From bc92e1a6b67b70ad74b397c6785f49797f9a67cd Mon Sep 17 00:00:00 2001 From: Cursor Agent Date: Thu, 6 Aug 2026 23:15:51 +0000 Subject: [PATCH 1/3] docs: review and decline three attached viral one-pagers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Rate Continuous Skill Acquisition, the Hoffman/Anthropic Karpathy-loop playbook, and the ChatGPT Andrew Ng compilation for Recertia applicability. None clear the bibliography bar (measurements or scrutinised practitioner argument); record scores and decline rationale in references.md §9 so the false Anthropic/recertia provenance claim is not re-imported later. Co-authored-by: Rob Schmidt --- docs/references.md | 39 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) diff --git a/docs/references.md b/docs/references.md index 99db9e5..521f7cd 100644 --- a/docs/references.md +++ b/docs/references.md @@ -272,3 +272,42 @@ cannot (refactor-plan B7). Kept here only as a pointer: - SkillsBench's null result → [`assumptions.md#a1`](assumptions.md#a1-machine-checkable-domains-show-a-positive-causal-lift-from-skill-retrieval). - Ratchet's evidence floor at our throughput → [`assumptions.md#a2`](assumptions.md#a2-ratchets-evidence-floor-is-reachable-at-our-traffic-volume). - No external precedent for the ADR-0005 self-modification boundary → [`assumptions.md#a3`](assumptions.md#a3-a-tiered-self-modification-boundary-is-sufficient-without-an-externally-reported-precedent). + +## 9. Attached one-pagers reviewed and declined (2026-08) + +Three short IEEE-styled one-pagers were reviewed for bibliography inclusion. None met the +threshold used elsewhere in this file — a design-shaping claim with measurements or a +practitioner argument that survived independent scrutiny — so they are **not** cited in §§1–5. +Scores below are 0–10 for topical *relevance* to Recertia and for *importance* as evidence +(quality, provenance, and whether anything here should change). Cite only if both are high. + +| One-pager | Relevance | Importance | Verdict | +| --- | --- | --- | --- | +| **Continuous Skill Acquisition** (anonymous poster: discover → filter → evaluate → MCP package → publish from the public OSS ecosystem) | 6 | 2 | Decline | +| **Graph Engineering: The Karpathy Loop, Improved 1000x… The Anthropic Playbook** (attrib. “David J.A. Hoffman”, “consultant in residence at Anthropic/recertia”) | 4 | 0 | Decline | +| **My 4 Steps — From Loop Engineering to Graph Engineering** (attrib. Andrew Ng; body states compiled by ChatGPT, July 2024, not authored by Ng) | 5 | 1 | Decline | + +**Continuous Skill Acquisition.** The pipeline rhyme with the Miner → validate → promote path +is real, and the insistence on static analysis / tests before install is compatible with our +gates. It is not design-shaping: there is no fetchable primary with measurements; MCP is a +connectivity interface, not our skill contract; and open-ecosystem crawl-and-install is the +wrong acquisition surface for us. SkillsBench (§1.1) is why the Miner prefers +`mined_from_human_artifact` in the owner’s repo over uncurated public skills. Adjacent real +work already covers the theme better (Dynamic Agent Skills survey §1.5; Voyager / MUSE in +§2). No change. + +**“Anthropic Playbook” / Hoffman.** Do not cite. Provenance fails: no verifiable Anthropic +affiliation for the named author, fabricated `recertia.{ai,com,org}` contact addresses and an +“Anthropic/recertia” residency claim that is not a Recertia affiliation, and a “1000×” +headline with no benchmark. Independent checks of the mid-2026 “graph engineering” PDF wave +found the same pattern — short compilations marketed under Anthropic/Karpathy branding while +disclaiming that affiliation on the cover. Recertia’s graph-execution grounding remains the +Kopadze practitioner note in §1.7 (fake edges, verifier isolation, hidden edges, silent +merges), not this artifact. No change. + +**“Andrew Ng / 4 Steps” ChatGPT compilation.