chore: daily update 2026-07-30

sources/git-repo.md

@@ -11,14 +11,17 @@ the vault, not summarize their content.
 - **Instance:** `obsidian-robin` (source id `obsidian-robin-1`)
 - **Repo path:** `/mnt/c/Users/20864/.openclaw/workspace`
 - **Branch:** `main`
-- **HEAD:** `21491578ec22733f15efda864e88e44823171e6a` (2026-07-28)
-- **Ahead of origin/main:** 7 commits (3 CellShot-TriArm-Demo plans + 4 new commits
-  dated 2026-07-28: Hodge 2025 / Lee 2025 / Parate 2020 deep reads + 董慧群
-  CelVivo distributor technical-contact profile, un-pushed as of this run).
+- **HEAD:** `1a81e485b6f8a4496122464f1151702d09f03cf0` (2026-07-29)
+- **Ahead of origin/main:** 9 commits (3 CellShot-TriArm-Demo plans + 4 dated
+  2026-07-28: Hodge 2025 / Lee 2025 / Parate 2020 deep reads + 董慧群
+  CelVivo distributor technical-contact profile + 2 dated 2026-07-29:
+  `81671c69` Organoids 2026 综述 feat deep-read + `1a81e485` SOP-rewrite
+  pass on the same file — un-pushed as of this run).
 - **Auth:** local path only — no remote, no credentials
-- **Latest fetch:** `2026-07-28T18:07:57Z` (HEAD advanced from `62877d1f` → `21491578`,
-  +4 commits; untracked working-tree list unchanged from the prior pull)
-- **Raw manifest:** `~/.openwiki/connectors/git-repo/raw/2026-07-28T18-07-57-114Z/manifest.json`
+- **Latest fetch:** `2026-07-29T18:07:33Z` (HEAD advanced from `21491578` →
+  `1a81e485`, +2 commits; untracked working-tree list unchanged from the
+  prior pull — same 5 carryover files + submodule `-dirty` suffix)
+- **Raw manifest:** `~/.openwiki/connectors/git-repo/raw/2026-07-29T18-07-33-803Z/manifest.json`
 - **Source config:** `~/.openwiki/connectors/git-repo/config.json`
 
 ## Working-tree status
@@ -49,17 +52,20 @@ the vault, not summarize their content.
     cheat sheet ("一天 6 次对话的完整沉淀"). Date stamp 2026-07-27.
     *(carryover from prior pull)*
 
-Since the prior fetch (2026-07-27T18:07:54Z), **4 new commits have landed on
-`main`** — HEAD advanced from `62877d1f` (2026-07-22) to `21491578`
-(2026-07-28). All four are dated 2026-07-28 and are well inside the 7-day
+Since the prior fetch (2026-07-28T18:07:57Z), **2 new commits have landed on
+`main`** — HEAD advanced from `21491578` (2026-07-28) to `1a81e485`
+(2026-07-29). Both are dated 2026-07-29 and are well inside the 7-day
 "vault fresh" window, so the staleness rule does not fire. Working-tree
 deltas this run: the `skills/openclaw-workspace` submodule `-dirty` suffix
 (unchanged commit hash; no content delta) plus the same 5 untracked files
-the prior pull already recorded. The three new 2026-07-28 vendor-paper
-deep reads landed as **committed** files this pull, not as untracked working
-files — they appear in the Recent commits table below, not above.
-Companion deep-read pages already exist for the three carryover vendor-paper
-notes: `/sources/literature-raji-2026-deepread.md`,
+the prior pull already recorded. The two new 2026-07-29 commits are a
+paired add+rewrite (`81671c69` feat deep-read of "Organoids for disease
+modeling and treatment 2026 综述" PMID 41545895 / PMC12828948 into
+`04-Literature/Deep-Reads/Organoid/`, then `1a81e485` fix pass to rewrite
+the same file using the `literature-deep-read-review` SOP — 按主题分块 +
+三段式 + 落款). The Organoid/ bucket is now +1 since this run lands
+one new committed file. Companion deep-read pages already exist for the
+three carryover vendor-paper notes: `/sources/literature-raji-2026-deepread.md`,
 `/sources/literature-tasevska-2025-deepread.md`, and
 `/sources/literature-cel-bosnjakovic-2025-pah-deepread.md`.
 
@@ -94,7 +100,7 @@ Markdown file counts per deep-read subfolder:
 
 | Subfolder | .md count |
 | --- | --- |
-| `Organoid/` | 658 |
+| `Organoid/` | 659 |
 | `Celvivo/` | 142 |
 | `3Brain/` | 136 |
 | `Cellink/` | 129 |
@@ -111,11 +117,13 @@ Markdown file counts per deep-read subfolder:
 | `Kidney/` | 2 |
 | `Retinal/` | 1 |
 
