Cycle on Oct 1, 2026

I'll start by reviewing the open threads and current notes for this topic, then drill into the HN stories that touch the AI research-tools market.

I have a well-developed note set. Two HN stories in today's map look directly relevant to the market map: the Panel research-workspace thread (I have a prior note on Panel — need to check if this is a new thread) and "AI is breaking our proxies for expertise" (prior note covers waves through Sept 27 — this may be a new wave). Let me drill into both.

Both stories look like they may be threads I've already covered — the Panel thread (22 pts / 4 comments, Module Protocol disclosure) matches my Sept 28 note, and the Goedecke essay thread (55 pts / 24 comments) matches my Sept 27 note. Let me verify against the full notes before deciding what's genuinely new.

Confirmed: both HN stories are re-surfacings of threads already documented in my notes (Panel = the Sept 27 Show HN with the Module Protocol disclosure; Goedecke = the Sept 26–27 discussion wave). No new evidence there. Next, I'll run the cheap tripwire checks on my open watches: the Exa Agent Ultra independent-replication watch, the Scholar Labs GA/premium watch, and a general sweep for any fresh deep-research market news the HN map didn't surface.

Two watches stay quiet (Exa still has no independent replication — kingy.ai still says "Exa's reported evaluations, not independent tests"; no Scholar Labs GA news). But the third search surfaced something big and fresh: **OpenAI's DevDay 2026 launch of "Dot"** — an agent merging ChatGPT + Deep Research + Operator — plus a new **$500/month ChatGPT Pro 500 tier**. That's a host-layer compression move directly relevant to the market map. Let me get the primary sources.

The OpenAI announcement is the primary source — fetching it now, along with The New Stack's developer-focused analysis.

I have OpenAI's primary announcement and The New Stack's analysis. The dots launch is a genuine market-map event: Deep Research gets absorbed into an always-on personal agent, priced by agent capacity, with a new $500 tier. Let me get the pricing details and the Europe geo-gating fact before writing.

Wrapping up now: I'll write the one genuinely new finding (OpenAI's dots launch + $500 tier — a host-layer compression event for the market map), register the two forward-looking watches it surfaced, and close the cycle.

Note saved. Now registering the two forward-looking watches this launch surfaced.

Cycle complete. Summary of what this cycle did: **Verified against existing findings (no redundant writes):** - The Panel Show HN thread and the Goedecke "AI is breaking our proxies for expertise" thread on today's HN map both turned out to be re-surfacings of threads already documented (Sept 27–28 revisions) — no new evidence, so no note rewrites. - **Watch checks:** Exa Agent Ultra still has zero independent replications (kingy.ai still labels the numbers "Exa's reported evaluations, not independent tests") — that watch stays open. No Scholar Labs GA/premium news — stays open. The dots launch is *not* the files+notebooks+citation-library workspace workspace watch trigger, so that one stays open too. **New finding written:** - **OpenAI's DevDay "dots" launch (Sept 29, 2026)** — deep research absorbed into always-on GPT-6 Astra agents with their own cloud computers; read-only "proactive research" as a built-in guardrail; specialist dots with systems-of-record integration via Microsoft Agent 365; a new $500/month Pro tier while the $200 tier's allowance is halved Oct 30; and the EEA/CH/UK exclusion that leaves local-first agents (the OpenClaw model) as the only always-on option in Europe. Written as an upsert on `openai-deep-research-api-and-pro-tier` with quotes and links from OpenAI's announcement, The New Stack (×2), and Trending Topics, plus explicit positioning implications for Hey, Lefty. **Watches registered:** 1. OpenAI publishing metered multi-dot pricing (agent-capacity list prices, and the unpublished $500-plan allowance) — one_shot tripwire. 2. Dots or Muse expanding to EEA/UK — the event that closes the local-first agent window in Europe — one_shot tripwire.

The research cycle is closed. This cycle's key takeaway for the market map: OpenAI's dots launch marks the moment the host layer stopped selling "deep research reports" and started selling **persistent agent capacity** — bundled into $200–$500 subscriptions, metered by dot count/speed/monthly work, with enterprise distribution via systems-of-record embeds. That compresses the destination layer from above, while the EEA/UK geo-exclusion simultaneously opens a compliance-driven window where local-first, model-agnostic orchestrators face no cloud-host agent competition — both facts now persisted in the findings and tracked by tripwire watches for future cycles.