Verdict: Thesis confirmed. The 2026-04-12 blind trace got most of the merchant supply-chain right at the first cut (HPLC purification → chromatography resins → injectable closures → pharma glass). The single biggest gap — the peptide CDMO layer (Bachem + Lonza, the two companies that actually synthesize the peptide for the principals) — was missed by the blind trace but closed by subsequent ambient-browse research between April and May 2026. The coverage state today is two paired watchlists (trace-glp1-manufacturing for the picks-and-shovels chain; peptide-economy for principals + next-gen pipeline + telehealth distribution); no active perspective exists, and the perspective decision (secular-obesity-buildout vs glp1-price-control-scenario vs stay watchlist-only) is deferred to user per the open [?] task in TASKS-RESEARCH.md:105.

What we're asking

Backlog item from research/projects/thesis-ideas.md:21 (Priority 1, BLIND TEST):

Thesis: GLP-1 drugs (Ozempic, Wegovy, Mounjaro) are in massive shortage. What physically bottlenecks production? Why blind: We cover LLY in our scans but know nothing about the peptide manufacturing supply chain.

Specific questions:

• What's the physical manufacturing process for GLP-1 peptides?
• Who makes the API (active pharmaceutical ingredient)?
• Who makes the injection devices (auto-injectors, pens)?
• Who makes the fill-finish (putting drug in vials/pens)?
• Where are the bottlenecks?

What we already had before today

This is post-trace consolidation. The 2026-04-12 trace ran blind; subsequent April–May 2026 research substantially filled in the gaps.

The physical manufacturing chain — what GLP-1 peptides actually need

GLP-1s (semaglutide, tirzepatide, dulaglutide) are chemically synthesized peptides — chains of ~30–40 amino acids made by solid-phase peptide synthesis (SPPS), then purified, formulated, fill-finished, and delivered via auto-injector or oral tablet.

The chain has six discrete bottleneck layers, in production order:

The bottleneck the trace nailed was HPLC + chromatography resins (Layers 3 — WAT + DHR/Cytiva — unanimous). The bottleneck the trace missed was Layer 1–2 — the peptide CDMO synthesis (Bachem + Lonza). Subsequent ambient research closed this via the dumb-money Discord BCHMY mention (Apr 14) and the picks-and-shovels research-picks run (May 5) which surfaced LZAGY.

State-of-coverage by ticker

Pricing not pulled fresh; this verdict is structural-state, not entry-timing. The watchlists already track day-to-day tape.

Picks-and-shovels (trace-glp1-manufacturing.json)

Principals + pipeline + telehealth (peptide-economy.json)

Verdict and reasoning

Manufacturing chain: CONFIRMED structurally. The 6-layer chain (Fmoc feedstock → SPPS synthesis → HPLC purification → bulk-drug-substance handling → fill-finish → devices) is real, well-documented, and matches what the merchant supply-chain literature says. Three layers are durably capacity-constrained: (a) Layer 5 (fill-finish — pharma glass + injectable closures) is the most-cited constraint in trade press because greenfield pharma-glass capacity has 4–6 year lead times; (b) Layer 2 (peptide CDMO synthesis) is constrained by the principals' decisions to build captive (NVO Hillerød $4B+ expansion) vs outsource to Bachem/Lonza; (c) Layer 3 (HPLC purification) is the yield-limiting step in SPPS economics.

Bottleneck name selection: CONFIRMED. WAT + DHR + WST + STVN are the right names at the right layers. The trace got 4/6 chain layers right on the first blind cut. STE, LZAGY, BANB.SW filled the remaining gaps via subsequent ambient research without needing a re-trace — the system did the work via the FOLLOWS / dumb-money-discord / picks-and-shovels CLI loop.

Coverage state today: COMPLETE for the merchant chain. Two paired watchlists cover the supply-chain (trace-glp1-manufacturing) and the principals/pipeline/distribution (peptide-economy). No major chain-layer gap remains. The merchant auto-injector layer (Phillips Medisize / SHL / Ypsomed) is private; BDX is the public proxy and that's likely the best we can do.

Perspective decision: DEFERRED to user. The open question TASKS-RESEARCH.md:105 lists three options — secular-obesity-buildout (multi-year cycle perspective on the duopoly + next-gen orals + RFK 503A peptide-list catalyst), glp1-price-control-scenario (Medicare price-cap scenario pair), or stay watchlist-only. The decision shape is perspective capacity, not coverage — the two watchlists are already paired lenses. This investigation does not make the perspective call.