** Reflection, tool use, planning, and +multi-agent collaboration are patterns we already use, but the attached PDF is an +unendorsed AI compilation and should not stand in for Ng. The primary academic anchors for +those patterns are already in §§3–4 (Reflexion, ReAct, Self-Refine). Ng’s own public +writing on agentic workflows is practitioner background at best; this PDF is neither that +nor a measurement. No change. From b0e1e799a605e8ca2f43abec6e3324d959da3ce1 Mon Sep 17 00:00:00 2001 From: Cursor Agent Date: Thu, 6 Aug 2026 23:22:49 +0000 Subject: [PATCH 2/3] docs: mark two attached compilations as fetched and read MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pull and OCR the Graph Engineering Drive PDF and the Andrew Ng one-pager; correct §9 provenance (independent July 2026 compilations, not Hoffman / recertia authorship). Continuous Skill Acquisition still has no public primary — leave unfetched and declined. Co-authored-by: Rob Schmidt --- docs/references.md | 82 ++++++++++++++++++++++++++-------------------- 1 file changed, 46 insertions(+), 36 deletions(-) diff --git a/docs/references.md b/docs/references.md index 521f7cd..1520f83 100644 --- a/docs/references.md +++ b/docs/references.md @@ -275,39 +275,49 @@ cannot (refactor-plan B7). Kept here only as a pointer: ## 9. Attached one-pagers reviewed and declined (2026-08) -Three short IEEE-styled one-pagers were reviewed for bibliography inclusion. None met the -threshold used elsewhere in this file — a design-shaping claim with measurements or a -practitioner argument that survived independent scrutiny — so they are **not** cited in §§1–5. -Scores below are 0–10 for topical *relevance* to Recertia and for *importance* as evidence -(quality, provenance, and whether anything here should change). Cite only if both are high. - -| One-pager | Relevance | Importance | Verdict | -| --- | --- | --- | --- | -| **Continuous Skill Acquisition** (anonymous poster: discover → filter → evaluate → MCP package → publish from the public OSS ecosystem) | 6 | 2 | Decline | -| **Graph Engineering: The Karpathy Loop, Improved 1000x… The Anthropic Playbook** (attrib. “David J.A. Hoffman”, “consultant in residence at Anthropic/recertia”) | 4 | 0 | Decline | -| **My 4 Steps — From Loop Engineering to Graph Engineering** (attrib. Andrew Ng; body states compiled by ChatGPT, July 2024, not authored by Ng) | 5 | 1 | Decline | - -**Continuous Skill Acquisition.** The pipeline rhyme with the Miner → validate → promote path -is real, and the insistence on static analysis / tests before install is compatible with our -gates. It is not design-shaping: there is no fetchable primary with measurements; MCP is a -connectivity interface, not our skill contract; and open-ecosystem crawl-and-install is the -wrong acquisition surface for us. SkillsBench (§1.1) is why the Miner prefers -`mined_from_human_artifact` in the owner’s repo over uncurated public skills. Adjacent real -work already covers the theme better (Dynamic Agent Skills survey §1.5; Voyager / MUSE in -§2). No change. - -**“Anthropic Playbook” / Hoffman.** Do not cite. Provenance fails: no verifiable Anthropic -affiliation for the named author, fabricated `recertia.{ai,com,org}` contact addresses and an -“Anthropic/recertia” residency claim that is not a Recertia affiliation, and a “1000×” -headline with no benchmark. Independent checks of the mid-2026 “graph engineering” PDF wave -found the same pattern — short compilations marketed under Anthropic/Karpathy branding while -disclaiming that affiliation on the cover. Recertia’s graph-execution grounding remains the -Kopadze practitioner note in §1.7 (fake edges, verifier isolation, hidden edges, silent -merges), not this artifact. No change. - -**“Andrew Ng / 4 Steps” ChatGPT compilation.