-Total ≈ 1,201 deep-read notes (+3 committed 2026-07-28 vendor/adjacent
-deep-reads since the prior pull: +2 `Other/` — Lee 2025 Front Vet Sci
+Total ≈ 1,202 deep-read notes (+3 committed 2026-07-28 vendor/adjacent
+deep-reads since the pull before that: +2 `Other/` — Lee 2025 Front Vet Sci
 canine MSC 3D spheroid + Parate 2020 SCRT PEMF MSC paracrine cartilage;
 +1 new `RegenMed/` bucket — Hodge 2025 Bio-Block MSC 4-week comparison;
-the Organoid/ bucket is unchanged since none of the new files land there).
+the Organoid/ bucket is now +1 from this run's two new commits
+(`81671c69` Organoids 2026 综述 feat deep-read + `1a81e485` SOP-rewrite
+of the same file) — neither `Other/` nor `RegenMed/` changed this run).
 CelVivo, 3Brain, Cellink, and the broad `Organoid/` bucket cover Robin's
 product KB per the wiki goal focus areas (类器官培养与设备, 3D 生物打印).
 The new `RegenMed/` bucket is application-domain adjacent (Bio-Block vs.
@@ -129,6 +137,8 @@ Pulled from the manifest's `recentCommits` field. Date stamps via
 
 | Date | Hash | Note |
 | --- | --- | --- |
+| 2026-07-29 | 1a81e485 | fix(literature): Organoids 2026 综述 - 按 literature-deep-read-review SOP 重写 (按主题分块 + 三段式 + 落款) |
+| 2026-07-29 | 81671c69 | feat(literature): 精读 Organoids for disease modeling and treatment 2026 综述 (PMID 41545895, PMC12828948) |
 | 2026-07-28 | 21491578 | feat(agent): 新增代理商技术同事董慧群档案 (推 CelVivo 产品) |
 | 2026-07-28 | 652e0e60 | feat(literature): Parate 2020 SCRT — PEMF 增强 MSC 旁分泌功能促软骨再生 (中文精读, 10 图 + 4 功能) |
 | 2026-07-28 | ee34cf83 | feat: 精读 Lee 2025 Front Vet Sci 犬脂肪源 MSC 3D 球体（Wiki.js Id=1645, NocoDB Id=3290） |
@@ -189,6 +199,14 @@ in the `git status` dirty tree):
 | `04-Literature/Deep-Reads/Other/2026-07-28_Lee_2025_FrontVetSci_canine_MSC_3D_spheroid.md` | added (commit `ee34cf83`) | Lee 2025 Front Vet Sci — 犬脂肪源MSC 3D球体 (Wiki.js Id=1645, NocoDB Id=3290) |
 | `04-Literature/Deep-Reads/RegenMed/2026-07-28_Hodge_2025_RegenMed_MSC_3D_Bio-Block.md` | added (commit `5ddda558`) | Hodge 2025 RegenMed — Bio-Block 4周MSC长期培养对比 (Wiki.js + SeaFile 7图 PDF) |
 
+The 2 commits added on 2026-07-29 brought this file into `main` (visible
+in `git log` and the manifest's `recentCommits` field, not in the
+`git status` dirty tree):
+
+| Path | Change | Note |
+| --- | --- | --- |
+| `04-Literature/Deep-Reads/Organoid/2026-07-29_Organoids_disease_modeling_treatment_2026_review.md` | added (commit `81671c69`) then modified (commit `1a81e485`) | Organoids for disease modeling and treatment 2026 综述 (PMID 41545895, PMC12828948); commit `1a81e485` is the SOP-rewrite pass (按主题分块 + 三段式 + 落款) |
+
 ## Notable entities surfaced by the vault (structural hints only)
 
 - **CelVivo industrial customers program** — a dedicated subfolder under

sources/hackernews.md

@@ -8,10 +8,26 @@ Compact evidence trail for the Hacker News connector (`hackernews-1`, Robin-cura
 - **Backend:** public HN feeds (`top`, `new`) + Algolia HN search queries
 - **Auth:** none — public feeds
 - **Window:** last 24 hours
-- **Latest fetch:** `2026-07-28T18:00:42Z` (latest 18:00 pull on 2026-07-28; prior 2026-07-27 18:00, 2026-07-26 18:16 and 18:00 pulls also documented below)
-- **Raw dumps:** `~/.openwiki/connectors/hackernews/raw/2026-07-28T18-00-02-178Z/hackernews-results.json` (current) · `~/.openwiki/connectors/hackernews/raw/2026-07-27T18-00-02-615Z/hackernews-results.json` · `~/.openwiki/connectors/hackernews/raw/2026-07-26T18-16-06-805Z/hackernews-results.json` · `~/.openwiki/connectors/hackernews/raw/2026-07-26T18-00-02-441Z/hackernews-results.json`
+- **Latest fetch:** `2026-07-29T18:00:02Z` (latest 18:00 pull on 2026-07-29; prior 2026-07-28 18:00, 2026-07-27 18:00, 2026-07-26 18:16 and 18:00 pulls also documented below)
+- **Raw dumps:** `~/.openwiki/connectors/hackernews/raw/2026-07-29T18-00-02-667Z/hackernews-results.json` (current) · `~/.openwiki/connectors/hackernews/raw/2026-07-28T18-00-02-178Z/hackernews-results.json` · `~/.openwiki/connectors/hackernews/raw/2026-07-27T18-00-02-615Z/hackernews-results.json` · `~/.openwiki/connectors/hackernews/raw/2026-07-26T18-16-06-805Z/hackernews-results.json` · `~/.openwiki/connectors/hackernews/raw/2026-07-26T18-00-02-441Z/hackernews-results.json`
 
