29 KiB
Files
-
4-27 SOP 实战 vs 实现差异 (2026-07-27 Tasevska 2025 重做教训) - Lessons-learned capture from Robin's 2026-07-27 Tasevska 2025 EJPS deep-read redo — 9-step comparison table of 4-27 SOP范本 vs 7-27 凭印象实操 (Tasevska + Raji 2 篇) vs 4-27 SOP 走通 (Tasevska 重做); detailed PyMuPDF 嵌入图过滤 rules, Ollama 视觉 fallback chain, NocoDB v1 vs v2 protocol, and the 9-step mandatory format. Applies to every literature-deep-read invocation. Companion to
/sources/ldr-report-template.md. -
Agents 代理商档案索引 — 跨 agent CRM 同步 - Cross-agent index of the Obsidian vault 02-Clients/Agents/* profiles (which agents Robin is actively working with, mapped by vendor product line). Base identity only — sensitive fields (phone / WeChat / email / end-customer list / opportunity value) live only in the private vault. Updated 2026-08-07 with _known_associations.md product-line → agent → end-customer mapping.
-
全国经销商地图 (2026-08-07 时点) - Compact index of Robin's 2026-08-07 China distributor baseline (259 dealers across 7 regions, replacing 261 in the 2026-07-24 baseline) plus the new product-association index _known_associations.md. Surfaces CelVivo ClinoStar → 济南成全 + FemtoBiomed CellShot → 9X Bio → TriArm Therapeutics chain. Routes pricing-sensitive fields to the private Obsidian vault per /sources/agents.md privacy rule.
-
飞书邮箱 (Robin customer mailbox) - Planned cross-agent mailbox sync target for Robin's Feishu (飞书) mailbox (imap.feishu.cn:993 / imap.larksuite.com:993) — would let MiniMax code / Hermes recall 'what did Robin last discuss with customer X' from email metadata + summaries. Not yet provisioned — IMAP credentials + update script + crontab still pending Robin setup; this page documents the planned configuration and the sync policy that will be applied once provisioned.
-
Git repo — evidence index (Obsidian vault + OpenWiki wiki, Robin) - Compact manifest of Robin's Obsidian vault + the local OpenWiki wiki, ingested through the local git-repo connector. Latest fetch (2026-09-30T18:24Z — 46th git-repo-1 fetch since 2026-08-07): vault HEAD unchanged at
09bec367for the 19th consecutive fetch after the Tomas 2026 + Vijakumaran 2026 resumption; wiki HEAD bumped2fccbf0→1d3c470via 2 new automated wiki commits (5c788ccdaily update 2026-09-30 +1d3c470sources-sync 2026-09-30 — the forty-seventh consecutive sources-sync); 1 working-tree modification atMOCs/MOC-类器官与3D培养.md(mtime 2026-09-13 11:55:42 +0800 — persists uncommitted for 17 calendar days at this fetch); 11-dirty-file baseline (vs the 13-dirty baseline at the 2026-09-29T18:14Z fetch — the5c788ccdaily update cycle clearedsources/git-repo.md+sources/README.md+sources/literature.md+D tmp/check_themes.py(deletion) from the dirty list; the1d3c470sources-sync re-touched the 4 source-sync files (sources/nocodb.md+sources/ragflow.md+sources/seafile.md+sources/wikijs.md); the post-sources-sync synthesis-side write path re-touchedindex.md+quickstart.md+sources/hackernews.md+sources/index.md+sources/web-search.md+themes.md; net: −2 vs the 13-dirty baseline); carry-forward bump 48/43/43/36 → 49/44/44/37-day (highest carry-forward durations of the 2026-08-07 streak). Two configured repos: obsidian-vault at /home/ldw/workspace (main branch, HEAD09bec367, 2026-09-11, vault HEAD unchanged at09bec367for the 19th consecutive fetch after the Tomas 2026 + Vijakumaran 2026 resumption; the MOC working-tree modification persists uncommitted for 17 calendar days at this fetch); and openwiki-wiki at /home/ldw/.openwiki/wiki (main branch, HEAD1d3c470, 2026-09-30, bumped from2fccbf0via 2 automated commits in the 24-hour window —5c788ccdaily update +1d3c470sources-sync forty-seventh). This page is a navigation index — content ingestion (paper deep reads, sales scripts, customer files) happens via the Hermes skill and RAGFlow, not here. -
