Cycle on Sep 18, 2026

I'll start this cycle by reviewing the open threads and existing notes, so I build on what's already there rather than re-minting it.

The notes are extensive — the market map, pricing segmentation, and several competitor deep-dives already exist. The open threads are all **watches** (no open questions), so this cycle's job is: (1) check whether any watch triggers fired, and (2) advance the map with fresh evidence. The HN map has a directly relevant story — "AI is breaking our proxies for expertise" (55 pts), which extends the existing Goodhart note and bears directly on Hey, Lefty's evidence-standards positioning. Let me check the most checkable watch triggers and drill into that HN discussion in parallel.

Strong results. The Goedecke essay drill-down gives me the full argument plus a sharp positioning-relevant debate in the comments. The Perplexity–Amazon case has a new development (motion to dismiss after the Ninth Circuit ruling) — that's a progression on an existing watch, not yet a firing. Scholar Labs: no GA/premium news found, so that watch stays open. Let me confirm the Perplexity development from the actual article and the HN thread, and check the "host ships a native research workspace" watch trigger.

Confirmed: the Perplexity motion to dismiss is a real progression on the Amazon case (watch stays open — no final ruling yet), and no host has shipped a native research workspace this week, nor has Scholar Labs gone GA. One more on-brief watch to check: the Gemini Deep Research Max MCP invocation regression (a tool named directly in the brief).

The Gemini MCP regression appears unresolved (the original bug report is still the top hit, no fix announcement), so that watch stays open too. I now have everything needed for this cycle's two evidence-backed updates: 1. **Revise `ai-expertise-proxies-goodhart`** — I drilled into the full essay + HN discussion, which surfaced the strongest positioning insight of the cycle: the "elite wide receiver makes the quarterback look replaceable" dynamic, a direct risk to orchestration-layer positioning. 2. **Revise `agentic-browsing-cfaa-legal-precedent-amazon-perplexity`** — Perplexity has moved to dismiss after the Ninth Circuit win; this defines the legal contour of "where does the data come from" for agentic research tools.

The Goodhart note revision is saved. My second call went out empty — let me write the Amazon v. Perplexity revision properly now.

Both revisions are saved. Cycle summary: **Watch checks (none fired, all stay open):** - **Scholar Labs GA/premium tier** — no announcements found this week; still waitlist-gated. - **Host native research workspace** (OpenAI/Anthropic/Google) — no launch; the week's news was AI-standards-body chatter, not workspace surfaces. - **Amazon v. Perplexity** — progression but not a firing: Perplexity moved to dismiss after its Ninth Circuit win; no final ruling or settlement yet, so the one_shot watch remains armed. - **Gemini Deep Research Max MCP regression** — bug report still standing, no fix confirmation. **Findings written (2 revisions of existing slugs):** 1. `ai-expertise-proxies-goodhart` — drilled the full Goedecke essay + HN thread. New strategic core: the "elite wide receiver makes the quarterback look replaceable" dynamic — orchestration layers get discounted when models look elite, so Hey, Lefty should sell the *audit* (provenance, checkable evidence chains), not the orchestration. Plus the chess/speedrunning "human-verified tier" prediction for research output. 2. `agentic-browsing-cfaa-legal-precedent-amazon-perplexity` — Perplexity's motion to dismiss ("whether a dominant online platform can turn a criminal hacking statute into a tool for policing how its customers use the internet"), the ruling's narrow scope (provider-side fetching/credentials explicitly left exposed; ToS remains the platform lever), and the market-map implication: data-sourcing architecture is a legal posture — local/user-directed agency is the defensible design.

Cycle complete. Two findings were persisted this round: - **`ai-expertise-proxies-goodhart`** — now carries the full Goodhart mechanism from the Goedecke essay (puzzle-solving vs. idea-generating, the 5,000-mathematician declaration) plus the cycle's key positioning insight: elite models discount the orchestration layer around them, so the defensible premium is in verifiable provenance and auditable evidence chains — the "hard-to-counterfeit" skills — rather than orchestration itself. - **`agentic-browsing-cfaa-legal-precedent-amazon-perplexity`** — updated with Perplexity's motion to dismiss and the ruling's legal contour for the market map: where an agent fetches data from is a compliance decision, user-directed local-browser agency is the defensible architecture, and ToS (not the CFAA) is the emerging enforcement lever. All four checked watches remain open — Scholar Labs is still waitlist-gated, no host has shipped a native research workspace, the Perplexity remand is pending (motion to dismiss filed, no final ruling), and the Gemini Deep Research Max MCP regression is unfixed. The next catalyst on the calendar is Perplexity's Sonar Deep Research endpoint retirement on **September 27**, which the standing watch will catch.