A small research outfit named Coracle Research published a 50-page technical disclosure on April 25, under 35 U.S.C. §102 ↗, making the architecture for an oral metabolic-peptide tablet permanent prior art. The document is a defensive patent publication, not a patent application. Its purpose is the opposite of a patent: to put the design into the public domain before any company can claim it. The architecture covers seven drugs (semaglutide ↗, liraglutide ↗, tirzepatide ↗, exenatide, retatrutide ↗, survodutide ↗, cagrilintide ↗) and predicts oral bioavailability of 2 to 7 percent, which is two to seven times what Rybelsus ↗ achieves today.
What a defensive publication is
Patents grant a monopoly. Defensive publications do the inverse. A document that meets the disclosure standards of 35 U.S.C. §102, dated and timestamped, becomes prior art. Once it is prior art, no one (including the original author) can later patent the same idea. Coracle's filing is also deposited as a Zenodo preprint ↗ under Creative Commons, which establishes the publication date in a venue independent of any future legal challenge.
The Rybelsus problem
Oral semaglutide (Rybelsus) is the only oral GLP-1 drug on the market. It uses a single absorption-enhancer strategy, sodium N-(8-[2-hydroxybenzoyl]amino)caprylate (SNAC), which carries the peptide across the gastric epithelium through the cells (transcellular). The bioavailability is roughly 1 percent. Patients have to swallow 14 mg orally to match 1 mg subcutaneous. So far, nobody has shipped a better oral architecture for this drug class.
The architecture
Coracle's disclosure stacks six subsystems in a single tablet. The mechanism trick is route-additivity. Transcellular absorption (through cells, the SNAC mechanism) and paracellular absorption (between cells, by opening tight junctions) are physically orthogonal pathways. Their bioavailabilities add. Rybelsus uses one route. This architecture uses both, plus mucus fluidization to thin the diffusion barrier, plus mucoadhesive chitosan to extend gastric residence two to four hours, plus a PLGA-PEG nanoparticle carrier as the payload vehicle.
The six components, by dose: SNAC at 200 to 400 mg for the transcellular route; CAGE ionic liquid (choline and geranate) at 30 to 80 mg for paracellular tight-junction opening and mucus fluidization; a PIP640 analog decapeptide at 2 to 10 mg as a selective claudin-2 modulator; a 50 to 200 nm PLGA-PEG mucus-penetrating nanoparticle carrier; a low-molecular-weight chitosan outer layer for mucoadhesion; and an optional pulsed-field ingestible device as a secondary embodiment. Predicted class-level nominal bioavailability is approximately 3.6 percent, with peptide-specific bands running 4 to 14 percent theoretical and 2 to 7 percent effective pre-bench.
What the disclosure does to the patent landscape
The document is engineered around existing claims. Tablet mass is specified at 650 mg or above, which sits outside Novo Nordisk's "at least 50 percent SNAC" composition floor in the Rybelsus patent family. Disintegration time is specified at 23 minutes or longer, which sits outside Novo's "22 minutes or less" continuation ceiling. A "Mitragotri-independent variant" is included for every peptide, which drops CAGE and substitutes augmented tight-junction modulation plus trimethyl chitosan, neutralizing any need for licensing from the Cage Bio / Harvard ionic-liquid patent family. Each peptide carries its own dose envelope (semaglutide 0.25 to 5 mg daily, tirzepatide 0.5 to 75 mg daily, retatrutide 0.5 to 75 mg daily, survodutide 0.1 to 30 mg daily, cagrilintide 0.125 to 15 mg daily, exenatide 0.01 to 7.5 mg two to three times daily), and each one carries a claim explicitly covering manufacture by any party other than Novo Nordisk as a generic oral tablet. Liraglutide became off-patent in August 2025; the disclosure now makes a generic oral liraglutide using this stack explicitly unblockable.
A note on the physics. Section 10.5 of the disclosure derives transcellular bioavailability from the fine-structure constant α = 1/137 and a cubic-packing symmetry argument for lipid bilayers, arriving at F_transcellular = 4α ≈ 0.029 and matching Rybelsus' published F within 1.4 percent. Either an unusually clean dimensional argument or numerology that landed on the right answer. The bench data, when it exists, will adjudicate.
Why this matters for the open peptide-therapeutic stack
We framed the wet-lab democratization arc yesterday ↗ as open weights, open compute, open data, open wet lab. Defensive publication is the same move applied to the patent layer. The closed-pharma advantage on oral GLP-1 has been a moat made of formulation IP, not biology. Coracle has put a candidate architecture for that moat into the public domain on the same week that a Nature Synthesis paper documented a $5,000 self-driving lab and a hobbyist sequenced their own genome with Claude. PeptideModel hosts the upstream peptide candidates: retatrutide is on the platform as pep-00018 (a triple agonist on GLP-1R ↗, GIPR ↗, and GCGR ↗), the GLP-1R corpus alone holds 224 cards, cagrilintide sits on CALCR ↗, and oxyntomodulin and glucagon variants populate GCGR. The open stack now has a legal node alongside the modeling, hardware, and protocol nodes.
Caveats
The numbers in the disclosure are predictions, not bench data. Coracle has filed claims and provided dose envelopes and failure-mode analysis at patent-grade specificity, but no in vivo bioavailability has been reported. A 4 to 14 percent theoretical figure can become 2 percent or 0.2 percent on the bench. The disclosure's value is structural, not experimental: it removes a future patent obstacle for any group, generic manufacturer, or compounding pharmacy that wants to develop oral versions of these drugs. Whether the architecture works in a tablet someone can actually take is a separate question that the disclosure does not answer.
What to watch
Three things. First, whether any independent group runs the architecture against retatrutide or tirzepatide in a Caco-2 model and posts the data publicly. Caco-2 is the standard in vitro screen for oral bioavailability and would partially validate the route-additive claim. Second, whether Novo or Lilly references the disclosure in subsequent patent filings; a defensive publication's effectiveness is partly measured by whether competitors stop trying to patent around it. Third, whether the off-patent generic liraglutide market sees this architecture appear in an actual product. If a compounder ships an oral liraglutide tablet using the disclosed stack, the legal moat will have been measurably breached.