-## Latest update — 2026-07-28 18:00 UTC
+## Latest update — 2026-07-29 18:00 UTC
+
+- **Coverage:** 2 feeds (`top`, `new`), 24 search queries (the same life-science + workflow set the 2026-07-28 18:00 pull used), 27 + 30 = 57 unique resolved feed stories, and the life-science / vendor / workflow lane hit table below.
+- **Life-science lane:** **zero search-query hits** across all 14 life-science / vendor queries — `organoid`, `3D bioprinting`, `3D bioprinter`, `bioprinting clinical`, `spheroid bioreactor`, `microgravity cell culture`, `CELVIVO`, `3Brain`, `Cellink`, `MEA neuroscience`, `organ-on-chip`, `liver organoid`, `brain organoid`, `spheroid drug screening` all returned no results. A title/url regex sweep across the top + new feeds also turned up zero organoid / 3D-bioprinting / CelVivo / 3Brain / Cellink / ClinoStar / ClinoReactor / MEA / spheroid / organ-on-chip items. The prior "no material HN footprint for the 4 vendor lines" finding still holds.
+- **Workflow lane:** **zero search-query hits** across all 10 workflow queries — `OpenWiki`, `OpenWiki personal brain`, `Hermes Agent`, `Claude Code agent workflow`, `Codex CLI`, `RAGFlow`, `agent memory wiki`, `agent personal brain`, `MCP connector agent`, `MiniMax` all returned no results. The feed sweep caught the recurring Kimi/K3 cluster (item `49098130` "Self-hosting Kimi K3: 20% more hardware cost, 20% better task resolution", 51 pts) — off-scope per brief (open-weight AI never earns a themes.md row).
+- **High-engagement top-feed items not promoted:** `49098510` "Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac" (317 pts) — local-LLM / hardware signal, off-scope; `49096188` "Document-borne AI worms can self-propagate through Copilot for Word" (249 pts) — security, off-scope; `49096969` "Handbook.md shows that long policy documents do not reliably govern agents" (207 pts) — adjacent to the existing `agent-memory-architecture` watchlist (a generic empirical result about agents and long policy docs), but not a direct change to Hermes / MiniMax / OpenWiki / RAGFlow / NocoDB / SeaFile, so it stays on the adjacent-not-promoted list rather than earning a themes.md row per the brief's "Workflow tooling (priority LOW) — surface story-worthy changes only" rule; `49098466` "Hugging Face: Anatomy of a frontier-lab agent intrusion" (87 pts) — security, off-scope; `49098198` "A.I. companies are recruiting electricians and carpenters by the thousands" (63 pts) — labor market, off-scope; `49098388` "GPT-5.6 vs. Claude Fable 5 for Physical AI" (41 pts) — model comparison, off-scope; `49099143` "Launch HN: Tokenless (YC S26) – Automatic model switching to save money" (32 pts) — generic YC launch, off-scope; `49098260` "Show HN: Qwen Scribe – local transcription and dictation for Apple Silicon" (27 pts) — local inference, off-scope; `49100035` "Infrastructure Patterns for Agentic Applications" (23 pts) — generic agent infra, off-scope; `49097908` "CipherX applies painless permanent tattoos with dissolving microneedle patches" (17 pts) — microneedle patches adjacent to bioprinting but the application is cosmetic tattooing, not 3D bioprinting, so off-scope per brief.
+- **New-feed items not promoted:** all 30 items were either 1–6 pts, off-scope news (Walmart EV charging, Next.js .next folder cleanup, Spiderman, Edible ants, Federal employees, etc.), or already off-scope clusters (rogue agent security at `49100529`, AI rental faces in China at `49100598`, AI on lost languages at `49100590`, the `49100656` "Give every AI project its own computer" generic agent post). None matches a Robin in-scope keyword.
+- **Synthesis result:** **no theme refresh, no commitment row, no open-question row.** `class-organoid-3d-bioprinting` last-seen stays at 2026-07-28 (item `49083799` from the prior pull — this 24h window is empty of organoid signal). The Kimi/K3 cluster continues to dominate the workflow lane at moderate engagement but stays off-scope per the brief.
+
+| Query lane | Queries run | Qualifying results |
+| --- | --- | --- |
+| Life sciences / vendors | `organoid`, `3D bioprinting`, `3D bioprinter`, `bioprinting clinical`, `spheroid bioreactor`, `microgravity cell culture`, `CELVIVO`, `3Brain`, `Cellink`, `MEA neuroscience`, `organ-on-chip`, `liver organoid`, `brain organoid`, `spheroid drug screening` | 0 |
+| AI workflow | `OpenWiki`, `OpenWiki personal brain`, `Hermes Agent`, `Claude Code agent workflow`, `Codex CLI`, `RAGFlow`, `agent memory wiki`, `agent personal brain`, `MCP connector agent`, `MiniMax` | 0 |
+
+Raw dump: `~/.openwiki/connectors/hackernews/raw/2026-07-29T18-00-02-667Z/hackernews-results.json`.
+
+## Update — 2026-07-28 18:00 UTC
 