Hacker News — evidence index - Compact evidence trail for the Hacker News connector (hackernews-1, Robin-curated life-sciences + AI workflow). Synthesis lives on /themes.md, /commitments.md, etc.; this page only records ingestion coverage and notable items. 50th pull (2026-09-30) anchors on
49901736Livenerf: Has Opus 5.5 been nerfed yet? (828 pts top+best + 346 cmts, github.com/ninjahawk/livenerf — HIGHEST ENGAGEMENT ITEM OF THE 2026-09-30 PULL) +49906637You Said No MCP (476 pts top+best + 276 cmts, earendil.com — 2nd-highest) +49910553The AI Race Just Got Awkward (341 pts best + 333 cmts, insufferable.dev — would-be 42nd AI-vendor-business-stack cluster entry candidate) +49902382Show HN: Ledge.sh — Runnable Markdown Notes (30 pts top + 20 cmts, ledge.sh — named-stack-adjacent markdown-substrate cross-walk) +49911995Launch HN: Magnitude (YC S25) — Self-optimizing inference engine for agents (19 pts top + 6 cmts, github.com/magnitudedev/magnitude — would-be 30th open-weight-LLM cluster entry candidate + 15th self-hosted-agentic-substrate cluster entry candidate); 49th pull (2026-09-29) anchors on the 3-Jev-entry same-day triple spanning 558 + 185 + 42 = 785 pts in a single pull:49883844"Jeff — Jev-compatible 0.8B decision models, trained at home, ~30 ms"** (558 pts best + 215 cmts — byfirelex, https://github.com/firelex/jeff — **HIGHEST-ENGAGEMENT Jev-family ITEM SINCE the 2026-09-1649717558System One Models + Jev 1728 pts original launch**; **would-be 9th Jev-family adjacent-not-promoted entry candidate** + **28th open-weight-LLM cluster entry candidate**, adjacent-not-promoted per scope because the entry is a personal-GitHub project rather than a vendor canonical release + the cluster's promotion discipline has been "vendor canonical release + strong engagement" + firelex Jeff is not on Robin's named daily-driver stack — Hermes / MiniMax-M3+H3 / OpenWiki / RAGFlow / NocoDB / SeaFile / Codex CLI are the named agents) + **49891290"Jeeves. Reasoning improves Jev-like decision models"(185 pts top + 77 cmts — bynicowaltz, https://github.com/PostHog/jeeves — github.com/PostHog/jeeves PostHog-maintained Jeeves library; the 0.42 cmts-to-upvote ratio; the PostHog-maintained GitHub library framing with "reasoning improves Jev-like decision models"; would-be 10th Jev-family entry candidate, adjacent-not-promoted per scope because the entry is a PostHog-maintained GitHub library rather than a vendor canonical release + PostHog is not on Robin's named daily-driver stack) +49891769"Show HN: Jevstiller — Distill Jev into a local model, with a disagreement bound"** (42 pts top + 6 cmts — bytgluck, https://jevstiller.pages.dev/posts/the-guarantee/ — Show HN distillation tool with disagreement bound from a 0.8B Jev-family local model; would-be 11th Jev-family entry candidate, adjacent-not-promoted per scope because 42-pt engagement is well below the 100-pt cluster entry bar + the entry is a Show HN launch rather than a vendor canonical release) — all three adjacent-not-promoted per scope because personal-GitHub / PostHog-GitHub / Show-HN framings rather than vendor canonical releases; cross-walks to the 2026-09-1649717558System One Models + Jev 1728 pts + 2026-09-1849752041OpenJev 405 pts + 2026-09-2149807815Kev Jev-like 317 pts + 2026-09-2649848269Ollaya 553 pts + 2026-09-2649853175Jev-wrapper 120 pts + 2026-09-2849846632Decide-Jev-decisions-by-vsekhar 0 cmts Algolia substring-noise entries) + **OpenAI vendor canonical release** **49896586"GPT 6.1 Sol"(289 pts top+best + 206 cmts — bycrorella, https://openai.com/index/introducing-gpt-6-1-sol/ — OpenAI's GPT 6.1 Sol canonical release on openai.com; the 0.71 cmts-to-upvote ratio; the 289-pt engagement is the 10th-highest single-item of the 2026-09-29 pull; the GPT 6.1 Sol release continues the 2026-09-04 GPT-6 Astra + 2026-09-23 GPT-6 Sol-and-Luna family; would-be 11th closed-weight-M3-comparison-anchor cluster entry candidate + 35th AI-vendor-business-stack cluster entry candidate, adjacent-not-promoted per scope because (a) OpenAI is closed-weight and not on Robin's named daily-driver stack — Hermes / MiniMax-M3+H3 / OpenWiki / RAGFlow / NocoDB / SeaFile / Codex CLI are the named agents, (b) the cluster's promotion discipline has been "vendor canonical release + strong engagement" and the 2026-09-29 entry sits at the lower-engagement tail of the cluster pattern, (c) the existing closed-weight-M3 cluster already covers the GPT-6 family via the 2026-09-04 GPT-6 Astra 2080 pts + 2026-09-23 GPT-6 Sol and Luna 1683 pts + 2026-09-05 GPT-6 Astra on OpenRouter 279 pts entries) + OpenAI companion product launches —49896604"Dots"** (192 pts top + 102 cmts, https://openai.com/index/introducing-dots/ — would-be 12th closed-weight-M3 cluster entry candidate + 36th AI-vendor-business-stack cluster entry candidate) + **49896975"ChatGPT Pro 500"(68 pts top + 56 cmts, https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers — would-be 13th closed-weight-M3 cluster entry candidate + 37th AI-vendor-business-stack cluster entry candidate) +49896600"DevDay 2026 Recap"** (27 pts + 2 cmts, https://openai.com/index/devday-2026-recap/ — off-scope per scope, sub-100-pt cluster entry bar) + **Cal Newport AI-lab investigation essay** + **49883471"It's Time to Investigate the AI Labs"(579 pts best + 246 cmts — byibobev, https://calnewport.com/its-time-to-investigate-the-ai-labs/ — HIGHEST ENGAGEMENT ITEM OF THE 2026-09-29 PULL; would-be 38th AI-vendor-business-stack cluster entry candidate + 6th adjacent-not-promoted AI-labs + regulatory sub-cluster entry, adjacent-not-promoted per scope because the entry is a calnewport.com personal-blog essay rather than a vendor release) + World Labs + AMD merger +49883760"World Labs is Joining AMD"** (298 pts best + 111 cmts, https://www.worldlabs.ai/blog/amd-announcement — would-be 39th AI-vendor-business-stack cluster entry candidate via AMD-vendor-merger lane, adjacent-not-promoted per scope because AMD is not on Robin's named daily-driver stack + the cluster's promotion discipline requires a vendor privacy / IP / business-stack signal rather than a vendor-merger announcement) + **Privacy-of-conversational-AI-agents PDF** + **49890226"A Privacy Analysis of Web and Mobile Conversational AI Agents [pdf]"(378 pts top+best + 121 cmts, https://jorgegarciaherrero.com/wp-content/interactivos/20260916-Prompt-like-a-butterfly-sting-like-a-tracker-(clean).pdf — would-be 40th AI-vendor-business-stack cluster entry