Robin/openwiki-wiki
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
5c70881b2827240167194c6ba095d7cd83b980df
openwiki-wiki/quickstart.md
T
OpenWiki Bot 5c70881b28 chore: daily update 2026-08-17
2026-08-17 02:05:16 +08:00

91 lines
67 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
type: "Reference"
title: "Robin's Local Knowledge Base"
description: "OpenWiki entrypoint and orientation — Robin's personal knowledge base with 5 synthesis themes (organoid equipment, bioprinting, AI agents, local LLM, personal research), navigation map, current connector status, and backlog."
tags: [quickstart, navigation, openwiki, status]
---
# Robin's Local Knowledge Base
> This is Robin's personal OpenWiki — a local, growing knowledge base that synthesizes notes, recurring themes, open questions, commitments, and personal logistics from configured sources.
## What this wiki is
- **Personal brain** — durable notes about Robin's life, work, people, projects, and recurring patterns.
- **Synthesis layer** — OpenWiki connector ingestion writes raw data under `~/.openwiki/connectors/<connector>/raw`; this wiki is the opinionated, deduplicated map on top of that evidence.
- **Living document** — pages evolve as new connector data lands. Each update run reviews `open-questions.md` and refreshes the relevant canonical pages.
## How this wiki is organized
| Page | Purpose |
| --- | --- |
| [/quickstart.md](/quickstart.md) | This page — orientation, navigation, status. |
| [/open-questions.md](/open-questions.md) | Unresolved questions about Robin's memory/wiki. Active, Answered, Stale. |
| [/themes.md](/themes.md) | Recurring topics and signals index — terse table. |
| [/commitments.md](/commitments.md) | Work tasks, follow-ups, approvals, scheduled items. |
| [/personal-logistics.md](/personal-logistics.md) | Personal errands, appointments, travel, life admin. |
| [/themes/01-organoid-equipment.md](/themes/01-organoid-equipment.md) | Theme 1 — CelVivo ClinoStar / ClinoReactor + 3Brain HD-MEA. |
| [/themes/02-bioprinting.md](/themes/02-bioprinting.md) | Theme 2 — Cellink BIO X / BIONOVA X / Lumen X 3D bioprinting. |
| [/themes/03-ai-agents.md](/themes/03-ai-agents.md) | Theme 3 — Hermes Agent + MiniMax code + Codex + OpenWiki. |
| [/themes/04-local-llm.md](/themes/04-local-llm.md) | Theme 4 — Ollama + bge-reranker + RAGFlow + LM Studio local inference. |
| [/themes/05-robin-research.md](/themes/05-robin-research.md) | Theme 5 — Robin's personal research: deep reads + sales ammunition + CRM + work rhythm. |
| [/sources/](/sources/) | Per-connector evidence indexes (one page per active connector). |
## Current status
- **Wiki state:** initialized — Hacker News, the local Obsidian vault, and Tavily web search are active and feeding `/sources/hackernews.md`, `/sources/git-repo.md`, `/sources/web-search.md`, and `/themes.md`. All 5 synthesis themes (`/themes/01-organoid-equipment.md`, `/themes/02-bioprinting.md`, `/themes/03-ai-agents.md`, `/themes/04-local-llm.md`, `/themes/05-robin-research.md`) are now generated.
- **Connectors:** Hacker News, git-repo, and web search active. The 2026-08-08, 2026-08-10, 2026-08-12, and 2026-08-15 web-search update attempts were skipped because the connector is not enabled and `TAVILY_API_KEY` is empty; the latest successful Tavily pull remains 2026-08-03 18:08 UTC. Gmail, Slack, Notion, and X are not yet configured — see "Adding sources" below.
**Latest update — hackernews-1 (2026-08-16T18:00Z).** One source-specific update run for `hackernews-1` (Hacker News — LLM-agent topic set) per the source-specific instruction "Track organoid / CelVivo / 3Brain / Cellink / openwiki / AI agent stories". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=hackernews` and wrote raw data at `~/.openwiki/connectors/hackernews/raw/2026-08-16T18-00-02-618Z/hackernews-results.json`. 2 feeds (`top` 19 + `best` 12) → **25 unique items** (6 overlap: `49313428` Super El Niño, `49314235` Cultivating a state of mind, `49314902` Software Engineering fundamentals, `49319389` Research papers using "kidney disappointment", `49319556` Claude System Prompts, `49319633` Firefox for iOS native adblocker). 6 search queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returned **zero hits across all six** — all six Algolia queries returned `nbHits=0`. The "no material HN footprint for the 4 vendor lines" finding continues to hold — this is the **fifth pull since 2026-08-07 to have a fully-zero Algolia lane** (after the 2026-08-08 03:15 + 2026-08-08 18:00 + 2026-08-13 + 2026-08-15 pulls; the 2026-08-10 + 2026-08-12 pulls had Algolia hits that were all string-collision noise rather than fully-zero); Tavily (web-search) remains the productive lane for vendor news. **Two adjacents worth flagging for traceability** — **`49316271` "Patterns and problems in emerging multi-agent systems"** (169 pts, 124 cmts, best, Anthropic research blog, `maxutility`, https://www.anthropic.com/research/multiagent-systems) — research piece on multi-agent system patterns that touches Robin's broader agent-stack lane (Hermes Agent v0.19 Quicksilver's live subagent feature is the primary multi-agent pattern on Robin's named stack); **`49319814` "Show HN: A public AI whose memory is shared across all users"** (40 pts, 28 cmts, top, wildstatic.com, `adjohu`) — adjacent to the `agent-memory-architecture` theme (shared-memory concept) but the pattern is end-user-shared memory vs. the OKF/Karpathy/MCP-Memory agent-side memory architecture. Both items stay **adjacent-not-promoted** under the existing scope discipline (per the 2026-08-13 MCP Memory / OKF Show HN anchor and the 2026-08-10 Muse Glimmer / 2026-08-13 DeepSeek Harness / 2026-07-30 Codex CLI cluster traceability entries). Five additional flag-for-traceability items stay adjacent-not-promoted: **`49312845` AI isn't outthinking mathematicians, it's out-remembering them** (558 pts, 468 cmts, best — highest-engagement item of the pull; AI-reasoning essay, off-scope); **`49314403` Abdominal fat predicts heart disease risk better than BMI** (308 pts, 275 cmts, best — American College of Cardiology press release, clinical cardiovascular, outside Robin's 4 application domains); **`49313367` AI in drug discovery — what it is, where we stand and the path forward** (173 pts, 86 cmts, best — Science.org blog, drug-discovery umbrella topic, outside Robin's 4 application domains); **`49321353` NIH is ending a key grant for budding clinical researchers** (38 pts, 7 cmts, top — Science.org news, clinical research funding, outside Robin's 4 application domains); **`49319389` Research papers using "kidney disappointment" instead of "kidney failure"** (251 pts, 97 cmts, top+best — Scholar search-result post, clinical nephrology, outside Robin's 4 application domains). **Key call**: this pull produces **no directly-relevant item for any `/themes.md` row** — the closest items either (a) fall outside Robin's 4 application domains (clinical cardiovascular / drug discovery / clinical research funding / nephrology) and stay off the `class-organoid-3d-bioprinting` watchlist per the "domain-anchored only" scope rule, or (b) sit on the AI-agent / multi-agent / shared-memory lane where the established scope rule keeps non-named-stack agent-architecture research adjacent-not-promoted unless there is a direct change to Hermes / MiniMax / OpenWiki / Codex CLI itself. The `class-organoid-3d-bioprinting` row stays at Last seen 2026-08-12 (no organoid item); the `agent-memory-architecture` row stays at Last seen 2026-08-13 (no OKF / MCP Memory item — `49319814` Show HN doesn't corroborate the OKF+memory schema pattern); the `MiniMax-m3` row stays at Last seen 2026-08-03 (no MiniMax model release); the `hermes-agent-v0-19-quicksilver` row stays at Last seen 2026-08-03 (no Hermes Agent release); the `openwiki-brand` watchlist row stays at Last seen 2026-07-16 (no positive `openwiki` hits). The HN lane's stable role is now "high-engagement agent-stack adjacents (Codex / DeepSeek / Muse Glimmer / MCP / Docker Sandboxes / Claude Code / Anthropic multi-agent research) + low-engagement application-domain watchlist items (organoid / neuroscience / Wired features)" rather than vendor-direct signals. Wiki edits this run: (a) `/sources/hackernews.md` — added new "Latest update — 2026-08-16 18:00 UTC" section ahead of the 2026-08-15 entry; demoted 2026-08-15 from "Latest" to a regular "Update" section; updated Connector section with the new raw dump path (`2026-08-16T18-00-02-618Z/hackernews-results.json`) and the new latest-fetch timestamp (`2026-08-16T18:00:39Z`); added a "Flag-for-traceability: Anthropic multi-agent systems research, Show HN shared-memory public AI" callout section; updated `tags` to include `2026-08-16`. **No `/themes.md` row change, no commitment row, no open-question row, no personal-logistics item** — the only synthesis-layer change is recording the fifth consecutive fully-zero Algolia lane streak in `/sources/hackernews.md`. Prior run was **git-repo-1 update (2026-08-15T18:07Z)** — see next paragraph.