 - **Coverage:** 2 feeds (`top`, `new`), 24 search queries (the same life-science + workflow set the 2026-07-27 18:00 pull used), 58 unique resolved feed stories, and the life-science / vendor / workflow lane hit table below.
 - **Life-science lane:** `organoid` returned 6 hits — one genuinely relevant (`49083799`, owlposting drug-discovery essay, 9 pts — see `class-organoid-3d-bioprinting (refreshed this run)` below); the rest were string collisions (Countify no-code computer vision, Shoebox Plus, "Bonfire" communities, anti-fraud, plus the YC S26 Rise Reforming launch also caught by the `Cellink` query). `3Brain` returned 8 hits (all string collisions: aeon bioenergetics essay, SpiNNaker2 chip, "local-first second brain" repo, "Use Your Brain" engineering blog, "Rachel has a new brain" reddit post, "Brain waves as noise" essay). `Cellink` returned 3 hits (Nutaan voice assistant, Rise Reforming — same as `organoid`, and a Kimi-K3 downloader). `liver organoid` and `brain organoid` each returned 1 hit — both string collisions (Shoebox Plus, Countify). `CELVIVO`, `3D bioprinting`, `3D bioprinter`, `bioprinting clinical`, `spheroid bioreactor`, `microgravity cell culture`, `MEA neuroscience`, `organ-on-chip`, `spheroid drug screening` all returned `nbHits=0` — the prior pulls' "no material HN footprint for the 4 vendor lines" finding still holds.