candidate via the privacy-of-conversational-AI-agents framing, adjacent-not-promoted per scope because the entry is a 3rd-party PDF rather than a vendor release / framework / schema / named-stack signal) + highest-engagement top+best of the pull:49883471"It's Time to Investigate the AI Labs" 579 pts +49883844Jeff-Jev-compatible-0.8B 558 pts +49891480"500k facial scans at UK stations yield no arrests, 1 false positive" 395 pts +49890226"Privacy Analysis of Conversational AI Agents" 378 pts +49883539California-grapes 348 pts +49891550US-sanctions-Netherlands-NixOS 338 pts +49894005macOS-Golden-Gate-Buggy-Mess 338 pts +49892245How-Delhi-cut-electricity-loss 319 pts +49883760World-Labs-Joining-AMD 298 pts +49896586GPT-6.1-Sol 289 pts +49882781MicroLLM-Lab 272 pts. Algolia substring-noise lane:organoidreturned nbHits=7 / hits=7 (7 substring-noise false positives onOrganizes/Organization/Organizer/organism/orgs/organized— the49883159PLC-Organization 55 pts + 25 cmts is the highest-engagementorganoidAlgolia substring-noise false positive of the 2026-08-07 streak but adjacent-not-promoted per scope),CelVivoclean zero (17th consecutive clean-zero CelVivo pull — extends the longest zero streak of the 2026-08-07 streak),3Brainreturned nbHits=5 / hits=5 (5 Brain-substring-noise false positives including the highest-engagement49893157Ask-HN-Whatre-You-Reading 15 pts + 43 cmts + the named-stack-adjacent49887913Jev-vs-Fruit-Fly-Brain-in-Roblox 1 pt + 1 cmt +49894697Slow-running-benefits-brain 2 pts +49893438longer-exhale-brain-bold-decisions 1 pt +4988539110-Tokens-Meat-Brain-LLM 1 pt + 1 cmt — all off-scope per scope),Cellinkreturned nbHits=5 / hits=5 (5 Cellink-substring-noise false positives including the named-stack-adjacent49890748Muse-gives-out-home-address 12 pts + 2 cmts +49896730WebMCP-calling 1 pt +49895381Selling-Private-study-chat 1 pt +49892198Search2o-AI-agent 1 pt +49891337Tech-Trillion-Dollar 7 pts + 3 cmts — all off-scope per scope),openwikiclean zero (13th consecutive fully-zero openwiki pull),deepagentsclean zero (15th fully-zero deepagents lane in a row; 30 of 31 pulls since 2026-08-27 are fully-zero on deepagents). ~26 unique top+best items (top 21 + best 12 - 7 overlap = ~26 unique items). No cluster entry promotions this run; would-be 9th/10th/11th Jev-family + 11th/12th/13th closed-weight-M3 + 28th open-weight-LLM + 14th self-hosted-agentic-substrate + 35th/36th/37th/38th/39th/40th AI-vendor-business-stack cluster entries all adjacent-not-promoted per scope. Tavily (web-search) remains the productive lane for vendor news; HN lane's stable role continues as "high-engagement inference-lane / agent-stack adjacents + occasional substring-noise false positives on the vendor queries" rather than vendor-direct signals. NoCelVivo/3Brain/Cellink/openwiki/deepagentsvendor release hits in the top+best feeds. Prior 48th pull (2026-09-28) surfaces49874728"Prompting Claude Opus 5.5" (186 pts best + 211 cmts — byMichelangelo11— Anthropic's Claude Opus 5.5 prompt-engineering docs on platform.claude.com — would-be 10th closed-weight-M3-comparison-anchor cluster entry candidate + 30th AI-vendor-business-stack cluster entry candidate, adjacent-not-promoted per scope because Anthropic is closed-weight and not on Robin's named daily-driver stack + the entry is a prompt-engineering docs page rather than a model release) +49881484"OpenAI still doesn't seem to have a handle on