The full method behind every Ostanes Intelligence report. Three-agent search, CPC code validation, chemical formula extraction, convex-hull analysis. No black boxes.
Why standard patent reports miss what matters
Patent landscape reports answer two questions well: who is filing and when. They produce assignee rankings, filing-trend curves, jurisdiction maps, technology-class heat-grids. For a portfolio-level view of competitive activity, that is genuinely useful work.
It is also incomplete in a way that matters more in deep-tech than anywhere else. Filing trends and assignee counts cannot tell you whether the science behind the claims is sound. A surge of filings in a materials class can mean a real technical breakthrough, or it can mean dozens of teams chasing a thermodynamic dead-end because the underlying chemistry was never tested against first principles.
For a VC writing a £5M cheque into a portfolio company, an R&D director allocating a year of bench time, or an IP attorney advising on freedom-to-operate, that distinction is the entire question. The standard report cannot answer it.
The Science-Patent Link
Ostanes Intelligence runs a four-stage pipeline that fuses patent data with computed materials physics and the published scientific record. The full process is below; every report we deliver is built on this foundation.
⬡ The Science-Patent Link · 4-stage pipeline
01 / SEARCH
Three-agent strategy
A Strategist, Critic, and Resolver agent triangulate on CPC codes and keyword variants. Reduces false negatives that single-prompt retrieval misses.
02 / FETCH
Multi-jurisdiction patent retrieval
Filings pulled from EPO, USPTO, and Google Patents BigQuery using the validated CPC codes. Full-text claims and abstracts retained.
03 / LINK
Claim-to-physics binding
Chemical formulae extracted from claims, looked up in the Materials Project for computed thermodynamic properties (formation energy, energy above hull, band gap, stability).
04 / SYNTHESISE
LLM-assisted pattern surface
Correlations, white spaces, implausibility flags, and inflexion points surfaced. Outputs reviewed by the founder before delivery (no auto-publish).
The four stages are sequential but not linear: stage 03 frequently sends signals back to stage 01 when extracted formulae reveal the original CPC search was too narrow, and the loop runs until the technical evidence stabilises. A typical Inflexion Report cycles through this loop three to five times before synthesis.
Two technical decisions shape what comes out of this pipeline. First, we treat energy above the convex hull as the authoritative stability signal rather than relying on inferred properties from the patent text. Second, we never publish a synthesis without founder review: the LLM surfaces patterns, and a human signs off on the conclusions.
A worked example: solid-state battery electrolytes
Solid-state battery (SSB) electrolytes are the cleanest illustration of why this method matters. The patent landscape since 2020 has been dominated by three material families: argyrodite-type sulfides (Li₆PS₅Cl and variants), garnet-type oxides (Li₇La₃Zr₂O₁₂, "LLZO"), and the orthorhombic phosphate β-Li₃PS₄.
By patent count alone, Li₆PS₅Cl is the runaway leader: it appears in well over half of the active SSB electrolyte filings across the three jurisdictions we index. A standard landscape report would conclude this is the consensus winner.
The Materials Project tells a different story. Li₆PS₅Cl sits 0.083 eV/atom above the convex hull. That is not catastrophic (it is well within the range where kinetic stability can compensate for thermodynamic metastability), but it does mean every patent claiming long-term operational stability for argyrodite-based cells is making a claim the calculated thermodynamics does not, on its own, support. LLZO at 0.041 is in much better shape. β-Li₃PS₄ sits exactly on the hull and is the least patented of the three.
The most-patented material in the field is also the least thermodynamically stable. The most stable is the least patented.
That is not a verdict. Argyrodite may well be the right answer once interfacial engineering and ionic conductivity are weighed in. But it is the kind of finding that should change how a VC due-diligences a portfolio company's electrolyte choice, how an R&D director scopes a multi-year programme, and how an IP attorney frames a freedom-to-operate opinion. None of that signal is visible in a standard landscape report.
What we do not do
Ostanes Intelligence does not provide legal advice. We do not opine on patentability, infringement, validity, or freedom-to-operate as legal conclusions. OSTANES LTD is not a registered patent attorney, and nothing in our reports should be read as a substitute for one.
What we provide is the technical analysis layer that makes legal advice more precise. Patent attorneys make the legal calls. We surface the scientific evidence that informs those calls, and we surface it before the IP landscape closes around a flawed assumption.
If you can tell us the technology question, we can tell you whether the science holds. The legal call belongs to your counsel. The technical truth is ours.