sources/web-search.md

@@ -225,3 +225,97 @@ The `agent personal memory wiki 2026` query returned five same-week sources on a
 - **No new vendor-anchored evidence** on CelVivo, 3Brain, Cellink, CellShot/TriArm, or Hermes this pull. The prior vendor-anchored items (CelVivo+TheWell, Hermes v0.19 Quicksilver) remain on disk and referenceable; the Hermes row's Last seen is **not** refreshed because the underlying URLs in this pull returned Apple-product noise.
 - **Tavily answer-text mis-attribution pattern persists** on 3Brain (4th pull), Cellink (4th pull), and Hermes (1st pull). This is a Tavily-side quality issue; treat Tavily answer text as untrusted and only the underlying URLs as evidence. Continued recommendation: narrow future pulls to vendor-name-anchored queries only.
 - **OpenWiki / RAGFlow / agent-memory lane behavior changed.** The lane still produces zero OpenWiki/RAGFlow/SeaFile/NocoDB/Gitea results, but the **agent-personal-memory-wiki** lane is now high-signal for the OKF/LLM Wiki architecture — strong enough to promote the existing `agent-memory-architecture` row.
+
+---
+
+## 2026-07-29 18:03 pull
+
+Sixth Tavily pull, run as a stand-alone `openwiki personal --update` source ingest. All 10 queries re-run. Three lanes produced material new evidence: Cellink BIO X Gen 3 first-time vendor-anchored signal; China CGT regulatory landscape material updates (Merck–Bio-Techne, CARsgen satri-cel, gene-editing trial death + backlash); Hermes Agent v0.19 documentation continues (configuration.md, api-server.md). The CelVivo/TheWell xeno-free and China Order No. 818 items remain pinned in `/commitments.md`. The agent-memory-architecture cluster continues to grow. Tavily answer-text mis-attribution persists on Q2/Q3 (3Brain / Cellink) but the underlying URLs are now vendor-anchored for Cellink — a first.
+
+### Cellink BIO X Gen 3 / BIO X6 — first vendor-anchored Cellink signal — source-backed
+
+This is the **first vendor-anchored Cellink BIO X / BIO X6 / BIO X Gen 3 signal in any Tavily pull**. Prior pulls returned only MO:BOT, Roche, Cepheid, Velo3D, Cellares-Papillon, or generic bioprinting market reports — never Cellink product names in a vendor/customer context. Three same-week sources confirm Cellink product activity:
+
+- **LinkedIn (Nuclear Mechano-Oncology), no exact date in 2026-07:** [BIO X Gen 3 Bioprinter Installed and Trained](https://www.linkedin.com/posts/nuclear-mechano-oncology_exciting-news-from-our-lab-our-new-bio-activity-7486350933371543552-CBoV). Score 0.601. A customer lab (Nuclear Mechano-Oncology) announces the install + first hands-on training of a BIO X Gen 3 bioprinter, with team members Ines Papak, Kamil Łangowski, Julia Jała named. Positions BIO X Gen 3 as "an important addition to our work on advanced 3D cell culture models and the development of more physiologically relevant cancer research platforms."
+- **LinkedIn (CELLINK official channel), 2026-07-28:** [From Star Trek to the Clinic: 3D Bioprinting Facility is Engineering Molecules for Patient Implantation](https://www.linkedin.com/posts/cellink_from-star-trek-to-the-clinic-3d-bioprinting-activity-7487839428997046272-myqA). Score 0.287. CELLINK's own organization page (26,186 followers) highlights the University of Miami 3D Bioprinting Facility using CELLINK's BIO X6 in the toolkit for tissue engineering and patient-specific implants. Companion article from InventUM.
+- **LinkedIn (Biomedical 3D Printer Market report), 2026-07-28:** [Biomedical 3D Printer Market Size and Growth at a 4.7% CAGR from 2026 to 2033](https://www.linkedin.com/pulse/biomedical-3d-printer-market-size-growth-47-bstsf). Score 0.659. Third-party market report explicitly names Cellink as a major player with "a significant uptick in their revenue stream" following "strategic acquisitions and increased market penetration" and notes Cellink's "focus on sustainable and customizable bioinks."
+- **apellaser.ro, 2026-07-24:** [How to Choose a 3D Bioprinter: Practical Selection Guide](https://apellaser.ro/en/bioprinter-selection-guide). Score 0.303. Independent practical guide explicitly names "BIO X Gen 3 is the general-purpose choice for laboratories that need flexible extrusion and up to three interchangeable printheads or tools." Vendor-anchored reference, not a press release.
+- **Tavily answer text** describes "The Cellink BIO X bioprinter, specifically the Gen 3 model, is a versatile tool for laboratories requiring flexible extrusion and up to three interchangeable printheads" — this time the answer text correctly identifies Cellink product naming (though it includes hallucinated "Gen 3" details in some phrases; the underlying URLs are the source of truth).
+
+**Robin relevance:** HIGH. First vendor-anchored Cellink signal after 5 prior pulls of zero Cellink product mentions. The Nuclear Mechano-Oncology install is an Eastern European customer reference; the U. Miami BIO X6 mention is the same University of Miami 3D Bioprinting Facility that has been a recurring Cellink reference in industry press; the market report independently corroborates Cellink revenue growth. Promoted to `/commitments.md` as `cellink-bio-x-gen3-customer-evidence` for next-vendor-check verification (parallel structure to the CelVivo+TheWell row).
+
+### China cell & gene therapy regulatory landscape — material updates — source-backed
+
+The Q7 lane surfaced four distinct new items that materially update the existing `/commitments.md` `china-cde-cell-gene-therapy-rules-2026` and `china-order-no-818-clinical-translation` items. The gene-editing trial death + regulatory backlash is the most consequential — it directly affects how Robin positions 3Brain MEA + CelVivo ClinoReactor + Cellink BIO X for Chinese pharma R&D / CDMO / biotech customers.