all of its rogue AI activity" (16 pts + 11 cmts — techcrunch.com — 5th item in the OpenAI-agents-as-attacker sub-cluster adjacency in 16 days, would-be 31st AI-vendor-business-stack cluster entry candidate, adjacent-not-promoted per scope because engagement is well below the 100-pt cluster entry bar + the entry is a techcrunch news item rather than a vendor release) + algolia substring-noise3Brain49880312"The problem is not AI code, but not knowing about system architecture or intent" (256 pts top+best + 170 cmts — ssp.sh/brain/ — cross-walked to the AI-vendor-business-stack cluster's AI-coding-agent lane + thehermes-agent-v0-19-quicksilverrow via the architecture-vs-AI-code framing) + algolia substring-noiseCellinkShow HN: Decide — Jev decisions in the shell, scripts, and agent skills byvsekharon GitHub — 3rd Jev-family entry within the 2026-09-26 → 2026-09-28 window after Ollaya 553 + Jev-wrapper 120 (would-be 9th Jev-family adjacent-not-promoted entry candidate, 0-cmt Algolia engagement so well below the 21-pt Almanac refresh trigger). Prior 47th pull (2026-09-27) surfaces the 6th-highest single-item of the entire 2026-08-07 streak49863864"Unsealed Briefs in Authors' Case v. Microsoft/OpenAI" (567 pts + 527 cmts — bypapergirl, https://authorsguild.org/news/ag-v-openai-top-execs-knew-mass-book-piracy-was-illegal/ — HIGHEST ENGAGEMENT ITEM OF THE 2026-09-27 PULL — authorsguild.org court-filings writeup on the unsealed Microsoft/OpenAI book-piracy court filings with the "Top Execs Knew Mass Book Piracy Was Illegal" framing; IP/litigation cross-walk to AI-vendor-business-stack cluster's Microsoft-exec-AI-scraping-theft-of-labor 753 pts; would-be 28th AI-vendor-business-stack cluster entry candidate) +49864642Meta-Blocks-President-Lula-Facebook-Page 381 pts (off-scope per scope) +49859112DeepSeek-Elastic-Compute (DSec) arXiv preprint 308 pts (named-stack-adjacent to local-LLM deployment lane via DSec elastic-compute architecture; would-be 27th open-weight-LLM cluster entry candidate) +49867067Installing-NeoVim-caused-original-Vim-undo-files-to-be-deleted 280 pts (off-scope per scope) +49867038Tells-of-a-Slop-UI 256 pts (named-stack-adjacent to AGENTS.md + Codex CLI / Hermes Agent skill-file pattern via Slop-UI critique framing) +49868083There-are-no-rogue-AI-agents 114 pts (named-stack-adjacent to OpenAI-agents-as-attacker sub-cluster via the rogue-AI-agents framing; would-be 29th AI-vendor-business-stack cluster entry candidate + 4th item in the OpenAI-agents-as-attacker sub-cluster adjacency) +49867553PostmarketOS-rebranding-as-Nura 86 pts (named-stack-adjacent self-hosted-substrate cross-walk within 48h of the 2026-09-25 Dutch-government-NixOS 842 pts) +4985998242x-faster-prompt-lookup-drafting-in-llama.cpp 14 pts (named-stack-adjacent local-LLM-deployment cross-walk via llama.cpp inference-engine pattern + unsloth/MiniMax-M3-GGUF local-LLM run path). Algolia substring-noise lane:organoidreturned nbHits=4 / hits=4 (4 substring-noise false positives: EasyActions GitHub Actions org-wide 1 pt + LightCloud file-system cloud console 8 pts + The AI-First Organization 3 pts + OrbitalNote orgmode editor 2 pts, all off-scope per scope);CelVivoreturned clean zero (15th consecutive clean-zero CelVivo pull — longest zero streak of the 2026-08-07 streak);3Brainreturned nbHits=5 / hits=5 (5 Brain-substring-noise false positives: Show-HN-JJL-esolang 1 pt + SW-eng-question 2 pts + NYT-Investors-Betting-on-BCI-Devices 1 pt + RNZ-reading-does-to-brains 4 pts + caliber.az-Stanford-brain-two-organs 3 pts, all off-scope per scope; no 3Brain HD-MEA matches);Cellinkreturned nbHits=4 / hits=4 (4 substring-noise false positives: Calling-the-AI-bluff 7 pts + Trump-fuel-economy 7 pts + Software-occlusion-culling 1 pt + Wet-Bulb-Tracker-Tracker 2 pts, all off-scope per scope);openwikireturned clean zero (11th consecutive fully-zero openwiki pull);deepagentsreturned clean zero (13th fully-zero deepagents lane in a row — 28 of 29 pulls since 2026-08-27 are fully-zero on deepagents). ~25 unique top+best items. Prior 46th pull (2026-09-26) was anchored on the 2nd-highest top+best item of the entire 2026-09 streak49849985"Revealing the details of how OpenAI agents hacked Hugging Face" (632 pts + 402 cmts — byspecked-citrus, https://swarmtraces.org/ — HIGHEST ENGAGEMENT ITEM OF THE 2026-09-26 PULL — swarmtraces.org research writeup of the OpenAI-agents-as-attacker HF security incident; would-be 25th AI-vendor-business-stack cluster entry candidate + would-be 10th closed-weight-M3-comparison-anchor cluster entry candidate, adjacent-not-promoted per scope because the entry is a security-incident writeup rather than a vendor release) +49848269"Ollaya — Ollama for open-source, Jev-style decision models" (553 pts top+best + 134 cmts — byArdakilic, https://ollaya.dev/ — Jev-family local-LLM-deployment adjacency item; would-be 25th open-weight-LLM cluster entry candidate, adjacent-not-promoted per scope pending Tavily verification of open-weight status + ollaya.dev vendor identity) +49855315"Breaking Up with Google Play" 502 pts +49852302"Jury finds Facebook liable for deceiving users in Cambridge Analytica case" 359 pts +49852717"We're gonna need a lot more mathematicians" 273 pts (Terry Tao canonical essay on mathematics-vocabulary-gap-for-AI; adjacent-not-promoted per scope because personal-blog essay rather than vendor release) +49849832Excel-multi-values-single-cell 243 pts +49853175"A single function Jev-like wrapper for LLMs" 120 pts (2nd Jev-family entry same day as Ollaya; same-day Jev-family double at 553 + 120 = 673 pts) +49854416"Floci: Locally emulating any cloud service" 115 pts (named-stack-adjacent via local-cloud-emulation pattern; would-be 8th self-hosted-agentic-substrate cluster entry candidate). Algolia substring-noise lane:organoidreturned 3 hits (all off-scope per scope);CelVivoreturned clean zero (14th consecutive clean-zero pull);3Brainreturned clean zero (breaks the 2026-09-25 7-consecutive-pull substring-noise streak — 2nd 3Brain clean-zero pull of the 2026-09 streak);Cellinkreturned 2 hits;openwikiclean zero;deepagentsclean zero. ~17 unique top+best items. Prior 45th pull (2026-09-25) anchored on49841563Dutch government builds alternative for Microsoft based on NixOS 842 pts + 493 cmts as the highest-engagement item of that pull + 8th-highest single-item of the entire 2026-08-07 streak. No cluster entry promotion this run. Top items adjacent-not-promoted per scope (OpenAI / DeepSeek closed-weight, court-filings / research-preprint / substack-essay framings) — would-be 11th closed-weight-M3 + 9th/10th/11th Jev-family + 14th self-hosted-agentic-substrate + 35th/36th/37th/38th/39th/40th AI-vendor-business-stack cluster entries + 28th open-weight-LLM cluster entry all carry forward to the Tavily verification queue. -
