chore: daily update 2026-07-30
This commit is contained in:
OpenWiki Bot
@@ -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`
|
||||
|
||||
Reference in New Issue
Block a user