**Latest update — git-repo-1 (2026-08-15T18:07Z).** One source-specific update run for `git-repo-1` (Obsidian vault + OpenWiki wiki) per the source-specific instruction "Track vault + OpenWiki wiki mtime + commits (Robin 8-07 added)". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=git-repo` and wrote manifest at `~/.openwiki/connectors/git-repo/raw/2026-08-15T18-07-49-711Z/manifest.json`. **Vault HEAD** unchanged at `297f106d2b137af3480261052ff080cbc48d5b48` (same as the 2026-08-12 and 2026-08-13 fetches) — **0 new vault commits** in the 2026-08-13T18:08Z → 2026-08-15T18:07Z 48-hour window; the 2026-08-12 Liang 2026 4-commit LDR cycle remains the most recent vault activity across both the 2026-08-13 and 2026-08-15 fetches. **No new untracked vault files** appeared in the 24-hour window — the untracked file set is unchanged from the 2026-08-13 fetch (23 untracked files, including the still-untracked `Review/zhou2023_computer_vision_microfluidics.md` first-entry-in-new-Review/-subfolder file, 8 × `02-Clients/Agents/区域-2026-0807.md` distributor-rebuild files, `__pycache__/...pyc` build noise, the `文悦精读流水线.yml` v1/v2/v3/v4/v5-clean variants, etc.). **Wiki HEAD** bumped from `1737454` (2026-08-13) to **`2507925e27b83b4f9d2e6f3b9c939d081a90f9f2`** (2026-08-15T02:31 +0800); **4 new wiki-side commits** in the 48-hour window — `021e929` chore: daily update 2026-08-14 (2026-08-14T02:18) + `6258717` chore(sources): refresh NocoDB/SeaFile/Wiki.js/RAGFlow counts (sources-sync 2026-08-14, 2026-08-14T02:30) + `173b58c` chore: daily update 2026-08-15 (2026-08-15T02:00, added OKF frontmatter to nocodb.md / ragflow.md / seafile.md / wikijs.md) + `2507925` chore(sources): refresh NocoDB/SeaFile/Wiki.js/RAGFlow counts (sources-sync 2026-08-15, 2026-08-15T02:31). **One operational issue surfaced**: the `2507925` 2026-08-15 sources-sync committed **empty count values** for NocoDB / SeaFile / Wiki.js / RAGFlow (NocoDB 6614 → 0 records, Wiki.js 83 → 0 unique pages, RAGFlow landing 200 → 000) — the on-disk source configs are unaffected (NocoDB / Wiki.js / RAGFlow backends at 10.10.10.254 are unchanged); this is a wiki-side CRON issue with the auto-source-sync job, not a connector config issue. Data is recoverable from `git show 1737454:sources/nocodb.md` etc. for any future recovery commit; the synthesis layer (themes / commitments / open-questions / personal-logistics) is independent of the source-page counters and is not affected. **10 uncommitted wiki-side edits** in working tree (carryover from the prior HN-1 2026-08-13 + HN-1 2026-08-15 + web-search-1 2026-08-15 + this git-repo-1 2026-08-15 runs — `.last-update.json` + `quickstart.md` + `sources/README.md` + `sources/hackernews.md` + `sources/index.md` + `sources/nocodb.md` + `sources/ragflow.md` + `sources/seafile.md` + `sources/web-search.md` + `sources/wikijs.md`; pending the next CLI orchestration commit). Wiki edits this run: (a) `/sources/git-repo.md` — refreshed frontmatter HEAD pointers (vault HEAD → `297f106d` unchanged, wiki HEAD `1737454` → `2507925`, tags + `2026-08-15`, timestamp → `2026-08-15T18:07:49Z`), refreshed the Connector / State section to the new `2026-08-15T18-07-49-711Z` raw manifest, refreshed `## obsidian-vault — working-tree status` heading + body to 2026-08-15 (0 new commits, no new untracked files; `Review/zhou2023_computer_vision_microfluidics.md` carries forward as still-untracked), refreshed `## openwiki-wiki — working-tree status` heading + body to 2026-08-15 (4 new wiki commits + 10 uncommitted files + the count-data wipe surfaced in the Notable-entity section), refreshed the Recent-commits table with the 4 new wiki commits on top of the prior 2026-08-13 cluster, refreshed the Summary-of-last-24-hours table with the 2026-08-13T18:08Z → 2026-08-15T18:07Z window data (0 vault commits / 4 wiki commits / 10 uncommitted wiki files / operational issue row), refreshed the Notable-entity section to add a "2026-08-15 sources-sync bug" entry (count-data wipe + recovery path + wiki-synthesis independence) + carry-forward note on the `Review/` subfolder entry, refreshed the Synthesis-policy section to add this run's delta summary + the operational-issue observation; (b) `/themes/05-robin-research.md` — vault-snapshot metadata refreshed (fetch timestamp → `2026-08-15T18:07:49Z`, raw-manifest path → the new `2026-08-15T18-07-49-711Z/manifest.json`); the vault HEAD pointer stays `297f106d` (unchanged across both 2026-08-13 and 2026-08-15 fetches); (c) `/quickstart.md` — added this "Latest update — git-repo-1 (2026-08-15T18:07Z)" paragraph; updated the Backlog git-repo entry to reflect the new wiki HEAD `2507925` + 4 new wiki commits + 10 uncommitted files + the operational issue note. **No `/themes.md` row change, no commitment row, no open-question row, no personal-logistics item** — the vault delta is empty (no new commits, no new untracked files) and the synthesis layer is independent of the source-page counters. Prior run was **web-search-1 attempted 2026-08-15 skipped** (fourth consecutive skip on the same code path) — see next paragraph.
**Latest update — web-search-1 attempted 2026-08-15 skipped.** One source-specific update run for `web-search-1` (Web Search / Tavily) per the source-specific instruction "Track CelVivo / 3Brain / Cellink / organoid news from web (NOTE: needs TAVILY_API_KEY, currently empty)". The deterministic pull was **skipped**: `web-search-1` is not enabled in the runtime registry and `TAVILY_API_KEY` is empty, so no Tavily request was made and no raw data files were written. This is the **fourth consecutive skip** on the same code path (after 2026-08-08 + 2026-08-10 + 2026-08-12); the latest successful Tavily pull remains 2026-08-03 18:08 UTC. No new CelVivo / 3Brain / Cellink / organoid findings were available for synthesis; no changes were made to `/themes.md`, `/commitments.md`, `/open-questions.md`, or `/personal-logistics.md` from this source attempt. Discord between on-disk `~/.openwiki/connectors/web-search/config.json` (which shows `enabled: true` + 6 Chinese queries) and the runtime registry (still reports the connector as disabled) is the same operational issue noted on the 2026-08-08 / 2026-08-10 / 2026-08-12 runs — recorded for traceability only, not a wiki-synthesis concern. Wiki edits this run: (a) `/sources/web-search.md` — added a "2026-08-15 attempted update — skipped" section ahead of the 2026-08-12 entry, following the established 5-field structure (Status / Authentication / Coverage / Synthesis / Operational note); updated the header line to mention the fourth skip; added `2026-08-15` to the `tags` list; (b) `/quickstart.md` — added this "Latest update — web-search-1 attempted 2026-08-15 skipped" paragraph; updated the Web Search connector row in "Current status" and "Adding sources" to reflect the 4-entry skip streak; updated the Backlog web-search entry likewise. **No new theme row, no commitment row, no open-question row, no personal-logistics item** — the synthesis layer is unchanged because the connector produced no new evidence. Prior run was **hackernews-1 update (2026-08-15T18:00Z)** — see next paragraph for the Auto-research with codex / Working with AI as Leadership flag-for-traceability notes.