** Reflection, tool use, planning, and -multi-agent collaboration are patterns we already use, but the attached PDF is an -unendorsed AI compilation and should not stand in for Ng. The primary academic anchors for -those patterns are already in §§3–4 (Reflexion, ReAct, Self-Refine). Ng’s own public -writing on agentic workflows is practitioner background at best; this PDF is neither that -nor a measurement. No change. +Three IEEE-styled posters/compilations were reviewed for bibliography inclusion. Two were +subsequently **fetched and read in full** (Google Drive 11-page PDF; LinkedIn one-pager +image + OCR). The third has **no public primary** under its distinctive phrases, so it was +assessed only from the chat attachment. None met the threshold used elsewhere in this file — +a design-shaping claim with measurements or a practitioner argument that survived independent +scrutiny — so they are **not** cited in §§1–5. Scores below are 0–10 for topical *relevance* +to Recertia and for *importance* as evidence. Cite only if both are high. Verification: +**[F, practitioner]** for the two fetched compilations (read directly; not academic +primaries); **[unfetched]** for Continuous Skill Acquisition. + +| Artifact | Fetched? | Relevance | Importance | Verdict | +| --- | --- | --- | --- | --- | +| **Continuous Skill Acquisition** (poster: discover → filter → evaluate → MCP package → publish from public OSS) | No public copy found | 6 | 2 | Decline | +| **Graph Engineering: The Karpathy Loop, Improved 1000x… The Anthropic Playbook** (11 pp.; cover: independently compiled July 2026, not affiliated with Karpathy/Anthropic) **[F, practitioner]** | Yes — [Drive PDF](https://drive.google.com/file/d/1-GOg0kxcp8tx1BMUECMj2yJq6JYGmfhb/view) | 5 | 2 | Decline | +| **My 4 Steps — From Loop Engineering to Graph Engineering** (one-pager; cover: independently compiled July 2026, not affiliated with Andrew Ng) **[F, practitioner]** | Yes — LinkedIn feed image + OCR | 5 | 1 | Decline | + +**Continuous Skill Acquisition [unfetched].** Distinctive phrases from the attachment +(“Agent tools are the open library for autonomous history”, Discovery/Evaluation/Deployment +layers with MCP skill packaging) return no public PDF, arXiv, or Drive hit. The pipeline rhyme +with Miner → validate → promote is real, and pre-install testing is compatible with our gates, +but without a fetchable primary there is nothing to cite. Even on the attachment’s own terms: +MCP is connectivity, not our skill contract; open-ecosystem crawl-and-install is the wrong +acquisition surface given SkillsBench (§1.1). Adjacent real work already covers the theme +(Dynamic Agent Skills §1.5; Voyager / MUSE §2). No change. Re-upload a PDF if a primary +exists. + +**Graph Engineering / “Anthropic Playbook” [F, practitioner].** Read cover-to-appendix. +Cover and acknowledgment both state it is an **independent synthesis**, not affiliated with +or endorsed by Karpathy or Anthropic. Content maps public sources: Karpathy `autoresearch` +(≈630 LOC, ~700 experiments / ~20 retained, `program.md` + `val_bpb` ratchet), AgentHub +(commit DAG as collaboration graph), Anthropic “Building Effective Agents”, Dynamic +Workflows (fresh-context parallel sub-agents), and the Knowledge Graph Construction Cookbook +(typed entities, provenance, multi-hop query). Useful engineering claims that survive the +provenance correction — measured loop before swarm, typed shared state over transcripts, +ground evaluators in edges not vibes, persist graph across sessions, keep human ownership of +objectives/ontology — largely restate what Recertia already took from Kopadze §1.7 and +ADR-0001 / plural memory. The “1000×” headline has no accompanying benchmark. Do not cite +this compilation; cite the primary sources or §1.7. No change. + +**“Andrew Ng / 4 Steps” [F, practitioner].