A healthcare PM wants to know whether GLP-1 obesity leaders are still structurally underappreciated - or whether payer friction, net price pressure, compounding dynamics, supply limits, and next-generation competitors will compress the upside. Vextrum turns the debate into a continuously updating intelligence universe: sources, entities, hypotheses, evidence, deltas, and analyst actions.
Illustrative case study. Public-source examples used for demonstration. Not investment advice.
Vextrum does not issue buy/sell recommendations. It structures public-source evidence, thesis movement, confidence, contradictions, and analyst actions for PM review.
Volume expansion, broader cardiometabolic indications, employer adoption, supply normalization, oral formulations
Coverage restrictions, price compression, compounded substitutes, supply bottlenecks, next-gen competitors
Market may be overweighting headline demand and underweighting access / pricing / persistence deltas
The platform starts not from a dataset, but from an investment controversy. The PM defines the debate, the securities universe, the critical assumptions, and the signals that would change their view.
Vextrum converts a PM's ambiguous research concern into a machine-readable intelligence program: watchlist, hypotheses, source priorities, event taxonomy, confidence scoring, and output cadence.
The onboarding agent generates a machine-readable intelligence program from the PM conversation — covering the full debate structure, source priorities, and output requirements.
5 pillars: Demand, Access, Pricing, Competition, Execution. Bull/bear/variant branches defined.
LLY, NVO, ZEAL, AMGN, VKTX, ALT, PFE + Zepbound, Wegovy, Ozempic, Mounjaro, Rybelsus, pipeline assets
7 hypotheses (H1-H7) with bull/bear classification, initial confidence, evidence requirements
25+ event types: regulatory, payer/access, pricing, clinical, competitor, company, channel, international
4-tier authority: FDA/CMS/filings (T1), PBM/ICER/journals (T2), trade press (T3), KOL/social (T4)
Source-linked, confidence-scored, hypothesis-tagged. Weak signals cannot auto-upgrade conclusions.
FDA/CMS action → immediate. PBM restriction → analyst review. Competitor readout → thread update.
Weekly: exec summary, what changed, confidence movement, bull/bear evidence, Red Team, watchlist
Vextrum presents the investment debate questions derived from the thesis capture. Each question is mapped to a pillar, evidence requirement, and materiality threshold. The PM reviews, edits, and approves before they drive the intelligence universe.
Vextrum generates a first-pass intelligence configuration that can be reviewed, edited, and approved. Every parameter traces back to the onboarding conversation.
Vextrum does not start from a fixed feed. It discovers, classifies, scores, and proposes a source universe based on the PM's actual investment debate.
The PM's custom intelligence infrastructure — sources discovered, classified, scored, and mapped to the five GLP-1 debate pillars. Not a generic feed; a purpose-built research network.
Every document is transformed into entities, relationships, evidence snippets, hypothesis links, and confidence updates. The ontology maps the full GLP-1 investment debate.
With the GLP-1 intelligence universe approved, ontology generated, and sources mapped — the pipeline activates. The system begins continuous monitoring of all approved sources.
Strategic Judge screening incoming GLP-1 signals against debate pillars. Multi-stage hybrid semantic + reranker scoring.
Entity resolution mapping drug names, brands, generics, mechanisms, tickers. Multi-stage pipeline: normalize → exact → alias → fuzzy → LLM → create.
Bayesian Nexus with confidence-weighted hypothesis scoring, causal chain detection, and Red Team falsification against the GLP-1 thesis.
Illustrative timeline — actual timing depends on source availability and crawl schedule
The daily command surface for the GLP-1 investment debate. What changed, which hypothesis gained or lost evidence, and what requires analyst action — organized around the five debate pillars.
The Situation Room synthesizes evidence into thesis-updating threads. Each thread carries bull/bear interpretations, confidence scores, evidence packs, Red Team triggers, and recommended analyst actions.
The Delta is a live tape of thesis-updating events. Every entry explains what changed, which pillar and hypothesis are affected, how confidence moved, and what analyst action is required.
The living memory of the GLP-1 intelligence universe. Every entity, relationship, and evidence snippet is linked to a hypothesis, a debate pillar, and an analyst action. The graph is the PM's research operating layer.
