FSLY Thesis Assessment

The single most important market in the set (information gain 1.00) and the most consequential for the thesis. The 22% probability that management quantifies AI traffic contribution implies a 78% probability that the central narrative driving the stock repricing remains unverified through H1 2026. Four of five analysis lenses independently flagged the AI narrative as E1-only evidence, and the Myth Meter estimated 20-30% of the stock surge was attributable to this unquantified AI overlay. With high model agreement (0.93), the ensemble is confident that management will likely continue declining to quantify AI's contribution -- consistent with the pattern of non-disclosure at Q4 2025 earnings, their most favorable disclosure moment. This maintains reversion risk on the AI-premium portion of the stock price.

The nearest-term falsification test for turnaround durability, and the market with the most balanced probability distribution. At 54%, the ensemble treats Q1 sustaining 18%+ growth as essentially a coin flip. This is revealing: FY 2026 guidance of $700-720M implies only 12-15% full-year growth, suggesting management itself expects deceleration from Q4's 23%. The gap between Q4's 23% and the 18% threshold means meaningful deceleration could still register as a 'pass.' Q4 2025 may have benefited from holiday streaming traffic and one-time factors. This market's near-50% probability prevents it from anchoring the thesis in either direction but makes Q1 2026 results the key catalyst for reassessment.

The strongest constructive signal in the market set for unit economics. At only 27% probability that price erosion reverts to double digits, the ensemble considers the improvement from mid-teens to mid-single digits likely to persist. This was the most significant new unit economics datapoint from Q4 2025 and the meta-synthesis identified it as the 'highest-impact unknown.' The ensemble's relatively low probability of reversion suggests structural pricing power rather than merely cyclical demand tightness, which would support the PLAUSIBLE unit economics assessment. However, at information gain 0.64, this is a medium-weight signal that partially offsets the escalation signals but cannot overcome the AI narrative gap and concentration concerns.

At only 32% probability of remediation, the ensemble expects the material weakness to persist in the FY 2025 10-K filing. This maintains a governance and process risk qualification on the otherwise strong operational execution narrative. Three of five lenses flagged this concern, and the RPO recast in Q3 2025 (due to termination-for-convenience clause calculation errors) raised additional questions about revenue measurement processes. While not a terminal risk -- the material weakness relates to process controls, not a restatement of financials -- it creates an uncomfortable juxtaposition with the 'first profitable year' narrative and suggests internal controls have not kept pace with operational improvement.

The highest-conviction de-escalation signal. With only 11% probability and the highest model agreement in the set (0.96), the ensemble strongly expects NRR to remain above 105%. The improvement from 102% to 110% appears durable, supported by multi-product adoption and platform switching costs. This dismisses the Gravy Gauge's worst-case scenario of a CONDITIONAL-to-FRAGILE REVENUE_DURABILITY downgrade. However, the Revenue Revealer's observation that gross retention actually declined (99.0% to 98.7%) while NRR expanded means expansion is masking rising churn -- a dynamic that could eventually surface even if NRR stays above 105%.

The 87% probability that no insider purchases appear through August 2026 reinforces the DIVERGING narrative-reality gap classification. With 19 sales and zero purchases, the insider pattern remains starkly one-directional. The new management team (CEO Compton, CFO Wong) has not demonstrated alignment through personal buying despite promoting a turnaround narrative. After a 71% stock surge, continued insider silence would be especially notable -- management is not endorsing with personal capital the valuation the market is assigning. While insider selling alone is not diagnostic (tax, liquidity, estate planning), the complete absence of buying across all insiders over an extended period is a pattern worth noting.

The only market with a majority-probability escalation outcome. At 60% probability, the ensemble considers it more likely than not that customer concentration will continue rising past 36%. Top 10 customers are growing 40% faster than the rest of the base, and the streaming entertainment cluster (~12% of revenue) creates effective single-customer concentration risk. While rising concentration in itself does not damage revenue -- it reflects large customers expanding rapidly -- it amplifies the impact of any single customer loss or renegotiation event. At 34% and rising, the revenue base is becoming more, not less, fragile despite the turnaround.

At only 18% probability, the ensemble does not expect the security revenue mix shift to reach the 25% threshold by Q4 2026. This is consistent with the Revenue Revealer's assessment that the ~2 percentage point per year pace of mix shift is 'too slow to materially change revenue structure within 2-3 years.' Security revenue growing at 32% YoY from a 21% base would need to accelerate meaningfully or network services would need to decelerate for 25% to be reached. This limits the structural revenue quality upgrade path -- the business remains approximately 76% usage-based, which constrains the CONDITIONAL-to-DURABLE upgrade the market appears to be pricing in.