** OCR of the circulating one-pager: “Independently +compiled, July 2026 — not affiliated with Andrew Ng and not endorsed.” It restates Ng’s four +patterns (reflection, tool use, planning, multi-agent) and the HumanEval anecdote (GPT-3.5 + +agentic workflow 95.1% vs GPT-4 zero-shot 67.0%) from his public Batch writing, then overlays +a loop→chain→network→graph ladder that Ng did not publish as an ordered method. Patterns we +already use are anchored in §§3–4 (Reflexion, ReAct, Self-Refine). Do not cite this +compilation in place of Ng or those primaries. No change. From aa24e094755f3c65fe093c76a8065b295aff841e Mon Sep 17 00:00:00 2001 From: Cursor Agent Date: Thu, 6 Aug 2026 23:25:19 +0000 Subject: [PATCH 3/3] docs: re-read Continuous Skill Acquisition from re-attached poster MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Update §9 with the eight-step Discovery/Extraction/Deployment pipeline from the re-attached CSA poster; still no public primary and still declined. Note Graph Engineering chat variants vs the fetched Drive PDF. Co-authored-by: Rob Schmidt --- docs/references.md | 82 +++++++++++++++++++++------------------------- 1 file changed, 37 insertions(+), 45 deletions(-) diff --git a/docs/references.md b/docs/references.md index 1520f83..d431108 100644 --- a/docs/references.md +++ b/docs/references.md @@ -275,49 +275,41 @@ cannot (refactor-plan B7). Kept here only as a pointer: ## 9. Attached one-pagers reviewed and declined (2026-08) -Three IEEE-styled posters/compilations were reviewed for bibliography inclusion. Two were -subsequently **fetched and read in full** (Google Drive 11-page PDF; LinkedIn one-pager -image + OCR). The third has **no public primary** under its distinctive phrases, so it was -assessed only from the chat attachment. None met the threshold used elsewhere in this file — -a design-shaping claim with measurements or a practitioner argument that survived independent -scrutiny — so they are **not** cited in §§1–5. Scores below are 0–10 for topical *relevance* -to Recertia and for *importance* as evidence. Cite only if both are high. Verification: -**[F, practitioner]** for the two fetched compilations (read directly; not academic -primaries); **[unfetched]** for Continuous Skill Acquisition. - -| Artifact | Fetched? | Relevance | Importance | Verdict | +Three IEEE-styled posters/compilations were reviewed for bibliography inclusion. None met the +threshold used elsewhere in this file — a design-shaping claim with measurements or a +practitioner argument that survived independent scrutiny — so they are **not** cited in +§§1–5. Scores are 0–10 for topical *relevance* and for *importance* as evidence. Cite only if +both are high. + +| Artifact | How read | Relevance | Importance | Verdict | | --- | --- | --- | --- | --- | -| **Continuous Skill Acquisition** (poster: discover → filter → evaluate → MCP package → publish from public OSS) | No public copy found | 6 | 2 | Decline | -| **Graph Engineering: The Karpathy Loop, Improved 1000x… The Anthropic Playbook** (11 pp.; cover: independently compiled July 2026, not affiliated with Karpathy/Anthropic) **[F, practitioner]** | Yes — [Drive PDF](https://drive.google.com/file/d/1-GOg0kxcp8tx1BMUECMj2yJq6JYGmfhb/view) | 5 | 2 | Decline | -| **My 4 Steps — From Loop Engineering to Graph Engineering** (one-pager; cover: independently compiled July 2026, not affiliated with Andrew Ng) **[F, practitioner]** | Yes — LinkedIn feed image + OCR | 5 | 1 | Decline | - -**Continuous Skill Acquisition [unfetched].** Distinctive phrases from the attachment -(“Agent tools are the open library for autonomous history”, Discovery/Evaluation/Deployment -layers with MCP skill packaging) return no public PDF, arXiv, or Drive hit. The pipeline rhyme -with Miner → validate → promote is real, and pre-install testing is compatible with our gates, -but without a fetchable primary there is nothing to cite. Even on the attachment’s own terms: -MCP is connectivity, not our skill contract; open-ecosystem crawl-and-install is the wrong -acquisition surface given SkillsBench (§1.1). Adjacent real work already covers the theme -(Dynamic Agent Skills §1.5; Voyager / MUSE §2). No change. Re-upload a PDF if a primary -exists. - -**Graph Engineering / “Anthropic Playbook” [F, practitioner].