**Latest update — hackernews-1 (2026-08-15T18:00Z).** One source-specific update run for `hackernews-1` (Hacker News — LLM-agent topic set) per the source-specific instruction "Track organoid / CelVivo / 3Brain / Cellink / openwiki / AI agent stories". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=hackernews` and wrote raw data at `~/.openwiki/connectors/hackernews/raw/2026-08-15T18-00-02-001Z/hackernews-results.json`. 2 feeds (`top` 30 IDs / 19 resolved + `best` 30 IDs / 4 resolved) → **19 unique items** (4 overlap: `49303202` Firefox uBlock Origin, `49304447` Going Dark law enforcement hacking, `49307592` "The other Sean Byrne doesn't exist", `49309549` Auto-research with codex 232x Faster Kernel). 6 search queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returned **zero hits across all six** — all six Algolia queries returned `nbHits=null`. The "no material HN footprint for the 4 vendor lines" finding continues to hold — this is the **fourth pull since 2026-08-07 to have a fully-zero Algolia lane** (after the 2026-08-08 03:15 + 2026-08-08 18:00 + 2026-08-13 pulls; the 2026-08-10 + 2026-08-12 pulls had Algolia hits that were all string-collision noise rather than fully-zero); Tavily (web-search) remains the productive lane for vendor news. **One high-engagement Robin-stack-adjacent in the 19 unique top+best items** — **`49309549` Auto-research with codex: How I achieved a 232x Faster Kernel** (275 pts, 75 cmts, top+best, sankalp.bearblog.dev, by `tosh`) — a personal-blog essay on using Codex (OpenAI coding agent) for autonomous GPU-kernel optimization. Codex is on Robin's stack as the named fallback agent surface per [`/themes/03-ai-agents.md`](/themes/03-ai-agents.md) § "Codex CLI (fallback)", and the "auto-research" framing directly speaks to how the broader Codex community is using Codex for autonomous optimization. **Adjacent-not-promoted** per the established scope rule that Codex-wrapper / TUI / usage changes stay adjacent unless there is a direct change to Hermes / MiniMax / OpenWiki / Codex CLI itself (per the 2026-07-27 / 2026-07-30 / 2026-08-03 / 2026-08-10 / 2026-08-12 / 2026-08-13 traceability entries in `/sources/hackernews.md`). Six additional flag-for-traceability items stay adjacent-not-promoted under the existing scope discipline: **`49309451` Working with AI Feels More Like Leadership Than Coding** (151 pts, 109 cmts, top — AI-coding-agent essay, off-scope); **`49310682` First At-Home Test for Infected Ticks** (111 pts, 42 cmts, top — Smithsonian, infectious-disease diagnostics, outside Robin's 4 application domains); **`49311651` Semaglutide linked to 26% lower 5-year predicted dementia risk** (95 pts, 59 cmts, top — Alzheimer's & Dementia journal, clinical neurology / pharmacology, outside Robin's 4 application domains); **`49312008` A controversial Alzheimer's surgery is said to reverse symptoms** (62 pts, 17 cmts, top — Nature news, clinical neurology, outside Robin's 4 application domains); **`49310495` Yadda 3.0.0: BDD in the Age of AI Agents** (37 pts, 16 cmts, top — BDD testing framework, off-scope); **`49310362` Could a computer scientist build a brain?** (32 pts, 19 cmts, top — computer-science neuroscience essay, outside Robin's 4 application domains). **Key call**: this pull produces **no directly-relevant item for any `/themes.md` row** — the closest items either (a) fall outside Robin's 4 application domains (clinical neurology / pharmacology / infectious disease diagnostics / computer-science neuroscience) and stay off the `class-organoid-3d-bioprinting` watchlist per the "domain-anchored only" scope rule, or (b) sit on the AI-agent / Codex-usage lane where the established scope rule keeps Codex surface / TUI / usage changes adjacent-not-promoted unless there is a direct change to Hermes / MiniMax / OpenWiki / Codex CLI itself. The `class-organoid-3d-bioprinting` row stays at Last seen 2026-08-12 (no organoid item); the `agent-memory-architecture` row stays at Last seen 2026-08-13 (no OKF / MCP Memory item); the `MiniMax-m3` row stays at Last seen 2026-08-03 (no MiniMax model release); the `hermes-agent-v0-19-quicksilver` row stays at Last seen 2026-08-03 (no Hermes Agent release); the `openwiki-brand` watchlist row stays at Last seen 2026-07-16 (no positive `openwiki` hits). The HN lane's stable role is now "high-engagement agent-stack adjacents (Codex / DeepSeek / Muse Glimmer / MCP / Docker Sandboxes / Claude Code) + low-engagement application-domain watchlist items (organoid / neuroscience / Wired features)" rather than vendor-direct signals. Wiki edits this run: (a) `/sources/hackernews.md` — added new "Latest update — 2026-08-15 18:00 UTC" section ahead of the 2026-08-13 entry; demoted 2026-08-13 from "Latest" to a regular "Update" section; updated Connector section with the new raw dump path (`2026-08-15T18-00-02-001Z/hackernews-results.json`) and the new latest-fetch timestamp (`2026-08-15T18:00:28Z`); added a "Flag-for-traceability: Auto-research with codex, Working with AI as Leadership, Yadda BDD for AI agents — adjacent, off-scope" callout section; updated `tags` to include `2026-08-15`. **No `/themes.md` row change, no commitment row, no open-question row, no personal-logistics item** — the only synthesis-layer change is recording the fourth consecutive fully-zero Algolia lane streak in `/sources/hackernews.md`. Prior run was **git-repo-1 update (2026-08-13T18:08Z)** — see next paragraph.
**Latest update — git-repo-1 (2026-08-13T18:08Z).** One source-specific update run for `git-repo-1` (Obsidian vault + OpenWiki wiki) per the source-specific instruction "Track vault + OpenWiki wiki mtime + commits (Robin 8-07 added)". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=git-repo` and wrote manifest at `~/.openwiki/connectors/git-repo/raw/2026-08-13T18-08-25-206Z/manifest.json`. **Vault HEAD** unchanged at `297f106d2b137af3480261052ff080cbc48d5b48` (same as the 2026-08-12 fetch) — **0 new vault commits** in the 24-hour window; the 2026-08-12 Liang 2026 4-commit LDR cycle remains the most recent vault activity. **1 new untracked file** appeared in the working tree: `04-Literature/Deep-Reads/Review/zhou2023_computer_vision_microfluidics.md` — **first entry in a new `Review/` subfolder** under `04-Literature/Deep-Reads/` (Zhou 2023 computer vision + microfluidics — Robin deep-read in progress, uncommitted). Also new untracked: `__pycache__/literature_deepread_v6.cpython-312.pyc` (Python build noise, ignored). **Wiki HEAD** bumped from `c57ead4` (2026-08-12) to **`1737454ad1b4f06ff9958cc9256c0c611262e44a`** (2026-08-13); **2 new wiki-side commits** — `8ddc6d9` chore: daily update 2026-08-13 (this committed the synthesis edits from the HN-1 2026-08-13 run: MCP Memory using OKF / agent-memory-architecture refresh + 1 new theme row extension on `/themes.md`) + `1737454` chore(sources): refresh NocoDB/SeaFile/Wiki.js/RAGFlow counts (sources-sync 2026-08-13). **8 uncommitted wiki-side edits** in working tree (carryover from the HN-1 2026-08-13 synthesis edits that landed between the `1737454` commit and this fetch — `.last-update.json` + `quickstart.md` + `sources/hackernews.md` + `sources/nocodb.md` + `sources/ragflow.md` + `sources/seafile.md` + `sources/wikijs.md` + `themes.md`; pending the next CLI orchestration commit). Wiki edits this run: (a) `/sources/git-repo.md` — refreshed frontmatter HEAD pointers (vault HEAD → `297f106d` unchanged, wiki HEAD → `1737454`, tags + `2026-08-13`), refreshed Connector / State section to the new `2026-08-13T18-08-25-206Z` raw manifest, updated `## obsidian-vault — working-tree status (2026-08-13)` to flag 0 new vault commits + the 2 new untracked files (Review/ zhou2023 + __pycache__), added the new files to the working-tree status table, refreshed the Summary-of-last-24-hours table with the 2026-08-12T18:08Z → 2026-08-13T18:08Z window data (no vault commits, 1 new untracked Review/ file, 2 new wiki commits), refreshed the Notable-entity section to add a `Review/` subfolder expansion entry (first review-type subfolder in the deep-read tree), refreshed the Synthesis-policy section to note this run is **a daily-cycle follow-up (no net-new external evidence)**; (b) `/themes/05-robin-research.md` — vault snapshot metadata refreshed (fetch timestamp → 2026-08-13T18:08:25Z, working-tree state 19 → 25 untracked files, raw manifest path updated), added a new `Review/` row to the deep-read breakdown table (1 file, `zhou2023_computer_vision_microfluidics.md`, marked as **untracked** — only counts once committed), updated the 30-day research output table with a new "Subfolder expansion (uncommitted)" row + extended the by-vendor table accordingly, ER-diagram entity label updated (deep-reads `~1,204` → `~1,205`), surrounding narrative extended with the "Robin appears to be expanding the folder taxonomy to separate review-type papers from original-research papers; whether this is a one-off subfolder or the start of a recurring folder pattern is a watchlist item for the next git-repo fetch" finding. **No `/themes.md` row change, no commitment row, no open-question row, no personal-logistics item** — the Review/ subfolder expansion is a single-file structural event, not a recurring signal; the next git-repo fetch will tell whether more papers land in this subfolder or it stays a single-entry watchlist. Prior run was **hackernews-1 update (2026-08-13T18:00Z)** — see next paragraph for the MCP Memory using OKF / agent-memory-architecture refresh detail.