Cultural-arbitrage cross-paradigm: OPEN. The Looksmaxxing × BANB cross-paradigm question (TASKS-RESEARCH.md:49) — whether cultural-demand signals sharpen the supply-chain thesis — is open and not in scope here.

RFK 503A peptide bulk-list catalyst: on-calendar. July 2026 FDA Pharmacy Compounding Advisory Committee meeting on whether peptides land on the 503A bulk list. Material outcome for HIMS specifically (compounded GLP-1 at scale via the Feb 2025 CA peptide facility acquisition); secondary for the wider chain. Listed in peptide-economy.json description; not its own task.

Coverage gaps / follow-ups

Zero new tasks filed. The chain is covered; the open decisions (perspective shape, cultural-arbitrage cross-paradigm) are already filed as [?] user-decisions in TASKS-RESEARCH.md. Specifically:

• TASKS-RESEARCH.md:49 — BANB.SW × Looksmaxxing cross-paradigm (open).
• TASKS-RESEARCH.md:105 — Peptide / GLP-1 economy perspective decision (open).

Both surface here for context but aren't re-filed.

Deliberately not filed (transparency):

• Captive vs merchant CDMO split deep-dive — would answer "what fraction of GLP-1 API is now in-house vs. Bachem/Lonza?" Useful but speculative; defer until a perspective decision lands.
• Pen-device merchant layer (Phillips Medisize / SHL / Ypsomed) coverage check — all private; no investable proxy beyond BDX at the needle layer. Not worth filing.
• RFK 503A peptide bulk-list calendar tracker — already implicit in peptide-economy.json description; not its own task until July 2026 approaches.

What didn't get done (transparency)

• No fresh OHLC fetch on the chain. Day-to-day tape lives in the watchlists; this is a structural verdict, not an entry-timing piece.
• No primary-source filing extraction (NVO 20-F, LLY 10-K, Bachem annual report). The chain-layer claims here are corroborated by trade-press and the prior trace; primary-source extraction is deep-dive work, not investigation work.
• No perspective creation. The open [?] user-decision (TASKS-RESEARCH.md:105) is deliberately left for user.
• No cross-paradigm decision. The Looksmaxxing × BANB cross-paradigm task stays open.
• No competitive comparison vs Saxenda generic biosimilars / Hims compounded GLP-1 vs branded. The thesis is principal-and-supply-chain; biosimilar/compounded competitive dynamics are a different layer.

Receipts

Source backlog item:

• research/projects/thesis-ideas.md:21 — Priority 1 Blind Test #2 (GLP-1 / Obesity Drug Manufacturing). To be updated to RESOLVED with verdict pointer.

Cited and consolidated:

• research/projects/autonomous-research/traces/2026-04-12-glp1-peptide-manufacturing-bottleneck.json — original blind trace (WAT/DHR unanimous; WST/STVN strong; BCHMF/TMO lone).
• stonks/watchlists/trace-glp1-manufacturing.json — 9-ticker picks-and-shovels watchlist.
• stonks/watchlists/peptide-economy.json — 10-ticker principals + pipeline + telehealth watchlist.
• TASKS-RESEARCH.md:49 — open cross-paradigm task (Looksmaxxing × BANB).
• TASKS-RESEARCH.md:105 — open perspective decision.
• Memory references: reference_realjgbanks_skip.md, reference_intheassembly_skip.md (skip-class accounts that surfaced peptide names; chain confirmed independently on substantive sources).

Filed follow-ups (this investigation):

• None new. Two [?] user-decisions already filed cover the open work.

Commands run for this artifact:

deno task beta:emit investigation \
  --topic "GLP-1 / peptide-economy manufacturing supply-chain blind-test verdict" \
  --slug "glp1-peptide-manufacturing-blind-test-verdict" \
  --tickers "WAT,DHR,WST,STVN,BDX,TMO,STE,LZAGY,LLY,NVO" \
  --watchlists "trace-glp1-manufacturing,peptide-economy"
# Note: BANB.SW referenced in prose; emitter validator generally drops foreign-suffix tickers from affects.tickers.