Vextrum enforces structured dialectic: the Blue Team builds the investment thesis, the Red Team challenges it. Every conclusion requires falsification rules.
When a material signal fires, Vextrum assembles a structured evidence pack: affected entities, pillar mapping, confidence scoring, source-linked claims, bull/bear interpretation, and analyst next steps.
| Source | Date | Extracted Claim | Entities | Hypothesis | Tier | Conf. | Limitation |
|---|---|---|---|---|---|---|---|
| FDA.gov 503B Notice | [citation slot] | Semaglutide/tirzepatide/liraglutide proposed for 503B exclusion | NVO, LLY, 503B | Supports H4 | T1 | High | Does not address demand conversion |
| STAT News analysis | [citation slot] | Compounding demand sizing, patient count estimates, telehealth adoption | telehealth, compounders | Context for H4, H2 | T3 | Med | Estimates vary widely; no claims data |
| Novo IR statement | [citation slot] | Reaffirmed patient access programs and supply expansion commitments | NVO, Wegovy, Ozempic | Supports H1, H4 | T1 | High | Company self-reporting; verify with claims |
| Telehealth commentary | [citation slot] | Channel adaptation strategies; potential 503A workarounds discussed | telehealth platforms | Bear risk for H4 | T4 | Low | Weak signal; single source, unverified |
| PBM / payer policy | [watch] | PA requirements, step therapy barriers, employer coverage decisions pending | PBMs, employers | Bear for H2, H3 | T2 | Gap | No data yet; explains why conversion is unresolved |
Illustrative evidence pack. Public-source examples used for demonstration.
The value of Vextrum is not that it detects a headline. The value is that it converts a source event into an auditable investment object: source → claim → entity mapping → hypothesis impact → Red Team objection → analyst task → PM brief.
Confidence is not a stock call. It is evidence confidence and thesis-confidence movement. The numbers have methodology, not vanity.
Hermes is a premium intelligence agent that operates across the entire platform — querying the knowledge graph, searching evidence, proposing configuration changes, and dispatching follow-up intelligence missions.
Vextrum is not a news feed, an expert network memo, or a generic RAG tool. It is a research operating layer built around a specific investment controversy.
News tells you what happened. Vextrum tells you which hypothesis changed, which source supports it, and what action is required.
Example: FDA compounding action → updates H4, creates channel migration task, flags demand conversion as unknown.
FDA compounding action lowers one regulatory overhang but does not prove branded revenue conversion. Vextrum splits the event into separate thesis effects.
This is what makes it smarter than news/RAG.
Alt-data provides a signal. Vextrum organizes the full investment controversy: thesis, sources, entities, evidence, confidence, Red Team, and PM outputs.
GLP-1 debate requires 41 source families, not a single data feed.
Generic RAG retrieves documents. Vextrum maps evidence into an ontology and links it to hypotheses, catalysts, and analyst tasks.
CMS source update → mapped to Ozempic/Wegovy entities → updates H2 and H3 → creates model sensitivity task.
Every claim stores 12 required fields: source, tier, date, entity, hypothesis, confidence, materiality, action, limitation. No unsupported conclusions.
Each PM/team defines its own debate, universe, materiality rules, output cadence, and evidence standards. No one-size-fits-all.
Falsification rules prevent confirmation bias. "Demand ≠ reimbursed demand." Weak signals cannot move high-confidence conclusions without corroboration.
High-impact changes require approval. Low-confidence T4 sources cannot auto-upgrade investment thesis. Four-tier action gate.
The platform compounds across deployments, but each intelligence room is built around a specific decision problem and PM-defined debate.
A 12-stage intelligence pipeline: from PM onboarding through source discovery, entity resolution, ontology mapping, hypothesis scoring, Red Team falsification, and PM-ready delivery.
A 2-4 week pilot around one defined healthcare investment controversy. Bring one live research debate. Vextrum will turn it into a source-linked, evidence-weighted intelligence room.
Vextrum will convert it into a source-linked, evidence-weighted intelligence room within 2–4 weeks.