**Latest update — hackernews-1 (2026-08-13T18:00Z).** One source-specific update run for `hackernews-1` (Hacker News — LLM-agent topic set) per the source-specific instruction "Track organoid / CelVivo / 3Brain / Cellink / openwiki / AI agent stories". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=hackernews` and wrote raw data at `~/.openwiki/connectors/hackernews/raw/2026-08-13T18-00-02-662Z/hackernews-results.json`. 2 feeds (`top` 25 + `best` 7) → **27 unique items** (5 overlap: `49285244` DeepSeek Harness developer preview, `49281916` Codex in ChatGPT desktop app for Linux preview, `49284774` Deutsche Bank yuan clearing, `49286341` Spaghettifying DRAM, `49285982` Gloomberb). 6 search queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returned **zero hits across all six** — all six Algolia queries returned `nbHits=0`. The "no material HN footprint for the 4 vendor lines" finding continues to hold (this is the third pull since 2026-08-07 to have a fully-zero Algolia lane — the prior 2026-08-08 03:15 + 2026-08-08 18:00 pulls were the previous two; the 2026-08-10 + 2026-08-12 pulls had Algolia hits that were all string-collision noise); Tavily (web-search) remains the productive lane for vendor news. **One directly-relevant adjacent** in the 27 unique top+best items — **`49286073` MCP Memory – Fast Agent Memory Using Google's OKF and SQLite FTS5** (46 pts, 22 cmts, Show HN, github.com/fellowgeek/mcp-memory) — the **first HN item since the 2026-07-16 `langchain-ai/openwiki` brand signal to explicitly name Google's OKF frontmatter** (the same schema OpenWiki uses at `/index.md` frontmatter `okf_version: "0.1"` and per concept page). Refreshes `agent-memory-architecture` row on `/themes.md` Last seen → 2026-08-13 (Evidence count → 16+, plus a new MCP Memory / OKF anchor in the row description and Evidence cell). Six flag-for-traceability items stay adjacent-not-promoted under the existing scope discipline ("open-weight AI never earns a `themes.md` row" + workflow-tooling priority LOW): **`49285244` DeepSeek Harness developer preview** (422 pts top+best, 197 cmts — 4th DeepSeek-vs-MiniMax observation in 10 days, after DeepSeek V4 Flash 0731 + Tavily benchlm.ai + DeepSeek V4 Pro 0813; open-weight agent runtime, off-scope); **`49281916` Codex in ChatGPT desktop app for Linux** (394 pts top+best, 269 cmts — Codex in ChatGPT Linux desktop app preview; Codex is on Robin's stack but the established scope rule keeps Codex surface changes adjacent-not-promoted); **`49289112` Gemini 3.7 Flash** (134 pts top, 84 cmts — off-scope); **`49285327` Choosing an AI model: one prompt, 11 models, different results** (115 pts top, 56 cmts — Netlify comparison, off-scope); **`49288889` Mistral OCR 4.1** (40 pts top, 6 cmts — off-scope); **`49283063` Launch HN: Bullet (YC S26) – A Faster Coding Agent** (17 pts top, 19 cmts — YC S26 launch, off-scope). **Key call**: MCP Memory using OKF (`49286073`) is the first HN-side corroboration of the OKF+memory schema that OpenWiki has been using since at least 2026-07-16. The 46-pt engagement is moderate (above the 1-2-pt adjacent items in the 2026-08-03 sweep, below the 100+ pt threshold for high-engagement vendor releases) — but the **direct OKF reference + MCP-server delivery channel + Show HN surfacing** makes it a stronger trigger for `agent-memory-architecture` Last-seen refresh than the prior 2-pt `49156055` verginglabs Markdown-wiki post (2026-08-03) or the 1-pt `49159148` governed/approval-gated-memory post (2026-08-03). No new theme row, no new commitment row, no new open-question row, no personal-logistics item. Wiki edits this run: (a) `/sources/hackernews.md` — added new "Latest update — 2026-08-13 18:00 UTC" section ahead of the 2026-08-12 entry; demoted 2026-08-12 from "Latest" to a regular "Update" section; updated Connector section with the new raw dump path (`2026-08-13T18-00-02-662Z/hackernews-results.json`) and the new latest-fetch timestamp (`2026-08-13T18:00:30Z`); added an "agent-memory-architecture (refreshed this run)" callout section for `49286073`; added a "Flag-for-traceability: DeepSeek Harness, Codex on Linux desktop, Gemini 3.7 Flash, Mistral OCR" callout section; updated `tags` to include `2026-08-13`; (b) `/themes.md` — refreshed `agent-memory-architecture` row: Last seen → 2026-08-13, Evidence count → 16+, row description extended with the MCP Memory / OKF evidence + Evidence cell gained a new HN anchor; all other theme rows unchanged; (c) `/quickstart.md` — added this "Latest update — hackernews-1 (2026-08-13T18:00Z)" paragraph ahead of the 2026-08-12 paragraph; updated the Backlog HN section's last-refreshed date from 2026-08-12 to 2026-08-13. Prior run was **web-search-1 attempted 2026-08-12 skipped** (third consecutive skip due to runtime registry not enabled + `TAVILY_API_KEY` empty); the latest successful Tavily pull remains 2026-08-03 18:08 UTC, and no new CelVivo / 3Brain / Cellink findings were synthesized from this skip. Prior successful run was **git-repo-1 update (2026-08-12T18:08Z)** — see second-paragraph detail in the prior-run history below.
**Latest update — git-repo-1 (2026-08-12T18:08Z).** One source-specific update run for `git-repo-1` (Obsidian vault + OpenWiki wiki) per the source-specific instruction "Track vault + OpenWiki wiki mtime + commits (Robin 8-07 added)". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=git-repo` and wrote manifest at `~/.openwiki/connectors/git-repo/raw/2026-08-12T18-08-49-138Z/manifest.json`. **Vault HEAD** bumped from `d97b39ed` (2026-08-10) to **`297f106d2b137af3480261052ff080cbc48d5b48`** (2026-08-12T10:24 +0800); 4 new commits landed in the 24-hour window — all on the **same single Liang 2026 Sci Rep liver organoid paper** (`04-Literature/Deep-Reads/Liver/2026-08-12_Liang_2026_SciRep_liver_organoid_fibrosis_platform.md`), walked through Robin's standard 3-stage LDR cycle: (a) **`e542c977`** adds the file (186 lines; Liang Y et al. Sci Rep DOI 10.1038/s41598-026-42990-2, PMID 41857098, Wiki.js Id=1716, NocoDB Id=770; 3 轨道 Hep-Orgs + Cho-Orgs + HSC + DAPT+Dex P1-P10 长程稳定成熟 + TGFβ 激活体外纤维化模型); (b) **`8699a154`** rewrites to page 946 9 项 SOP 重写版 (105+/122- lines; 6 核心发现 三段式 tables + 7 SeaFile Fig + 4-star "与我们的产品" table + 4-paper 参考文献 + 转发语); (c) **`aacba7aa`** + **`297f106d`** are the page 934 修齐 micro-revisions (frontmatter title 改中文 / H1 去括号 / 元数据用 bold 单行 / 📄PDF 文件名). **This run brings net-new external evidence into the vault** — Liang 2026 is the first paper whose `related:` field already names `[[MOC-类器官与3D培养]]` on the day of publication (the prior 2026-08-10 first-MOC framing was only prospective). **Wiki HEAD** bumped from `2bc03f3` (2026-08-09) to **`c57ead4bf115dbca5101720db63a7730476d6894`** (2026-08-12T02:30 +0800); 2 new commits on `main` — `431ee6a` chore: daily update 2026-08-12 + `c57ead4` chore(sources): refresh NocoDB/SeaFile/Wiki.js/RAGFlow counts (sources-sync 2026-08-12). **10 uncommitted wiki-side edits** in working tree (carryover from the prior HN + web-search synthesis runs this same 24-hour window — `.last-update.json` + `quickstart.md` + `sources/hackernews.md` + `sources/literature.md` + `sources/nocodb.md` + `sources/ragflow.md` + `sources/seafile.md` + `sources/web-search.md` + `sources/wikijs.md` + `themes.md`; pending the next CLI orchestration commit). Wiki edits this run: (a) `/sources/git-repo.md` — refreshed frontmatter HEAD pointers (vault HEAD → `297f106d`, wiki HEAD → `c57ead4`, tags + `2026-08-12`), refreshed Connector / State section to the new `2026-08-12T18-08-49-138Z` raw manifest, updated `## obsidian-vault — working-tree status (2026-08-12)` HEAD + Ahead-of-origin (12 → 13 commits) + added a Liang 2026 row to the working-tree status table (the file moved from untracked → committed via `e542c977`), refreshed the Recent commits table with the 4 new commits on top of the prior 2026-08-10 cluster, refreshed `## openwiki-wiki — working-tree status (2026-08-12)` HEAD + the uncommitted-10-files enumeration for this window's synthesis-layer carryover, added a new **Liang 2026 liver organoid fibrosis platform (2026-08-12)** Notable-entity entry (the paper's 4-star ClinoStar + ClinoReactor + PhaseHolographic + 3Brain HD-MEA table; MOC linkage evidence; customer-entry scenarios), refreshed the Synthesis-policy section with this run's wiki edits summary + 2 process signals worth watching; (b) `/themes/05-robin-research.md` — vault HEAD `d97b39ed` → `297f106d`, "Latest cluster (2026-08-10)" → "Latest cluster (2026-08-12)" with Liang 2026 cluster summary, ER-diagram entity label updated (vault HEAD + deep-reads `~1,202` → `~1,204`), 30-day research output table extended with a new top-row entry for the Liang 2026 cluster, by-vendor table updated to flag CelVivo `Most recent` → `2026-08-12 (Liang 2026 — ClinoStar rating 5 + ClinoReactor rating 5)` and CelVivo additions row extended to include Liang 2026, total deep-read count `1,202` → `1,203` and the `Liver/` subfolder count `6` → `7`, surrounding paragraph extended with the "Liang 2026 first paper with active MOC `related:` back-link on publication day" finding; (c) `/themes/01-organoid-equipment.md` — added Liang 2026 to the CelVivo evidence anchors (ClinoStar rating 4 + ClinoReactor rating 4 + PhaseHolographic rating 3 + 3Brain HD-MEA rating 3 for 肝-神经互作延伸), extended "What to read next" with the Liang 2026 vault-path row; (d) `/commitments.md` — added `liang-2026-celvivo-clinostar-positioning` (Owner: me, fold Liang 2026 长程成熟 + 体外纤维化模型 story into CelVivo ClinoStar / ClinoReactor sales narrative; parallel to existing `miao-2025-celvivo-clinostar-positioning`), tags updated. **No `/themes.md` row change** — the Liang 2026 evidence is a single-paper flywheel-sync output, not a recurring signal; it routes into the existing `class-organoid-3d-bioprinting` evidence without needing a new theme row. **No `/open-questions.md` row, no `/personal-logistics.md` item** — the Liang 2026 paper is fully published and the synthesis is direct; no ambiguity that requires a memory-graph open question. Prior successful run was **hackernews-1 update (2026-08-12T18:00Z)** — see the next paragraph's "Prior run was" detail.