+
+- **Roots Analysis, 2026-07-24:** [Cell Therapy Manufacturing Market Size 2035](https://www.rootsanalysis.com/reports/cell-therapy-manufacturing/285.html). Score 0.669. Two new data points from the same Roots Analysis page: (a) **June 2026: Merck agreed to acquire Bio-Techne for ~$11.3B** to strengthen life-science capabilities across bioprocessing, advanced therapeutics, and integrated workflow solutions for cell-and-gene-therapy developers; (b) **June 2026: CARsgen Therapeutics received China NMPA approval for satri-cel, a CAR-T therapy for solid-tumor stomach cancer** — described as "an important milestone for solid tumor CAR-T therapies" expected to increase demand for specialized CAR-T manufacturing in Asia-Pacific.
+- **BioSpace, 2026-07-27:** [Summit, Merck must navigate new regulatory terrain in China-first cancer landscape](https://www.biospace.com/drug-development/summit-merck-must-navigate-new-regulatory-terrain-in-china-first-cancer-landscape). Score 0.535. Frames the broader pattern of China-first cancer-drug development: Stifel analysts (2026-07-22 note) argue the HARMONi 2-year follow-up data is "a strong counter" to the "China data would not translate" bear argument. Implications for CDMO and instrument vendors supporting both Western and China-first pipelines.
+- **Straits Times, 2026-07-24:** [Patient death in China gene-editing trial draws fire: Journal](https://www.straitstimes.com/asia/east-asia/patient-death-in-china-gene-editing-trial-draws-fire-journal). Score 0.313. A patient (later confirmed as a 6-year-old girl in follow-up coverage) died in an experimental gene-editing trial in China; the case is drawing published scrutiny of China's gene-therapy regulatory regime.
+- **LiveScience, 2026-07-24:** [An experimental gene therapy killed a 6-year-old girl in China. Her parents are seeking accountability](https://www.livescience.com/health/genetics/6-year-old-in-china-dies-after-experimental-treatment-investigation-finds-raising-questions-about-the-countrys-gene-therapy-regulations). Score 0.145. Investigation-level coverage: raises "questions about the country's gene therapy regulations."
+- **SCMP, 2026-07-29:** [Is China's biotech boom at risk? Fear of policy backlash after death in gene-editing trial](https://amp.scmp.com/news/china/science/article/3362253/chinas-biotech-boom-risk-fear-policy-backlash-after-death-gene-editing-trial). Score 0.203. Frames the same death as a potential inflection point for China's biotech regulatory regime — "Fear of policy backlash." Published same day as this Tavily pull, which is what surfaces it.
+
+**Robin relevance:** HIGH (the gene-editing trial death item especially). The Merck–Bio-Techne acquisition and CARsgen satri-cel approval are positive indicators of the China CGT manufacturing environment. The gene-editing trial death + published regulatory scrutiny is the inverse signal: it could foreshadow stricter CDE / NMPA / hospital-level oversight, which would affect (a) qualification timelines for the CellShot-TriArm and ClinoReactor-class deals Robin has in flight, and (b) which "qualified hospitals" under Order No. 818 remain eligible. Promoted to `/commitments.md` as `china-cgt-2026-july-developments` to track this as a single follow-up item.
+
+### Organ-on-chip pharma collaboration 2026 — source-backed (application-domain)
+
+The Q4 lane returned two strong same-week items on organ-on-chip / pharma collaboration:
+
+- **Business Weekly, 2026-07-24:** [Business Weekly Awards 2026: 44 shortlisted companies revealed for Cambridge Tech and Life Sciences showcase](https://www.businessweekly.co.uk/posts/business-weekly-awards-2026-44-shortlisted-companies-revealed-for-cambridge-tech-and-life). Score 0.754. **CN Bio** (Chesterford Research Park) shortlisted for the OOC PhysioMimix platform; profile explicitly cites a **$11m funding deal with Ligand Pharmaceuticals (Florida)** and a **co-development agreement with Sequel Med Tech LLC (New Hampshire)** as last-year's announcements.
+- **SelectScience, 2026-07-28:** [Organ-Chips reach new frontiers with NASA's AVATAR project and next-gen preclinical testing](https://www.selectscience.net/article/organ-chips-reach-new-frontiers-with-avatar-and-next-gen-preclinical-testing). Score 0.736. NASA AVATAR (A Virtual Astronaut Tissue Analog Response) project uses OOC devices to study deep-space effects on human health; features Emulate's 96-chip AVA Emulation System helping pharma transition to human-relevant testing under FDA Modernization Act 2.0 / UK guidelines.
+- **Nature, 2026-07-24:** [Engineering human Gut-on-a-Chip culturomics](https://www.nature.com/articles/s44385-026-00099-1) reappears at score 0.367 (re-mention from pulls 1 + 5).
+
+**Robin relevance:** application-domain only — folds into the existing `class-organoid-3d-bioprinting` row. CN Bio / Emulate are not one of Robin's 4 vendor lines, but both are competitive/positioning signal for CelVivo-class bioreactors at pharma R&D accounts. NASA AVATAR is a high-visibility OOC application milestone.
+
+### Hermes Agent v0.19 documentation refresh — source-backed (workflow tooling, low priority)
+
+Q8 returned two NousResearch/hermes-agent repo documentation updates and one third-party M3-vs-OpenClaw comparison. The docs confirm v0.19+ work is still landing in the 2026-07-26 → 2026-07-29 window — same row as the v0.19 Quicksilver signal, more sources.
+
+- **GitHub (NousResearch/hermes-agent `configuration.md`), 2026-07-26:** [Iteration Budget](https://github.com/NousResearch/hermes-agent/blob/main/website/docs/user-guide/configuration.md). Score 0.590. Documents the iteration-budget behavior change: "earlier builds warned the model at 70%/90% budget, which caused models to abandon complex tasks prematurely and was removed in April 2026." Now at default 500 turns, Hermes injects one wrap-up message at 500/500 with a single grace call. Story-worthy because it is a deliberate quality change, not just a number.
