agents with a similar hard-gate policy check, or are you scaling trust through a different multiplier?