**Latest update — web-search-1 attempted 2026-08-12 skipped.** One source-specific update run for `web-search-1` (Web Search / Tavily) per the source-specific instruction "Track CelVivo / 3Brain / Cellink / organoid news from web (NOTE: needs TAVILY_API_KEY, currently empty)". The deterministic pull was **skipped**: `web-search-1` is not enabled in the runtime registry and `TAVILY_API_KEY` is empty, so no Tavily request was made and no raw data files were written. This is the third consecutive skip on the same code path (after 2026-08-08 + 2026-08-10); the latest successful Tavily pull remains 2026-08-03 18:08 UTC. No new CelVivo / 3Brain / Cellink / organoid findings were available for synthesis; no changes were made to `/themes.md`, `/commitments.md`, `/open-questions.md`, or `/personal-logistics.md` from this source attempt. Discord between on-disk `~/.openwiki/connectors/web-search/config.json` (which shows `enabled: true` + 6 Chinese queries) and the runtime registry (still reports the connector as disabled) is the same operational issue noted on the 2026-08-08 and 2026-08-10 runs — recorded for traceability only, not a wiki-synthesis concern. Wiki edits this run: (a) `/sources/web-search.md` — added a "2026-08-12 attempted update — skipped" section ahead of the 2026-08-10 entry, following the established 5-field structure (Status / Authentication / Coverage / Synthesis / Operational note); updated the header line to mention the third skip; added `2026-08-12` to the `tags` list; (b) `/quickstart.md` — added this "Latest update — web-search-1 attempted 2026-08-12 skipped" paragraph ahead of the existing "Last update" paragraph; updated the Web Search connector row in "Current status" and "Adding sources" to reflect the 3-entry skip streak; updated the Backlog web-search entry likewise. **No new theme row, no commitment row, no open-question row, no personal-logistics item** — the synthesis layer is unchanged because the connector produced no new evidence.
- **Last update:** see [/.last-update.json](/.last-update.json) — this run is **hackernews-1 update (2026-08-12T18:00Z)** for the LLM-agent topic set per the source-specific instruction "Track organoid / CelVivo / 3Brain / Cellink / openwiki / AI agent stories". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=hackernews` and wrote raw data at `~/.openwiki/connectors/hackernews/raw/2026-08-12T18-00-01-956Z/hackernews-results.json`. 2 feeds (`top` 25 + `best` 10) → **30 unique items** (5 overlap: `49271994` AI middle class software engineering, `49272832` Tailscale SQLite WAL-Reset bug, `49270953` 2026 Eclipse Webcams, `49273165` license plate warrant, `49274600` DeepSeek V4 Pro 0813). 6 search queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returned **zero vendor-anchored hits** — `CelVivo` returned `nbHits=0`; the 11 `3Brain` hits and 4 `Cellink` hits are string-collision noise on "brain" / author fragments; the 7 `organoid` hits include one real organoid feature (`49266068` "AI Is Dead. Organoids Are Alive", 5 pts, Wired) + 6 string-collision noise hits. `deepagents` returned `nbHits=0`. `openwiki` returned 1 hit (`49272534` "Qwen 3.8-27B goes openweight in 2 days", 15 pts, huggingface.co/Qwen/Qwen3.8-27B) — **string collision** on the literal "open" prefix + "openweight" substring; not the `langchain-ai/openwiki` brand signal from 2026-07-16. **One confirmed organoid hit** → `class-organoid-3d-bioprinting` row on `/themes.md` Last seen → 2026-08-12 (Evidence count → 27+, plus a new Wired source anchor). All other theme rows unchanged. Six flag-for-traceability items stay adjacent-not-promoted under the existing scope discipline ("open-weight AI never earns a `themes.md` row" + workflow-tooling priority LOW): **`49274600` DeepSeek V4 Pro 0813** (274 pts top+best — third observation in 8 days on the DeepSeek V4 → MiniMax-M3/M3-H3 competitive-benchmarks lane, after `49214008` DeepSeek V4 Flash 0731 ARC-AGI + the Tavily benchlm.ai comparison; **recurrence trigger met per the "third corroborating item" criterion** but the annex would be DeepSeek-V4-Pro-0813-specific rather than the always-on-agent-model cluster framed by the 2026-08-10 Muse Glimmer entry, so no row description extension is added this run); **`49265051` WorldClaw Agentic 3D** (265 pts best, Tencent Hunyuan3D — 3D-scene generation, not 3D bioprinting); **`49267928` llama.cpp** (333 pts best, llama.app promo page — **strongest adjacents to `/themes/04-local-llm.md` (canonical backend for Ollama / LM Studio)** but third-party domain rather than upstream release, so no theme row added per the open-weight-AI scope rule); **`49273478` Qwen3.8-2.4T-A95B** (200 pts top, open-weight AI, off-scope); **`49272484` "My Agent Setup"** (73 pts top — adjacent agent tooling); **`49273175` "Hax" coding agent in C** (39 pts top — adjacent agent tooling). **Key call**: this is the **first pull in the 6-query LLM-agent topic set that produces a confirmed organoid hit** (after six consecutive no-promotable-hits pulls since 2026-08-07); the Wired organoid feature is a recurring-application-domain signal rather than a vendor launch / customer win / breakthrough / new product release, so it refreshes Last seen on the existing theme row rather than earning a new theme row or commitment row. The DeepSeek V4 Pro 0813 + WorldClaw + llama.cpp + GitHub-PR/Issues cluster continues the prior pull's pattern of "no promotable vendor-channel signal but adjacent tooling + competitive-benchmarks inference is the dominant workflow signal" — this is normal cadence for a 24-hour window against the 6-query set. Wiki edits this run: (a) `/sources/hackernews.md` — added new "Latest update — 2026-08-12 18:00 UTC" section ahead of the 2026-08-10 entry; demoted 2026-08-10 from "Latest" to a regular "Update" section; updated Connector section with the new raw dump path (`2026-08-12T18-00-01-956Z/hackernews-results.json`) and the new latest-fetch timestamp (`2026-08-12T18:00:30Z`); added a "class-organoid-3d-bioprinting (refreshed this run)" callout section for `49266068`; added a "Flag-for-traceability: DeepSeek V4 Pro 0813, WorldClaw 3D, llama.cpp, GitHub PR/Issues incident" callout section; updated `tags` to include `2026-08-12`; (b) `/themes.md` — refreshed `class-organoid-3d-bioprinting` row: Last seen → 2026-08-12, Evidence count → 27+, row description extended with the Wired organoid feature + Evidence cell gained a new Wired anchor; all other theme rows unchanged; (c) `/quickstart.md` — added this "Latest update — hackernews-1 (2026-08-12T18:00Z)" paragraph ahead of the existing "Last update" paragraph; updated the Backlog HN section's last-refreshed date from 2026-08-10 to 2026-08-12. **No commit row, no open-question row, no personal-logistics item** — the only synthesis-layer change is the `class-organoid-3d-bioprinting` Last-seen refresh. Prior run was **git-repo-1 update (2026-08-10T18:06Z)** for the Obsidian vault + OpenWiki wiki per the source-specific instruction "Track vault + OpenWiki wiki mtime + commits (Robin 8-07 added)". Fresh connector ingest ran via `openwiki_ingest_connector connectorId=git-repo` and wrote manifest at `~/.openwiki/connectors/git-repo/raw/2026-08-10T18-06-18-966Z/manifest.json`. **Vault HEAD** bumped from `d7df3830` (2026-08-08) to **`d97b39edb328b7e15d28de4a54a6d0bc2dd51274`** (2026-08-10T14:03Z); 3 new commits landed in the 24-hour window — all frontmatter-hygiene + MOC-discipline, no new external evidence: (a) **`6b717aef`** MOC 方案 1 样品 — **first MOC** created at `04-Literature/Deep-Reads/MOCs/MOC-类器官与3D培养.md` (95 lines, 