+- **GitHub (NousResearch/hermes-agent `features/api-server.md`), 2026-07-27:** [POST /v1/runs/{run_id}/stop + /approval endpoints](https://github.com/NousResearch/hermes-agent/blob/main/website/docs/user-guide/features/api-server.md). Score 0.367. Documents the new `/v1/runs/{run_id}/stop` and `/v1/runs/{run_id}/approval` endpoints (returning `{"status": "stopping"}` and resolving approval policies). Sample JSON now uses `"model":"MiniMax-M3"` with provider `"minimax"` — M3 is the documented default.
+- **MyClaw.ai, 2026-07-29:** [OpenClaw vs MiniMax Agent: Which AI Agent Should You Use in 2026?](https://myclaw.ai/blog/openclaw-vs-minimax-agent). Score 0.543. Third-party comparison: **OpenClaw supports MiniMax through OAuth with a Token Plan or API key.** M3 handles primary reasoning; M2.7 variants remain available. Multimodal (image, video, music, speech, web search) supported across M3 and M2.7.
+- **Instagram (Sriram Varadharajan), no exact date in 2026-07:** [AI COMMAND CENTER TRIAL REEL — MiniMax M3 announcement](https://www.instagram.com/reel/DbVydtNj46Z). Score 0.497. Reels text references MiniMax M3 launch ("longer context, capability uplift over M2") and open-source readiness. Adjacent to MiniMax-M3 theme row.
+
+**Robin relevance:** WORKFLOW TOOLING (low priority). The configuration.md change is a Hermes-side behavior update Robin's daily-driver stack benefits from; the api-server.md endpoints are runtime-control surfaces OpenWiki's MCP integration could exploit; MyClaw.ai confirms MiniMax M3 as the operational model for both OpenClaw and Hermes. Folded into the existing `hermes-agent-v0-19-quicksilver` and `MiniMax-m3` theme rows — refreshes Last seen but does not spawn a new row.
+
+### Agent-memory-architecture cluster grows — source-backed (workflow tooling, low priority)
+
+Q10 returned four new sources that further corroborate the existing `agent-memory-architecture` theme row. Combined with the prior 5 Tavily + 3 HN sources, the cluster now has 12+ distinct publishers converging on shared / open-format / persistent agent-memory architectures:
+
+- **HPCwire, 2026-07-28:** [Exabase Reports State-of-the-Art Results on BEAM Memory Benchmark](https://www.hpcwire.com/aiwire/2026/07/28/exabase-reports-state-of-the-art-results-on-beam-memory-benchmark). Score 0.624. Exabase M-1 achieves highest reported score at every scale: **76.9% at 100K tokens, 75.0% at 1M tokens, 68.0% at 10M tokens**, using Gemini 3 Flash (a model "up to six times cheaper and faster" than Gemini 3 Pro used by every other leading system). 20% fewer tokens per query than next-best system.
+- **Hackernoon (Moorcheh.ai CTO), 2026-07-28:** [Agent Memory Has a Lock-In Problem. Open Formats Are How We Fix It.](https://hackernoon.com/agent-memory-has-a-lock-in-problem-open-formats-are-how-we-fix-it). Score 0.605. Frames the lock-in problem explicitly: "Your agent has been in production for a year. … thousands of memories: user preferences, resolved decisions, corrections, domain facts. Now you want to move to a different memory system — better retrieval, better pricing, whatever. What's your migration path?" Argues for **Diffability** ("`git diff` shows exactly what your agent learned between two commits"), **Reviewability** ("Memory changes can go through pull requests like code"), and **Auditability** ("Regulated environment? Your …"). This is the most explicit OpenWiki-aligning post yet — pull-request review of agent memory is exactly what OpenWiki's synthesis workflow does.
+- **Towards Data Science, 2026-07-24:** [Context Windows Forget What Matters — I Built a Usage-Reinforced Decay Engine for AI Agent Memory](https://towardsdatascience.com/context-windows-forget-what-matters-i-used-a-140-year-old-psychology-paper-to-fix-ai-memory). Score 0.576. Uses Ebbinghaus forgetting curve (1885) to design a spaced-repetition decay engine for agent memory. Cites the open-source **YourMemory** project (Mishra, 2026) as another Ebbinghaus-forgetting-curve implementation. Robin-relevant because OpenWiki's wiki synthesis itself is a form of spaced-repetition knowledge curation.
+- **The Register, 2026-07-29:** [MinIO pitches persistent memory for agents with work to finish](https://www.theregister.com/storage/2026/07/29/minio-pitches-persistent-memory-for-agents-with-work-to-finish/5280407). Score 0.425. MinIO **AIStor Memory** for long-running multi-step agent workflows. Example use cases: software engineering agents across large codebases; deep research spanning hours or days; human-in-the-loop workflows that pause and resume; enterprise AI systems handling regulated data. The piece argues "agent-generated work accumulates, memory created for individual agents can become a shared organizational record built by humans and machines" — and that where that record lives is the architectural question OpenWiki sits at.
+
+**Robin relevance:** WORKFLOW TOOLING (low priority). The cluster is now 12+ sources. The Hackernoon/Moorcheh.ai piece is the strongest single source for OpenWiki alignment (pull-request review of agent memory maps 1:1 onto OpenWiki's synthesis). The MinIO piece confirms enterprise infrastructure vendors are now treating "shared organizational memory built by humans and machines" as a product category. Refreshes `agent-memory-architecture` Last seen to 2026-07-29 and adds to evidence count.
+
+### 3Brain / CellShot / installed-base / OpenWiki-RAGFlow / Tavily answer-text quality — same pattern
+
