Three public LinkedIn posts from the last two days, one theme: the personal agent - the kind that manages your messages, calendar and errands rather than your codebase - just became a real product category. Anthropic shipped one, a feedback company ranked nine of them, and the legal-AI corner of the market raised nine figures to prove the money is following. Short quotes, links and what I think. Nothing here comes from anyone's private feed.
Each item started as a public LinkedIn post; I could not open the posts themselves (LinkedIn requires a login), so every item below is verified against the public first-party source it points to: Anthropic's own Claude Duet announcement and product page, Enterpret's Feedback Bench site with its full published method and data, and Legora's own press release. Where a number is the company's own claim - ARR, post and user counts - the item says so, and one figure from the LinkedIn summary could not be confirmed on the public site. I read these pages on October 22, 2026.
"My read is that the personal agent wave could get much bigger than the coding agent wave, simply because it's accessible to a much larger audience, including nontechnical users. But security and trust are serious concerns."
The launch behind the post is real. Anthropic's announcement, published October 21, 2026, opens a research preview of Claude Duet, a standalone app on iOS, Android and claude.ai where Claude works as a proactive agent: a morning brief on what needs your attention across connected apps, monitoring through the day, and task execution that ranges from scheduling to building dashboards - with the stated design rule that it checks with you before acting on your behalf. It is built on Claude Opus 5.5, works through more than 6,000 connectors and plugins per the product page, and is waitlisted, starting with US Pro and Max subscribers under existing usage limits. Reganti's post is an opinion, and a reasonable one: the addressable audience for "manage my day" is everyone with a phone, not everyone with a repository. Her caveat is the load-bearing one. An agent that reads your messages and acts across 6,000 connectors concentrates exactly the permissions that make prompt injection and account compromise expensive, and "asks before it acts" is a design promise that users will test in production. Not verified: any independent evaluation of how often Duet's judgment calls are right - it is one day old.
"We analyzed 218,149 public posts from 132,844 users... Meta's Muse leads overall, with the highest customer love on 22 of 53 criteria."
The board behind the post is Feedback Bench 1.1, built by Enterpret - a feedback-analytics vendor, which discloses that itself and says no vendor pays to be listed - covering the window August 31 to October 18, 2026. It ranks nine personal agents on public posts from Reddit, X, YouTube and the app stores, labelling each against 53 criteria in 9 areas, with the unit of counting being the user-week so one loud person cannot dominate. The method, formulas and per-criterion data are all public as Markdown, JSON and CSV, which is how a ranking like this earns the right to be cited. The standings as published: Muse (Meta) first with a Feedback Score of 70.5, Hermes Agent (Nous Research) second at 59.6, then Grok Bot (xAI), Hark, OpenClaw, Instinct, Gemini Spark, ChatGPT Dots and Poke. Two precisions on the LinkedIn summary. The site's exact claim is that Muse sits clearly in the top half on 22 of 53 criteria (and clearly in the bottom half on 12) - "highest customer love on 22" is the post's shorthand, not the site's wording. And the 218,149 posts / 132,844 users totals appear in the LinkedIn post but not on the public ranking or method pages I checked, so treat them as Enterpret's own report of its corpus. The head-to-head claim does check out precisely: on the Hermes Agent page, of 331 posts that judge both Hermes and OpenClaw, Hermes comes out ahead in 160 and OpenClaw in 33 - that is 82.9% of the posts that pick a side, the "83%" in the summary. One more honest note: automatic labelling does the scoring, X volume is capped per feed, and the method page itself says popularity counts who posts, not who uses - so read this as a ranking of public sentiment, properly done, not of product quality.
The funding item, and the numbers are confirmed by Legora's own press release of October 21, 2026: a $550 million Series E at an $8.8 billion post-money valuation, led by Atreides Management, with new investors including Baron Capital, Capital Group, Fidelity Management & Research Company, Glade Brook Capital Partners and J.P. Morgan, and every major existing investor - Accel, Benchmark, Bessemer, General Catalyst, ICONIQ, Insight, Menlo, Redpoint, Salesforce Ventures, Sands Capital - participating. The company-sourced growth figures in the release: ARR above $200 million, having gone from $1 million to $100 million in under 18 months, with the previous round valuing it at $5.6 billion; the US is now its largest market by revenue, with more than half of the AmLaw 50 as customers. The capital funds US and international expansion plus further investment in the Legora aOS, its "agentic operating system" for legal work. The context that makes this more than a money story: it lands the same week Anthropic ships a general personal agent and a benchmark formally ranks the category - the "agents that do multi-step professional work" thesis is getting funded, shipped and measured simultaneously. The standard caveats apply: ARR and customer counts are the company's own, valuations are set by the round's participants, and engagement counts on the LinkedIn post itself were not visible to me.