沙牛《AI 时代的 Obsidian 公开课》课时 1 pattern; `type: moc`; `related:` already names `[[MOC-3Brain-HD-MEA]]` as next planned MOC plus `[[02-Clients/Agents/_known_associations]]` + `[[02-Active-Deals/CellShot-TriArm-Demo-计划-v2-老师版-2026-07-20]]`); 7 existing deep-reads got `related:` back-link frontmatter; (b) **`3b929c2f`** wang2024 方案 2 样品 — `wang2024-org-on-chip.md` got 6-field frontmatter (doi/journal/year/volume/pages/authors/pmid) + 7 image alt descriptions; (c) **`d97b39ed`** Phase A 批量 — 6 deep-reads received the 6-field frontmatter: Organoids 2026 review (Exp Hematol Oncol, PMID 41545895), CelVivo PAH Metabolomics (bioRxiv), Shin 2024 thalamocortical organoids 22q11ds (3Brain folder), Fernandez Vallone 2026 Cells THSDC, Liu 2026 NBME shape-conformal organoid, 类器官综述_CMOS_HD_MEA_类器官电生理_2026-06-05. **Wiki HEAD** stayed at **`2bc03f3ddb54e44d53dfed9366daeee109b82e87`** (2026-08-09T18:30Z, the sources-sync refresh recorded in the prior 2026-08-10 update — no new wiki-side commits landed in the window). Wiki edits this run: (a) `/sources/git-repo.md` — refreshed frontmatter HEAD pointers + added the 3 new commits to the recent-commits table + added a "Phase A metadata batch + first MOC (2026-08-10)" Notable-entity entry + added a "Process signal worth watching" pointer in the synthesis-policy section; (b) `/themes/05-robin-research.md` — vault HEAD `d97b39ed` (12 commits since 2026-08-03) + added a new top-row entry to the 30-day research-output table for "Phase A metadata batch + first MOC" + extended the by-vendor table with a "(Process discipline)" row + added 2 backlog items: "MOC pattern extension (2026-08-10 onward)" and "Phase A frontmatter batch rollout (2026-08-10 onward)"; (c) `/themes/05-robin-research.md` Mermaid ER diagram — added the `MOC` entity to the vault children. **No theme-row promotion, no commitment row, no open-question row, no personal-logistics item** — the new commits are process discipline, not new external evidence, and the prior 2026-08-08 update already captured the substantive distributor / CRM / Liu-2026-roadmap signal. Prior run was **hackernews-1 update (2026-08-10)** for the LLM-agent topic set per the source-specific instruction "Track organoid / CelVivo / 3Brain / Cellink / openwiki / AI agent stories". 2 feeds (`top` 27 + `best` 9) → **31 unique items** (5 overlap: `49241679` Muse Glimmer, `49244569` Kinney Drugs AI phone-assistant, `49245023` Mars Bar 1991, `49245359` Cognitive Commons, `49246057` Stoa Markets GPU marketplace). 6 search queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returned **zero vendor-anchored hits** — all `organoid`/`3Brain`/`Cellink` hits are string-collision noise (e.g. `49245612` PicPocket "organize", `49243294` "brainbaking.com", `49246057` Stoa Markets GPU marketplace substring-hit on "Cellink"); `openwiki` + `deepagents` returned `nbHits=0`; `CelVivo` returned `nbHits=0`. The "no material HN footprint for the 4 vendor lines" finding still holds; Tavily (web-search) remains the productive lane for vendor news. Five high-engagement AI-agent / inference items are flagged for traceability but stay adjacent-not-promoted under the existing scope discipline ("open-weight AI never earns a `themes.md` row" + workflow-tooling priority LOW): **`49241679` Muse Glimmer** (759 pts, 420 cmts — Meta research blog, 30B open-weights model "optimized for always-on local agent workflows", highest engagement of the pull) + **`49241749` Muse Glimmer developer-portal variant** (9 pts, 2 cmts, same Meta release); **`49239751` Docker Sandboxes** (517 pts, 312 cmts — disposable isolated sandboxes for AI agents); **`49239021` Auto mode now default in Claude Code** (268 pts, 286 cmts — Anthropic product update; Claude Code is on Robin's stack but the established scope rule keeps wrapper/TUI changes adjacent-not-promoted); **`49245437` Ante offline coding agent** (66 pts, 39 cmts — Rust single-binary offline agent). **Key adjacent call:** Muse Glimmer (Meta, 30B, open-weights, always-on-agent framing) is the second open-weight agent-model release in the same week-and-a-half as the 2026-08-03 MiniMax-H3 launch cluster — this establishes a recurrence pattern on the "always-on agent model releases" lane but does not meet the brief's "recurrence bar" by itself (one observation = not enough; a third corroborating item on a future pull would justify an annex on the `/themes.md` `MiniMax-m3` row description). Wiki edits this run: (a) `/sources/hackernews.md` — added new "Latest update — 2026-08-10 18:00 UTC" section ahead of the 2026-08-08 18:00 entry; demoted 2026-08-08 18:00 from "Latest" to a regular "Update" section; updated Connector section with the new raw dump path (`2026-08-10T18-00-02-327Z/hackernews-results.json`) and the new latest-fetch timestamp (`2026-08-10T18:00:52Z`); added two flag-for-traceability callout sections for Muse Glimmer and the Claude Code / Docker / Ante adjacent cluster; updated `tags` to include `2026-08-10`. **No theme-row promotion, no commitment row, no open-question row, no personal-logistics item** — the empty-pulls-on-vendor-queries pattern continues to be normal cadence (this is the **fourth consecutive "no promotable hits" pull** under the new 6-query set). Prior run was **git-repo-1 update (2026-08-08)** for the Obsidian vault + OpenWiki wiki. Vault HEAD bumped from `1a81e485` (2026-07-29) to `d7df3830` (2026-08-08); 9 new commits landed on `main` in the last 5 days, with 19 untracked files in the working tree (8 of which are the new `02-Clients/Agents/区域-2026-0807.md` distributors-rebuild + the new `_known_associations.md` association index). The 2026-08-07 distributor refresh (Hermes hook 4) replaces the 2026-07-24 baseline (261 dealers → 259 dealers: 华北 74 → 72) and identifies **济南成全生物科技有限公司 (曹沛起)** as the only CelVivo ClinoStar-listed agent in the 259-dealer list (cross-region: 华北 + 山东); the same hook also maps FemtoBiomed CellShot → 9X Bio (汇佰生物) → TriArm Therapeutics (翁靖傑, 128 万 Phase1 / 8-10 至 8-14 现场 demo), and finds 3Brain / Cellink / FemtoBiomed direct agents → 0 in the 259 baseline. Wiki edits this run: (a) `/sources/git-repo.md` — fully rewritten from the 2026-08-03 stale state to the 2026-08-08 fetch (vault HEAD `d7df3830`, 9 new commits, 19 untracked files, 4 vendor-paper deep-reads since 2026-08-03, 8 distributors-rebuild + association index); (b) `/sources/distributors-china-2026.md` — replaced 2026-07-24 baseline with 2026-08-07 baseline (259 dealers; full per-region Δ table; new `_known_associations.md` index table); (c) `/sources/literature.md` — added Rosero 2026 Microbioreactor + Fernandez Vallone 2026 Cells THSDC papers to the 30-day table; (d) `/themes/05-robin-research.md` — vault HEAD updated to `d7df3830`; CRM section updated to reflect 2026-08-07 `_known_associations.md` partial resolution for 董慧群 → 济南成全; distributor map section updated with 8-07 baseline numbers + 4-vendor break-out partial resolution; 30-day research output table extended with 2026-08-06 Parfitt (DJ1 LOF midbrain organoid PD) + 2026-08-08 Rosero (Microbioreactor Mini Review) + 2026-08-08 Fernandez Vallone (Cells THSDC); (e) `/open-questions.md` — updated `agent-dong-huiqun-celvivo` with 2026-08-07 partial resolution (济南成全 candidate); added 2 new questions: `chengdu-baile-vs-lidesai-for-ou-xianhong` (成都 dealer 2选1 for 西南医大殴贤红), `distributors-3brain-cellink-zero-coverage` (259 dealer list shows 0 三Brain / 0 Cellink / 0 FemtoBiomed direct agents — channel gap or xlsx gap?); (f) `/commitments.md` — added 2 new commitments: `dong-huiqun-ji-nan-chengquan-link` (confirm 董慧群 ↔ 济南成全 