Rubric gaps — review and third-wave decision (2026-10-01)

Hand-written record (not generated). Owner approval: “add all. approved” (2026-10-01).

What was reviewed#

The vault had 17 rubrics (121 articles), all built on one frame: a hazardous chemical → a proven biological alternative. The review compared that taxonomy with the vendor index (1,816 companies) and with the standard scope of the bioeconomy, using:

• index topic tags on the 172 companies that had no rubric — dominated by seeds-breeding, sugar-palm-agri-processing, cea-vertical-farming, agtech-digital, plant-based, cultivated-meat, gm-gene-editing;
• keyword counts over company cards for candidate themes (rough; candidates only, not verified);
• the vault’s own trends (BIOECONOMY-TRENDS.md): trend 1 “host immunity and genetics” and trend 2 “diagnose → threshold → treat” had no collection of their own.

Decision: 9 new collections + 2 food-ingredient articles (26 articles)#

Deliberately not added#

• Human biopharma and human diagnostics — out of scope for the whole vault.
• Seaweed and algae as a rubric — they are a feedstock used across many rubrics (FI06, NU01, LE01, cosmetics, feed), so they are better treated as a topic running through those rubrics.
• Herbicide-tolerant crops — a genetic trait that enables herbicide use rather than replacing it (see the seeds-genetics skipped list).

Method used (same rules as the earlier waves)#

• Every cited supplier was checked against its own card text (Focus / Profile / official website, via tools/rubrics.card_text) and is active. Where nothing qualified, the article carries an Honest finding gap note instead of a padded table (NU04, LE01, RM02, CE01, FI08 partly, SG03 partly).
• All 95 planned journal references were checked against Crossref before writing (90 matched automatically; 1 is classed as a report only because the author name starts lower-case — du Jardin; 3 confirmed by direct DOI lookup and added to tools/hand-verified.json; 1 that could not be found in Crossref was dropped). Nightly ref_verify after the build: 439 verified / 0 check.
• Government funding signals: tools/grants_scan.py (OpenAIRE + NSF, free), candidates reviewed by hand, 2026-10-01.
• No paid API calls were made.

Tooling changes#

• tools/rubrics.py: 9 new rubric rules (topic tags + product-specific keywords; generic words such as “fertilizer”, “methane”, “greenhouse” deliberately excluded so incumbent-chemistry makers are not pulled in).
• Collection lists extended in index_check.py, ref_verify.py, supplier_index.py, build_dashboard.py, build_priority_board.py, sellable_now.py.

Supplier gaps worth a targeted rubric-seed pass (paid; owner-gated)#

1. Enteric-methane feed additives and Asparagopsis growers (LE01).
2. Struvite / recovered-phosphate producers and licensees (NU04).
3. Phytoremediation and mine-rehabilitation contractors (RM02).
4. Coir, wood-fibre and bark substrate makers (CE01).
5. Immobilized-lipase suppliers for oil interesterification (FI08).

Follow-up 2026-10-01/02: targeted supplier passes (paid, owner-approved “do all”)#

• RUBRIC_SECTORS gained 5 sectors in all three datasets (methane-feed, struvite, phytoremediation, growing-media, lipase-oils); the BIO pre-filter and the vetting scope text were widened to match.
• rubric-seed: 13 + 22 + 40 countries × 5 new sectors, $2.71, 0 errors.
• Gap-only discover (only the 5 new seed sets fed the pool; wrapper kept in the session scratchpad): vetted the top 150/150/200 candidates, 178 new companies accepted (Asia 75, LatAm 54, Africa & ME 49) → index 1,816 → 1,994. Research + status top-up: $10.00 + $7.01. Every new company has a status check; 73 still have one or more research steps incomplete (budget stop) — a later research/discover run fills them from cache at low cost.
• Results per gap:
  • CE01 peat-free media — closed: 14 coir/coco-substrate makers now cited (Sri Lanka, India, Bangladesh, Peru, Mexico).
  • RM02 phytoremediation — partly closed: 4 mine-rehabilitation/revegetation contractors + 1 public agency cited; no metal phytoextraction supplier.
  • NU04 struvite, LE01 enteric-methane additives, FI08 enzymatic interesterification — still no qualifying supplier; these look like genuine regional gaps (the technology vendors sit in Europe, North America, Australia, Japan).
• For the owner (removal is an owner decision, nothing was deleted): vetting accepted some doubtful companies from the lipase-oils and phytoremediation seeds — e.g. edible-oil mills (Golden Oil Industries, Vester Oil Mills, United Oil Company, Iran Oil Products Company), and Kenana Engineering (Sudan) whose URL is a redirect link. Giloil Company Ltd (Kenya) now has two cards (giloil-company-ltd and giloil-company-ltd-2). Use sources/overrides.json exclude if you want any of them out.
• Bug found and fixed: re-selecting an already-discovered company overwrote its hand-fixed Latin name with the raw Exa name, so two companies fell back to the degenerate id company. Fixed in discover() (all three datasets: a re-selected company keeps its existing name and aliases); Kubota Environmental Service, NCVC and the new Shandong Chuangxin Humic Acid Technology were given Latin names, reusing their caches.

Addendum 2026-10-02: rubrics derived from the bioecon.ru taxonomy comparison#

Compared the vault’s 26 rubrics with the bioecon.ru cluster spine (11 sectors → 46 clusters, hugo_skill/data/clusters.yaml) and with the companies still lacking a rubric. Result:

The 26 companies without a rubric are mostly newly discovered firms whose free keyword topic step never ran (dairies, food processors, labs, consultancies) — no new theme among them. Running exa_osint.py topics for them is the fix (keyword rules are free; Exa is called only for untagged cards).

Addendum 2026-10-02: North American suppliers close most remaining gaps#

The USA & Canada dataset (256 companies) supplied products that no Africa/ME, LatAm or Asia company offered. Cited after checking each card’s own profile/website text (not the seed-map Focus line): struvite — Ostara (NU04); Asparagopsis feed additives — CH4 Global, Blue Ocean Barns (LE01); bio-cement — Biomason, Prometheus Materials (C03); bioleaching — BacTech, Endolith, Universal Bio Mining (M01, M02); Wolbachia — MosquitoMate (E02); beet de-icers — K-Tech (E04); fermentation equipment — Paul Mueller, PBS Biotech (EQ02); plus 21 rows in other articles. Still open everywhere: enzymatic interesterification lipases (FI08), classical weed biocontrol (A06–A09), Coniothyrium (A12). The regional gaps remain for the other three regions — which matters under the RU/EAEU territory option (see TARGET-MARKET-DECISION.md).