** Read cover-to-appendix. -Cover and acknowledgment both state it is an **independent synthesis**, not affiliated with -or endorsed by Karpathy or Anthropic. Content maps public sources: Karpathy `autoresearch` -(≈630 LOC, ~700 experiments / ~20 retained, `program.md` + `val_bpb` ratchet), AgentHub -(commit DAG as collaboration graph), Anthropic “Building Effective Agents”, Dynamic -Workflows (fresh-context parallel sub-agents), and the Knowledge Graph Construction Cookbook -(typed entities, provenance, multi-hop query). Useful engineering claims that survive the -provenance correction — measured loop before swarm, typed shared state over transcripts, -ground evaluators in edges not vibes, persist graph across sessions, keep human ownership of -objectives/ontology — largely restate what Recertia already took from Kopadze §1.7 and -ADR-0001 / plural memory. The “1000×” headline has no accompanying benchmark. Do not cite -this compilation; cite the primary sources or §1.7. No change. - -**“Andrew Ng / 4 Steps” [F, practitioner].** OCR of the circulating one-pager: “Independently -compiled, July 2026 — not affiliated with Andrew Ng and not endorsed.” It restates Ng’s four -patterns (reflection, tool use, planning, multi-agent) and the HumanEval anecdote (GPT-3.5 + -agentic workflow 95.1% vs GPT-4 zero-shot 67.0%) from his public Batch writing, then overlays -a loop→chain→network→graph ladder that Ng did not publish as an ordered method. Patterns we -already use are anchored in §§3–4 (Reflexion, ReAct, Self-Refine). Do not cite this -compilation in place of Ng or those primaries. No change. +| **Continuous Skill Acquisition** (poster: Scan→Filter→Curate→Extract→Evaluate→Generate→Store→Publish from public OSS) | Chat attachment (twice); **no public PDF/Drive/arXiv** under distinctive phrases | 6 | 2 | Decline | +| **Graph Engineering: The Karpathy Loop, Improved 1000x… The Anthropic Playbook** | Chat attachment + full [Drive PDF](https://drive.google.com/file/d/1-GOg0kxcp8tx1BMUECMj2yJq6JYGmfhb/view) (11 pp.) + OCR **[F, practitioner]** | 5 | 2 | Decline | +| **My 4 Steps — From Loop Engineering to Graph Engineering** | Chat attachment + LinkedIn one-pager OCR **[F, practitioner]** | 5 | 1 | Decline | + +**Continuous Skill Acquisition.** Re-read from the re-attached poster. Architecture is three +layers / eight steps: Discovery (Scan GitHub trending & tech reports → Filter duplicates / +malware → Curate by risk/utility), Extraction (Workflow Extraction → Skill Evaluation via +lint/tests → Skill Generation as YAML/JSON), Deployment (Store in org library → Publish to +agents). Abstract claims agents must discover workflows across public repos; index terms +include “remote sensing” (domain-mismatched tell). Still **not citable**: no fetchable +primary, no measurements, and open-ecosystem crawl-and-install contradicts SkillsBench (§1.1) +— Miner prefers `mined_from_human_artifact` in the owner’s repo. Pipeline rhyme with Miner → +validate → promote / Curator is already covered by Dynamic Agent Skills (§1.5) and Voyager / +MUSE (§2). No change. A durable PDF upload would not change the importance score without +measurements. + +**Graph Engineering / “Anthropic Playbook” [F, practitioner].** The circulating Drive PDF’s +cover and acknowledgment state **independent synthesis, July 2026 — not affiliated with +Karpathy or Anthropic**. Content maps public sources (`autoresearch` / AgentHub; Anthropic +workflows + KG Cookbook). Useful surviving claims (measured loop before swarm, typed shared +state, edge-grounded evaluators, persistence, human ownership of objectives) restate Kopadze +§1.7 and ADR-0001. The “1000×” headline has no benchmark in the PDF. Chat re-attachments +sometimes present a different surface branding (Anthropic mark, “2024” header, role-swarm +diagram); treat those as the same viral compilation family, not as Anthropic authorship. Do +not cite; cite primaries or §1.7. No change. + +**“Andrew Ng / 4 Steps” [F, practitioner].** OCR: “Independently compiled, July 2026 — not +affiliated with Andrew Ng and not endorsed.” Restates Ng’s four patterns and the Batch +HumanEval anecdote; the loop→graph ladder is the compiler’s overlay. Patterns already +anchored in §§3–4 (Reflexion, ReAct, Self-Refine). Do not cite this compilation. No change.