link), `distributors-2026-08-07-refresh` (落地 2026-08-07 distributor baseline to wiki + 跨工具). Prior run was **web-search-1 update attempt (2026-08-08)** — **skipped** because the connector is not enabled and `TAVILY_API_KEY` is empty; no raw Tavily evidence was written and no CelVivo / 3Brain / Cellink findings were added to `/themes.md`, `/commitments.md`, `/open-questions.md`, or `/personal-logistics.md` from this source attempt. Wiki edits in that prior run were limited to: (a) `/sources/web-search.md` — added a "2026-08-08 attempted update — skipped" section recording the connector status + the missing API key + the absence of new 24-hour evidence; replaced the `openwiki_generated` placeholder front matter with explicit `type` / `title` / `description` / `tags`; (b) `/sources/README.md` — moved Web Search from "not configured" to "active" with the latest-successful-pull and skip note; refreshed the connector roll table with the same status; replaced placeholder front matter; (c) `/themes.md` — replaced placeholder front matter; repaired six broken internal-link markers in the theme table (the heading anchors `2026-07-27-1804-pull`, `2026-08-08`, `china-cde-cell-gene-therapy-classification-source-backed`, `MiniMax-m3-watchlist`, `hermes-agent-best-models-eval-source-backed-workflow-tooling-low-priority`, `MiniMax-h3-launch-cluster-refreshed-this-run` did not match the current headings in the source pages); no theme rows were added or promoted (no new Tavily evidence); (d) `/quickstart.md` (prior run) — added a "Latest update — web-search-1 attempted 2026-08-08 skipped" paragraph ahead of the existing hackernews / literature / web-search prior-run history; updated the wiki state note to mention the 2026-08-08 skip; updated the Web Search connector row in "Adding sources" with the missing-API-key caveat. Prior successful source run was **hackernews-1** (`2026-08-08T18:00:44Z`) — third consecutive "no promotable hits" pull under the simplified 6-query LLM-agent topic set (after 2026-08-07 12:29 and 2026-08-08 03:15 UTC); 2 feeds (`top`, `best`) → 23 unique items; all 6 queries (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`) returned `nbHits=null` / 0 hits. Zero vendor-anchored life-science items, zero positive `openwiki` hits, three flag-for-traceability items (`49214468` Managing AI Coding Costs at Scale 290 pts best databricks blog, `49220609` OpenAI accidental attack on Hugging Face timeline 192 pts top simonwillison.net, `49215786` Claude Bluetooth-tracking anecdote 100 pts top) are adjacent-only and stay off `/themes.md` per the brief's "Workflow tooling (priority LOW) — surface story-worthy changes only" rule and the "low-engagement items are watchlist by default; promotion requires recurrence, strong engagement, or corroboration" rule. The third consecutive empty-pull under the new 6-query set **confirms the empty-pulls-on-vendor-queries pattern is normal cadence**, not an outage — no config-side intervention needed. `/sources/hackernews.md` updated with a new "Latest update — 2026-08-08 18:00 UTC" section + raw-dump pointer; `openwiki-brand` watchlist row stays at Last seen 2026-07-16 (no positive `openwiki` hits this pull — confirms the absence is normal cadence). Prior run was **literature-flywheel 2026-08-08** (today's flywheel-sync filled in 3 missing sinks: Liu 2026 Nat Biomed Eng shape-conformal porous framework + Miao 2025 Cell v2 vascularized lung/gut organoids + Sharma 2026 Cell Stem Cell vascularized retinal organoids). Wiki index rebuilt: `/sources/literature.md` reorganized with proper frontmatter + the 3 new papers added to "最近 30 天精读" (mtime 倒序) + the "按 Vendor 统计" table extended with 2026-08-08 additions per vendor; three new deep-read synthesis pages created (`/sources/literature-sharma-2026-cell-stem-cell-vascularized-retinal-organoid-rgc-deepread.md`, `/sources/literature-miao-2025-cell-vascularization-v2-deepread.md`, `/sources/literature-liu-2026-nbme-shape-conformal-organoid-deepread.md`); `/themes/01-organoid-equipment.md` CelVivo + 3Brain evidence anchors extended with the 3 new papers + the CelVivo + 3Brain vascularized organoid workflow chain; `/themes/02-bioprinting.md` application-domain evidence extended with Liu 2026 microlattice fabrication evidence; `/themes/05-robin-research.md` 30-day research output table extended with 2026-08-08 additions per vendor; `/themes.md` `class-organoid-3d-bioprinting` row Last seen → 2026-08-08 (Source now includes `literature-flywheel`, Evidence count → 26+). Three new commitments added: `liu-2026-hd-mea-roadmap-implication` (verify with 3Brain whether the Liu 2026 shape-conformal framework is on their product roadmap, HD-MEA rating 5), `miao-2025-celvivo-clinostar-positioning` (fold Miao 2025 vHLPO co-differentiation story into CelVivo ClinoStar sales narrative, ClinoStar rating 5), `robin-picked-papers-nocodb-table` (capture the structural change — Liu 2026 is the first entry in the new NocoDB table `Robin_Picked_Papers`, created 2026-08-08 per Robin instruction). Two new open questions added: `liu-2026-rogers-lab-vs-3brain-relationship` (partnership / licensing / competitive pressure?), `sharma-2026-retinal-organoid-mea-cross-sell` (does the 60-channel MCS MEA route transfer cleanly to 3Brain HD-MEA?). Prior run before that was **web-search-1** (`2026-08-03T18:08:50Z`) — Tavily **recovered from the 3-day outage streak** (pulls 8/9/10 on 2026-07-31 / 2026-08-01 / 2026-08-02 all returned `"error": "fetch failed"`); pull 11 is the first successful pull after the streak and surfaces three scope-aligned findings: (a) **CELLINK × Volumetric "new Lumen X" release** (3DPrint.com 2026-07-28, score 0.819); (b) **3Brain Utrecht Summer School "Human Brain Models" course** LinkedIn post 2026-07-29 (score 0.808); (c) **CelVivo ClinoStar 2 + Thailand Lab 2026** (labindia-analytical.com News & Events 2026-07-31, score 0.696). Tavily answer-text quality improved on 3Brain and Cellink this pull but remained partially fabricated on the Hermes/M3 lane ("release date of June 1, 2026") — treat Tavily answer text as untrusted; trust the underlying URL + score + published_date.
## How to use this wiki
- **Ask questions about the wiki** — Robin can ask things like "what's on my plate this week?" or "what recurring themes have you noticed?" and answers come from these pages first.
- **Update on demand** — `openwiki personal` opens the personal chat; `openwiki personal --update "<message>"` ingests configured connectors and refreshes synthesis pages.
- **Trust labels** — facts use `confirmed`, `source-backed`, `watchlist`, or `saved-context` so Robin can see how solid each claim is.
## Adding sources
When Robin is ready, configure connectors to start populating the wiki:
- **Gmail** — `openwiki auth google` (needs Google OAuth refresh token)
- **Slack** — `openwiki auth slack` (needs Slack user token with `search:read` for definitive self-message search)
- **Notion** — configure Notion MCP via `~/.openwiki/connectors/notion/`
- **X / Twitter** — `openwiki auth x` (needs `OPENWIKI_X_ACCESS_TOKEN`)
- **Web Search** — set `TAVILY_API_KEY` (active; latest successful pull 2026-08-03 18:08 UTC; the 2026-08-08, 2026-08-10, and 2026-08-12 update attempts were skipped because the connector is not enabled in the runtime registry — see [/sources/web-search.md](/sources/web-search.md))
- **Hacker News** — public feeds, no auth needed
- **Local git repo** — point the `git-repo` connector at a repo path
Auth instructions and env var names live in the OpenWiki operations docs and `~/.openwiki/skills/write-connector.md`.