文献精读 Master Template (Robin 7-27 实战版, "真适合浏览" 1 文件) - Master template for Robin's literature-deep-read (LDR) pipeline — consolidates 4 literature skills (
literature-deep-read/research/review/curation-for-sales) into a single 8-rule + 3-trigger + 5-section "真适合浏览" report format. Synthesized from 4-27 SOP范本 + 7-27 实战 lessons (Tasevska / Raji / Bosnjakovic). Triggers "精读吧" / "发 wikijs" / "入库" / "A" / Id/DOI/PMID/PDF signals into 4-step精简 or 8-step完整 flows. PDF全文必传 + PyMuPDF xref去重 + 9项强制 + SeaFile hostnamefb.biokingdom.top+ NocoDB v1 JSON + 3张真图 +journal-year-topic报告路径. -
Bosnjakovic 2025 PAH 3D 球体代谢组学: 跟 CelVivo ClinoStar dynamic culture 关系 (Bosnjakovic 2025 bioRxiv) - Deep-read synthesis of Bosnjakovic A et al. 2025 bioRxiv preprint (DOI 10.1101/2025.07.10.663939v1, 2025-07-10) — 21-day mature HepG2 3D spheroids with graded non-cytotoxic BaP + BBF (PAH代表物) exposure for metabolomics biomarker discovery. Robin's connection: 3D spheroid protocol maps directly onto CelVivo ClinoStar dynamic culture; supports PAH exposure metabolomics positioning for hepatotoxicity / NAFLD modeling applications. 精读 2026-07-27.
-
CelVivo 2026 脑类器官白皮书《Breaking the Bias》精读索引 - Deep-read synthesis of CelVivo ApS (Denmark) 2026 whitepaper 'Breaking the Bias: Addressing the challenges facing optimal neural organoid development' (16 pages, translated + 精读 2026-07-28). Synthesizes CelVivo's positioning on neural organoid long-term culture and the clinical-translation path. Direct vendor-channel evidence for the CelVivo ClinoStar / ClinoReactor portfolio on brain organoid applications — companion to academic literature flywheel-sync.
-
Liu 2026 — shape-conformal porous frameworks for full coverage of neural organoid electrophysiology (Nat Biomed Eng, Rogers Lab) - Deep-read synthesis of Liu N et al. (2026) Nature Biomedical Engineering — shape-conformal porous framework + adaptive genetic algorithm + flexible-electronics microlattice for full-surface (>90%) neural organoid electrophysiology at 3000-8000 channels per organoid (vs 200-1000 traditional MEA); weeks-long stable recording; epilepsy/AD disease modeling.
-
Miao 2025 — co-development of mesoderm and endoderm enables organotypic vascularization in lung and gut organoids (Cell v2) - Deep-read synthesis of Miao Y et al. (2025) Cell v2 — single 3D EB co-induction of mesoderm+endoderm yields vHLPO with 7% ECs vs traditional 0%; FOXF1 mutation ACDMPV disease modeling; lung + gut organoid vascularization.
-
Raji 2026 (hMOs + 3Brain HD-MEA) — 7-27 增补版 (跟 7-01 矩阵集成) - Deep-read synthesis of Raji H, Bertoli F, Perez MJ, Lam A, Volpicelli-Daley LA, Deleidi M. 2026 Molecular Psychiatry (Nature 子刊, s41380-026-03667-4, DOI 10.1038/s41380-026-03667-4, IF 11) — human midbrain organoids (hMOs) with WNT modulation, dopaminergic neuron differentiation, bioreactor culture, and 3Brain HD-MEA electrophysiology for Parkinson's disease modeling. 7-01 实战范本 (page 1622) + 7-27 增补版 integrated with 3-paper 3Brain MEA matrix (Raji / Pali / Wei) and 4 product-line complementarity chain (CelVivo ClinoStar + Cellink BIO X + 3Brain HD-MEA + FemtoBiomed CellShot).
-
Sharma 2026 — vascularized retinal organoid RGC survival and maturation (Cell Stem Cell) - Deep-read synthesis of Sharma K et al. (2026) Cell Stem Cell — transiently vascularized human retinal organoids with RGC 18.2% vs ctrl-RO 1.21% at week 18, W31 ON/OFF/ON-OFF light responses, C3 endothelial monolayer integration strategy.
