Consensus Synthesis: Eli Lilly (LLY) Proxy Analysis

---

Consensus Summary

All four models independently reached the same foundational conclusion: this is not a traditional proxy contest. The filings under review concern Eli Lilly's proposed acquisition of Centessa Pharmaceuticals plc (DFAN14A) and supplemental materials for Lilly's 2026 Annual Meeting (DEFA14A). There is no activist investor challenging Lilly's management or board. The analytical question therefore reduces to evaluating the strategic and financial merits of the Centessa acquisition and, separately, assessing Lilly's own annual meeting governance matters — the latter of which cannot be fully evaluated given the absence of substantive content in the available DEFA14A excerpt.

On the Centessa acquisition itself, the models are broadly aligned: the strategic rationale is sound, the financial terms are appropriately structured, and governance processes surrounding the transaction are adequate. Three of four models recommend supporting management, with one (Gemini) appropriately noting that the traditional "support/oppose" framing does not cleanly apply to an acquisition context.

---

Model Comparison

---

Points of Agreement

1. Correct Framing of the Filing Type

All four models correctly identified that this is an acquisition transaction rather than an activist contest. This is a critical analytical prerequisite, and unanimity here is meaningful — no model was misled by the DFAN14A filing designation into treating this as a hostile or contested situation.

2. Strategic Rationale is Compelling

Every model endorsed the core strategic logic: orexin receptor biology is well-validated, the sleep medicine market represents genuine unmet medical need, and Lilly's neuroscience infrastructure (including donanemab/Kisunla for Alzheimer's) positions it credibly to accelerate Centessa's pipeline. The characterization of cleminorexton (ORX750) as a potential best-in-class OX2R agonist across narcolepsy type 1, narcolepsy type 2, and idiopathic hypersomnia was accepted across all analyses without material skepticism.

3. Financial Terms Are Defensible

All models accepted the 40.5% premium to the 30-day VWAP as reasonable for a Phase 2-stage biopharma asset. The CVR structure ($9.00/share contingent on milestone achievements, against $38.00 upfront) was viewed consistently as an appropriate risk-sharing mechanism that avoids overpaying for speculative value in the upfront consideration while preserving Centessa shareholder participation in success.

4. Governance Process Is Adequate

Models uniformly noted that dual board approvals, the High Court sanction requirement under the scheme of arrangement structure, and the 24.1% voting support agreements from sophisticated institutional investors (Medicxi Ventures, Index Ventures, General Atlantic) represent sound governance. No model flagged material governance red flags.

5. CVR Milestone Risk Is the Primary Variable

All models that addressed risk in detail identified clinical and regulatory execution risk — particularly the IH regulatory pathway and Phase 3 success rates — as the dominant uncertainty. CNS Phase 2 to Phase 3 transition rates are historically below 50%, a risk Claude explicitly quantified and others implicitly acknowledged.

6. DEFA14A Content Gap

Claude and Grok both explicitly noted the limitation that the DEFA14A for Lilly's own annual meeting contained insufficient substantive content for individual ballot item analysis. This is a shared analytical constraint that tempers confidence across the board.

---

Points of Divergence

1. Confidence Level — Significant Disagreement

This is the most pronounced divergence. OpenAI assigned 9/10 confidence, Grok assigned 8/10, while Claude and Gemini each assigned 5/10. The gap reflects fundamentally different assessments of analytical completeness:

• OpenAI and Grok appear to weight the internal consistency of the available information — strategic alignment, financial structure, governance approvals — as sufficient to generate high confidence.
• Claude and Gemini weight the missing information more heavily: incomplete DEFA14A content, absence of Phase 3 data for cleminorexton, no independent fairness opinion details, and the inherent uncertainty of clinical-stage asset valuation. This is arguably the more rigorous analytical posture, as high confidence derived from one-sided management solicitation materials carries confirmation bias risk.

Consensus view: The lower confidence scores from Claude and Gemini are more epistemically appropriate given the information constraints. The true consensus confidence should be calibrated closer to the 5-6 range rather than the 8-9 range.

2. Framing of the Recommendation

Gemini took the principled position that recommending "Support Management" vs. "Support Activist" is a category error in this context, since there is no activist. The other three models defaulted to "Support Management" as the closest applicable label. Both positions are defensible — Gemini's framing is more analytically precise, while the other models' framing is more pragmatically actionable for institutional voters.

3. CVR Risk Quantification

Claude specifically suggested discounting CVR value by 30-50% in risk-adjusted terms, arriving at effective consideration of approximately $42-$44/share. The other models acknowledged CVR uncertainty qualitatively but did not quantify it. This level of analytical specificity represents a meaningful difference in rigor, though the underlying judgment is consistent across models.

4. Competitive Landscape Emphasis

Claude placed noticeably more weight on Takeda's TAK-861 as a competitive risk, noting that first-mover advantage is critical in rare CNS disease markets. The other models mentioned competition more briefly. This is a substantive analytical difference — competitive positioning in narcolepsy will materially affect commercial outcomes and CVR achievability.

---

Consensus Recommendation

Support Management

Strength: Moderate

Three of four models explicitly recommend supporting management's strategic direction in connection with the Centessa acquisition, and Gemini's "N/A" position does not represent opposition — it represents a framing objection rather than a substantive disagreement with the acquisition's merits. The "Moderate" rather than "Strong" designation reflects:

1. The acquisition is strategy-affirming but carries material clinical execution risk
2. The full Lilly annual meeting proxy cannot be assessed from available materials
3. High-confidence endorsement based primarily on management's own solicitation materials would be analytically unsound
4. Institutional investors should conduct independent diligence on cleminorexton Phase 2a data and competitive positioning before finalizing any votes

---

Confidence Score

Confidence: 6/10

Rationale: The directional consensus (support management / no activist to oppose) is clear and unanimous, warranting a score above the midpoint. However, the confidence ceiling is constrained by: (1) the one-sided nature of available information — all materials originate from management and lack independent counter-analysis; (2) the substantive gap in the DEFA14A content for Lilly's own annual meeting; (3) the inherent uncertainty in Phase 2-to-Phase 3 CNS development probability; and (4) the significant divergence among models on individual confidence scores, which itself signals analytical uncertainty. A score of 6/10 appropriately reflects a well-supported directional conclusion under conditions of materially incomplete information.