## Backlog
Deferred until Robin configures connectors or asks for specific documentation:
- **Gmail evidence index** — `/sources/gmail.md` (no Gmail connector configured yet)
- **Slack evidence index** — `/sources/slack.md` (no Slack connector configured yet)
- **Notion evidence index** — `/sources/notion.md` (no Notion MCP configured yet)
- **X / Twitter evidence index** — `/sources/x.md` (no X connector configured yet)
- **Hacker News evidence index** — `/sources/hackernews.md` (active; last refreshed 2026-08-16 18:00 UTC — source-specific LLM-agent topic set (`organoid`, `CelVivo`, `3Brain`, `Cellink`, `openwiki`, `deepagents`); **fifth pull since 2026-08-07 to have a fully-zero Algolia lane** (after 2026-08-08 03:15 + 2026-08-08 18:00 + 2026-08-13 + 2026-08-15 — the 2026-08-10 and 2026-08-12 pulls had Algolia hits that were all string-collision noise rather than fully-zero) — vendor-anchored footprint remains null on the HN lane and Tavily (web-search) remains the productive lane; **two adjacents worth flagging for traceability surfaced this pull: `49316271` "Patterns and problems in emerging multi-agent systems"** (169 pts, 124 cmts, best, Anthropic research blog, `maxutility`, https://www.anthropic.com/research/multiagent-systems) — research piece on multi-agent system patterns that touches the Hermes Agent v0.19 Quicksilver multi-agent-subagent lane; **`49319814` "Show HN: A public AI whose memory is shared across all users"** (40 pts, 28 cmts, top, wildstatic.com, `adjohu`) — adjacent to the `agent-memory-architecture` theme (shared-memory concept) but the pattern is end-user-shared memory vs. the OKF/Karpathy/MCP-Memory agent-side memory architecture. **Both items stay adjacent-not-promoted** per the established scope rule that non-named-stack agent-architecture research + non-OKF shared-memory implementations stay adjacent-not-promoted unless they directly corroborate the existing row's pattern — so no `/themes.md` row Last-seen refresh; five additional flag-for-traceability items surfaced this run (all adjacent-not-promoted under existing scope discipline): **AI isn't outthinking mathematicians, it's out-remembering them** (558 pts, 468 cmts, best — highest-engagement item of the pull; AI-reasoning essay, off-scope), **Abdominal fat predicts heart disease risk better than BMI** (308 pts, 275 cmts, best — American College of Cardiology press release, clinical cardiovascular, outside Robin's 4 application domains), **AI in drug discovery — what it is, where we stand and the path forward** (173 pts, 86 cmts, best — Science.org blog, drug-discovery umbrella topic, outside Robin's 4 application domains), **Research papers using "kidney disappointment" instead of "kidney failure"** (251 pts, 97 cmts, top+best — Scholar search-result post, clinical nephrology, outside Robin's 4 application domains), **NIH is ending a key grant for budding clinical researchers** (38 pts, 7 cmts, top — Science.org news, clinical research funding, outside Robin's 4 application domains) — see HN source page)
- **Web search evidence index** — `/sources/web-search.md` (active; last successful pull 2026-08-03 18:08 UTC — first pull after the 3-day outage streak; CELLINK × Volumetric new Lumen X + 3Brain Utrecht Summer School + CelVivo ClinoStar 2 / Thailand Lab 2026 surfaced, see source page; 2026-08-08, 2026-08-10, 2026-08-12, 2026-08-15, and 2026-08-16 update attempts skipped — connector not enabled in runtime registry, `TAVILY_API_KEY` missing; disk config shows `enabled: true` + 6 Chinese queries but the runtime registry disagrees — operational concern, not a wiki-synthesis concern)
- **Git repo evidence index** — `/sources/git-repo.md` (active; last refreshed 2026-08-15 18:07 UTC — covers Robin's Obsidian vault (HEAD `297f106d` 2026-08-12, **unchanged across both the 2026-08-13 and 2026-08-15 fetches** — 0 new vault commits in the 2026-08-13T18:08Z → 2026-08-15T18:07Z 48-hour window) + OpenWiki wiki (HEAD `2507925` 2026-08-15, **4 new wiki commits since 2026-08-13**: `021e929` daily update 2026-08-14 + `6258717` sources-sync 2026-08-14 + `173b58c` daily update 2026-08-15 + `2507925` sources-sync 2026-08-15); 13 vault commits since 2026-08-03 fetch including the **2026-08-12 Liang 2026 Sci Rep liver organoid fibrosis platform 4-commit LDR cycle** (`e542c977` adds file 186 lines → `8699a154` page 946 9 项 SOP 重写版 105+/122- → `aacba7aa` frontmatter title 改中文 1 line → `297f106d` page 934 修齐 5+/4-) — **net-new external evidence** (DOI 10.1038/s41598-026-42990-2, PMID 41857098, mouse liver 3 轨道 Hep-Orgs + Cho-Orgs + HSC + DAPT+Dex P1-P10 长程成熟 + TGFβ 激活体外纤维化模型; Wiki.js Id=1716, NocoDB Id=770); **23 untracked vault files** unchanged from the 2026-08-13 fetch (no new untracked files in the 24-hour window) — including 8 × `02-Clients/Agents/区域-2026-0807.md` distributors-rebuild + `_known_associations.md` association index + **the first entry in the new `Review/` subfolder** (`04-Literature/Deep-Reads/Review/zhou2023_computer_vision_microfluidics.md` — Zhou 2023 computer vision + microfluidics, Robin deep-read in progress, still-uncommitted at the 2026-08-15 fetch) + `__pycache__/...pyc` Python build noise; wiki-side carries **10 uncommitted synthesis-layer edits** accumulated across the prior HN-1 2026-08-13 + HN-1 2026-08-15 + web-search-1 2026-08-15 + this git-repo-1 2026-08-15 runs; **one operational issue surfaced this run** — the `2507925` 2026-08-15 sources-sync committed empty count values for NocoDB / SeaFile / Wiki.js / RAGFlow (NocoDB 6614 → 0 records, Wiki.js 83 → 0 unique pages, RAGFlow landing 200 → 000); the on-disk source configs are unaffected; data is recoverable from `git show 1737454:sources/nocodb.md` etc. — recorded for traceability on `/sources/git-repo.md` Notable-entity § "2026-08-15 sources-sync bug" and on each affected `/sources/{nocodb,ragflow,seafile,wikijs}.md` page; the synthesis layer (themes / commitments / open-questions / personal-logistics) is independent of the source-page counters and is not affected)
- **Project / people / company pages** — only created when durable entities emerge from ingested data
These are intentionally blank until there is real evidence to put in them. Populating them with placeholders would dilute the synthesis layer..esis layer..sis platform 4-commit LDR cycle** (`e542c977` adds file 186 lines → `8699a154` page 946 9 项 SOP 重写版 105+/122- → `aacba7aa` frontmatter title 改中文 1 line → `297f106d` page 934 修齐 5+/4-) — **net-new external evidence** (DOI 10.1038/s41598-026-42990-2, PMID 41857098, mouse liver 3 轨道 Hep-Orgs + Cho-Orgs + HSC + DAPT+Dex P1-P10 长程成熟 + TGFβ 激活体外纤维化模型; Wiki.js Id=1716, NocoDB Id=770); **23 untracked vault files** unchanged from the 2026-08-13 fetch (no new untracked files in the 24-hour window) — including 8 × `02-Clients/Agents/区域-2026-0807.md` distributors-rebuild + `_known_associations.md` association index + **the first entry in the new `Review/` subfolder** (`04-Literature/Deep-Reads/Review/zhou2023_computer_vision_microfluidics.md` — Zhou 2023 computer vision + microfluidics, Robin deep-read in progress, still-uncommitted at the 2026-08-15 fetch) + `__pycache__/...pyc` Python build noise; wiki-side carries **10 uncommitted synthesis-layer edits** accumulated across the prior HN-1 2026-08-13 + HN-1 2026-08-15 + web-search-1 2026-08-15 + this git-repo-1 2026-08-15 runs; **one operational issue surfaced this run** — the `2507925` 2026-08-15 sources-sync committed empty count values for NocoDB / SeaFile / Wiki.js / RAGFlow (NocoDB 6614 → 0 records, Wiki.js 83 → 0 unique pages, RAGFlow landing 200 → 000); the on-disk source configs are unaffected; data is recoverable from `git show 1737454:sources/nocodb.md` etc. — recorded for traceability on `/sources/git-repo.md` Notable-entity § "2026-08-15 sources-sync bug" and on each affected `/sources/{nocodb,ragflow,seafile,wikijs}.md` page; the synthesis layer (themes / commitments / open-questions / personal-logistics) is independent of the source-page counters and is not affected)
- **Project / people / company pages** — only created when durable entities emerge from ingested data
These are intentionally blank until there is real evidence to put in them. Populating them with placeholders would dilute the synthesis layer..
Reference in New Issue View Git Blame Copy Permalink