[chaitanyagiri]

Day 30 evening run of the content engine: one new versus page and two fact refreshes from the refresh queue (the queue holds more than 5 items, so the run ships 2 fixes and 1 new page, ENGINE.md section 5b).

What changed

• New: pi-vs-claude-code (BACKLOG b072). Price, models, built in features and permissions, with a dated table and a pick list. Title card plus 3 scenes (box, plugs, gate), one tap interaction each. Demand: Search Console 13 to 26 Oct shows "pi agent" at 3,356 impressions, position 6.9, landing on what-is-pi-agent at 0.15% clicks; versus pages are the format the tag review favours.
• Fix: codex-cli-vs-claude-code (refresh queue: over the word band, FAQ tier fact). Plans, Plus limits per model (GPT-6.1 Terra added), Pro $100, $200 and $500, Max 5x and 20x, computer use and AGENTS.md rechecked against OpenAI's and Anthropic's pages on 27 Oct. The FAQ no longer says the tiers line up at 5x and 20x on each side. The old 29 Sep help screen run and the Pro 200 pause section were removed. 1,498 to 1,150 words. House line on the Munder Difflin mentions.
• Fix: how-to-use-claude-code-plan-mode (refresh queue: version fact, word band, click rate). States both sources for auto as the starting mode: the docs say v2.1.283 or later, and the changelog says 2.1.283 for third party providers and 2.1.284 for every plan and provider. New title and description, the 11 Sep test run section removed, and the best practices quote replaced because it is no longer on the live page. 1,685 to 1,159 words. Munder Difflin source facts reread on main.
• what-is-pi-agent: house line, plus a link to the new page.
• media.json: one entry, inserted after pi-agent-alternatives, so it does not collide with seo-engine: day 30 AM, brief 27 Oct and Claude Cowork alternatives #795's entries.
• docs/ rebuilt.

Checks

• gate.mjs: PASS on all four changed posts (pi-vs-claude-code 951 words; codex-cli-vs-claude-code 1,150; how-to-use-claude-code-plan-mode 1,159; what-is-pi-agent 1,180, length warning only).
• One independent fact review: FAIL on 7 lines (Pi page), 3 (Codex), 2 (plan mode); PASS on what-is-pi-agent. All fixed with the reviewer's wording or tighter. Main fixes: Claude Code now starts in auto, so it is not "asks before it acts"; Bedrock, Agent Platform and Foundry can map aliases to older Claude versions; annual prices; "not in the CLI" softened to what OpenAI documents.
• npm run build clean (424 files).
• OpenSEO credits used: 0.

Before

Page	Before (main, built)
/blog/pi-vs-claude-code/
/blog/codex-cli-vs-claude-code/
/blog/how-to-use-claude-code-plan-mode/
/blog/what-is-pi-agent/

After

Page	After (built locally)
/blog/pi-vs-claude-code/
/blog/codex-cli-vs-claude-code/
/blog/how-to-use-claude-code-plan-mode/
/blog/what-is-pi-agent/

Open engine PRs: #795 (day 30 AM) and this one. Both rebuild docs/, so whichever merges second may need a rebase. The founder merges.

🤖 Generated with Claude Code