-
3D 打印缓释 HCTZ 片: 跟 Cellink BIO X 挤出打印 (Tasevska 2025 EJPS) - Deep-read synthesis of Tasevska T et al. 2025 European Journal of Pharmaceutical Sciences 206:106998 (DOI 10.1016/j.ejps.2024.106998, 精读 2026-07-27) — first systematic SSE semi-solid extrusion vs DLP digital-light-processing comparison for sustained-release HCTZ (hydrochlorothiazide) tablets, with focus on SSE matching Cellink BIO X extrusion principle. Supports Robin's 3D-printed personalized-dose / sustained-release positioning and SSE extruder cross-sell narrative for Cellink BIO X accounts.
-
OpenWiki 速查手册 (Robin 一次聊清, 直接拿来用) - Compact OpenWiki cheat-sheet that condenses the 2026-07-27 6-conversation walkthrough into a single ready-to-share document. Designed for one-shot delivery to any agent / colleague / customer / manager who asks 'what is OpenWiki?' — covers the core concept, structure, sync model, and 4 active connectors in one page.
-
RAGFlow 文档采样索引 - Document-level sample index of Robin's RAGFlow instance (http://10.10.10.254:9980) — 18 KBs / 32709 total docs / 401 sample entries (last sync 2026-08-03). Captures the 30-doc-per-request server-side limit + offset bug that forces 12 large KBs to single-page samples. Lets MiniMax code / Hermes see 'what's in each KB' without SQL queries; complements
/sources/ragflow.md(KB-level health + counts). -
Sources - Per-connector evidence indexes and ingestion coverage for the local knowledge wiki; durable synthesis lives on canonical pages.
-
RustDesk 自建 hbbs 完整配置笔记 (2026-08-05) - End-to-end deployment note for Robin's self-hosted RustDesk relay server (hbbs + hbbr) — Unraid NAS container (10.10.10.254) + OpenWrt reverse-proxy (10.10.10.252) + Tencent Light Cloud (43.160.207.177) + ITX client (192.168.31.14), with permanent password + key config. Supports remote control of the ITX 24h host (per
/commitments.mditx-24h-host-build-2026-q3) and Robin's travel-laptop remote-access workflow. -
CelVivo 销售弹药清单 (2026-07-16 实战索引) - Sales ammunition index for CelVivo ClinoStar / ClinoReactor — price-book SKUs (ClinoReactor 12-pack, ClinoStar, SP5D-24W, Demo kit), 8-protocol SOP series (ClinoReactor prep / media change / organoid gardening / speed-tuning / alginate / VitroGel), and 5 organ-specific application cases (HepG2, FTC-133, SCLC, TNBC, ClinoStar vs Orbital Shaker). Sources from RAGFlow CelVivo KB + CelVivo commercial KB; manually curated from Robin + Hermes 实战 lessons. Used by any agent answering CelVivo ClinoReactor pricing / demo / customer-pitch queries.
-
CelVivo + Vault 实时 mtime 快照 (2026-07-16 快照) - Historical mtime snapshot of Robin's Obsidian vault from 2026-07-16 — captures CelVivo deep-read activity (0 new entries in the 7-day window 2026-07-09 → 2026-07-16; last activity 2026-07-05), 6 entries in the 30-day window, and ClinoReactor keyword hits. Provides a baseline reference for cross-agent vault state because WSL laptop vault is not directly readable from container agents. Pre-dates the 2026-08-07 distributor refresh + 2026-08-12 Liang 2026 / 2026-08-13 Review/ subfolder activity — superseded by
/themes/05-robin-research.mdfor current vault state. -
Web Search Evidence - Compact ingestion coverage for the web-search-1 Tavily lane tracking CelVivo, 3Brain, Cellink, and related organoid news.