+- **3Brain CorePlate MEA 2026:** Nature [portable semi-dry EEG device paper](https://www.nature.com/articles/s41598-026-52672-8) reappears at score 0.187 — **fifth pull in a row with the same mis-attribution**. Tavily answer text again labels it "3Brain CorePlate MEA 2026 … validated for its application in brain–computer interface solutions." Underlying paper is a portable EEG headset, not 3Brain MEA chips. Do not promote as 3Brain evidence.
+- **Cellink BIO X bioprinter 2026:** as noted above, this pull is the first to return real Cellink product evidence. Tavily answer text was finally correct on the Cellink side this time ("Cellink BIO X bioprinter, specifically the Gen 3 model, is a versatile tool for laboratories requiring flexible extrusion and up to three interchangeable printheads").
+- **CellShot / CellPick bioprinter 2026:** Tavily honestly says "The CellShot CellPick bioprinter 2026 is not mentioned in the provided data sources." Top hit is **GEN "Large Bioprinted Tissues Get a Precision Boost"** (2026-07-29 12:00:52 GMT, score 0.259) — application-domain news, no Femtobiomed vendor signal. The same GEN article reappears in Q1 (0.037), Q2 (0.032), and Q5 (0.259). Worth a note in `class-organoid-3d-bioprinting` evidence column.
+- **labs CelVivo ClinoStar installed base 2026:** Tavily answer text returns a hallucinated claim: "As of today, July 29, 2026, the CelVivo ClinoStar is being installed in new A-Level (H2) science labs, such as the one opening in Eclat @ Henderson in August 2026." **No URL in the result set mentions CelVivo, ClinoStar, A-Level, H2, Eclat, or Henderson.** This is a fabricated Tavily answer — flag and ignore. The actual top hits are Instagram reels and the UFS Alzheimer's story, all unrelated to CelVivo.
+- **Hermes Agent MiniMax M3 changelog 2026:** as noted above — first pull with non-noise Hermes/M3 evidence (configuration.md, api-server.md, MyClaw.ai, Instagram reel). Promotes the existing rows instead of spawning new ones.
+- **OpenWiki connectors Tavily RAGFlow 2026:** recall still broken. Returns Roche ADLM 2026 (0.043), Nature chemical-reaction pathway algorithm (0.029), SmartBall X water infrastructure (0.025), exposure-management AI-attack cybersecurity (0.023), HID workforce unit transfer (0.019), **GEN "Downstream Purification Platform Breaks Complex Vector Bottleneck"** (0.017), RightShip maritime (0.016), insurance payment (0.016). Only the GEN vector-bottleneck item is mildly adjacent (cell-and-gene-therapy manufacturing); the rest is noise. Recommend dropping this lane from future Tavily runs.
+
+### Synthesis result
+
+- **Three new scope-aligned findings:**
+  1. **Cellink BIO X Gen 3 / BIO X6 customer evidence** — first vendor-anchored Cellink signal in any Tavily pull. Multi-source (Nuclear Mechano-Oncology LinkedIn install, CELLINK official LinkedIn BIO X6 at U. Miami, third-party market report, appellaser selection guide). Promoted to `/commitments.md` as `cellink-bio-x-gen3-customer-evidence`.
+  2. **China CGT regulatory landscape updates** — Merck–Bio-Techne $11.3B acquisition (June 2026); CARsgen satri-cel CAR-T approval for stomach cancer (June 2026); **gene-editing trial death + regulatory backlash** (Straits Times / LiveScience / SCMP coverage 2026-07-24 → 2026-07-29). Promoted to `/commitments.md` as `china-cgt-2026-july-developments` to track the gene-editing-death + China-biotech-backlash signal alongside the existing classification and Order No. 818 items.
+  3. **Hermes Agent v0.19 docs continue** — `configuration.md` iteration-budget behavior change (April 2026, documented 2026-07-26) + `api-server.md` new stop / approval endpoints (2026-07-27). Refreshes Last seen for `hermes-agent-v0-19-quicksilver`. MyClaw.ai confirms MiniMax M3 + M2.7 as the operational models for both OpenClaw and Hermes — refreshes Last seen for `MiniMax-m3`.
+- **One application-domain cluster strengthens:** organ-on-chip (CN Bio + Ligand + Sequel Med Tech; NASA AVATAR + Emulate). Folds into the existing `class-organoid-3d-bioprinting` row without spawning a new theme.
+- **Agent-memory-architecture cluster grows to 12+ distinct sources.** Hackernoon/Moorcheh.ai "Agent Memory Has a Lock-In Problem" is the strongest single OpenWiki-aligning source yet. Refreshes Last seen for `agent-memory-architecture`.
+- **Tavily answer-text mis-attribution continues** on Q2 3Brain (5th pull). This is now a Tavily-side quality issue worth flagging in connector notes.
+- **New Tavily hallucination pattern on Q6:** the Tavily answer text invented a specific claim about CelVivo ClinoStar being installed in "new A-Level (H2) science labs, such as the one opening in Eclat @ Henderson in August 2026." No underlying URL supports this. Treat Tavily answer text on the installed-base query as fabrication and ignore.
+- **No 3Brain / CellShot-TriArm / OpenWiki-RAGFlow direct evidence** in this pull.
+
+## Source coverage (cumulative)
+
+- Pull 1: fetched 2026-07-26T17:39:11Z — raw `~/.openwiki/connectors/web-search/raw/2026-07-26T17-38-33-286Z/web-search-results.json`
+- Pull 2: fetched 2026-07-26T18:07:27Z — raw `~/.openwiki/connectors/web-search/raw/2026-07-26T18-07-14-335Z/web-search-results.json`
+- Pull 3: fetched 2026-07-26T18:16:21Z — raw `~/.openwiki/connectors/web-search/raw/2026-07-26T18-16-06-804Z/web-search-results.json`
+- Pull 4: fetched 2026-07-27T18:04:46Z — raw `~/.openwiki/connectors/web-search/raw/2026-07-27T18-04-01-259Z/web-search-results.json`
+- Pull 5: fetched 2026-07-28T18:04:26Z — raw `~/.openwiki/connectors/web-search/raw/2026-07-28T18-03-43-560Z/web-search-results.json`
+- Pull 6: fetched 2026-07-29T18:03:26Z — raw `~/.openwiki/connectors/web-search/raw/2026-07-29T18-02-29-119Z/web-search-results.json`
+- Queries: pulls 1, 2, 4, 5, 6 = 10 queries each; pull 3 = 6 queries (vendor-anchored only)
+- Instance: `web-search-1`