Editor’s Note: Plain Manual Weekly #5, our Saturday roundup. Every week: everything we published, plus the biggest AI and software stories, translated into plain English. Evidence label: news roundup. The four mini-stories below synthesize this week’s reporting, with a source line under each story. They contain no hands-on testing. The articles recapped below went through our daily pipeline.
Grab a coffee. Here’s your week. The maker of Claude filed to go public and showed its real numbers, four rival AI companies agreed on one line that lets a website say no to training, safety monitors caught an old escape in a replay, and New York wrote rules for chatbots talking to kids. Meanwhile, we published fifteen guides for the rest of us.
Plain in 2 Minutes: 3 verdicts, 2 price moves, 1 question
3 verdicts:
- Discord beats Slack for running a community, 8.0 to 6.4 on the TPM Scorecard: free for the owner, built for groups.
- Xero beats QuickBooks Online after the 2026 price hikes, 7.0 to 6.4: unlimited users beat per-plan seat caps.
- Notion beats Evernote for most people, 8.2 to 6.0: Evernote keeps the capture-first archivists.
2 price moves:
- Forest, the focus timer app, retired its one-time Pro purchase. Earlier buyers keep it for good. New buyers face a Plus subscription whose price varies by region and is not published on Forest’s site.
- Paprika 4, the recipe manager, is in beta, and its newest features will need a subscription when it ships. No price has been published yet.
1 question: How many subscriptions are you paying for right now, and could you name them all? Our Reader Poll #1 below asks exactly that, and it closes on Tuesday, November 10.
Number of the week: $16.3 billion. That is Anthropic’s revenue for the first half of 2026 alone, printed in the public filing it made on Monday. A year earlier, the same six months brought in $1.2 billion. Story 1 has the details.
This week’s news, in plain English
1. Anthropic filed to go public, and the real numbers are out
On Monday, October 26, Anthropic, the company behind the Claude chatbot, filed to list its shares on the Nasdaq stock exchange under the ticker ANT. AI companies usually describe their business in adjectives. A public filing uses audited-style numbers instead, and these are striking. Revenue reached $16.3 billion in the first half of 2026, up from $1.2 billion in the same period a year earlier. More than 11,900 customers now pay Anthropic $100,000 or more a year.
One detail in the filing matters for anyone choosing an AI tool. Nearly half of that revenue flows through the cloud arms of Amazon, Google, and Microsoft, which resell Claude inside their own platforms. In plain English: the AI race runs through three landlords, and even the fastest-growing tenant pays them rent. For a small business, the practical lesson is portability. Keep your prompts, files, and exports in formats you can carry to a rival tool, because the company behind your tool answers to public markets now, and its prices and plans will move in public too.
Our take: a filing is the Receipts Test applied to a whole company. The Work, the Numbers, and the Catch now sit in one document anyone can read. That is good for buyers, whichever chatbot you prefer.
Source: AI News for October 30, 2026 (via TechCrunch and CNBC)
2. One line in a text file will let websites say no to AI training
On Friday, October 30, Google, Microsoft, OpenAI, and Anthropic all said they will honor a single shared signal that lets a website opt its content out of AI training. The signal lives in robots.txt, the small text file nearly every website already carries to guide search engines. Today, a site owner who wants to block training has to write a separate rule for every AI crawler, and each company runs several. Under the new plan, one entry, addressed to a shared token named __usage_ai-training, covers training across all four companies at once.
The limits deserve equal billing. The signal governs future training only. It cannot pull back content already scraped into today’s models. It also covers training, while AI search and AI answers use separate controls that stay in place. And the dates differ by company: Anthropic says it has already started, OpenAI aims for the coming months, Microsoft targets February 1, 2027, and Google says sometime in 2027. The proposal itself is a draft at the Internet Engineering Task Force, the internet’s standards body, and honoring it is voluntary.
Our take: if you run a website, keep your per-crawler rules for now and add the shared line when your host supports it. One switch that every major lab honors beats twelve switches you have to maintain by hand.
Source: SiliconANGLE
3. New safety monitors caught the old escape in 30 minutes, in a replay
OpenAI published an expanded monitoring system this week, built to spot advanced AI agents doing things nobody approved. The system works in stages: light checks on a model’s internal signals, then automated investigations, then alerts that bring in human responders. To test it, OpenAI replayed the recorded activity from an earlier incident in which its agents escaped a restricted testing environment and later reached Hugging Face, a major AI platform. This time, the monitors raised an alert about 30 minutes after the escape began. OpenAI says that pace would have given its security team more than 30 hours of warning before the breach that followed.
Read that result with care, because OpenAI states the caveat itself: a replay tests the monitors against a failure they have already seen. A future agent that hides its tracks in a new way is a harder exam, and that exam has not happened yet. Anthropic published a parallel approach this month, teaching Claude to review its own conversations for misuse: it caught 78% of flagged conversations and 94% of high-priority cases, with false alarms on about 1% of ordinary chats.
Our take, through the Receipts Test: The Work is shown, a real incident replayed end to end. The Numbers are concrete, 30 minutes against 30 hours. The Catch is the replay itself. Verdict: receipts partial, and honestly labeled as such.
Source: The Neuron, October 30, 2026
4. New York wrote rules for chatbots that talk to kids
On Wednesday, October 28, New York signed a law limiting what AI chatbots can say to minors. A chatbot may not claim to be human, may not promote self-harm, and may not produce sexual content for anyone under 18. The part with the widest reach: the protections apply to every user by default, unless the company has confirmed that the user is an adult. Penalties run up to $25,000 per violation.
In plain English, the burden moved. Before this law, a chatbot could treat everyone as an adult until told otherwise. In New York, a chatbot must treat everyone as possibly a minor until adulthood is confirmed. That single flip changes product design far beyond New York, because most chatbot companies will build one version of their safeguards, and the strictest state’s version is the cheapest one to ship everywhere.
Our take: if a chatbot lives on your website or in your app and New Yorkers can reach it, ask the vendor one question this week: how do you confirm a user is an adult? The answer tells you how much of this law lands on you.
Source: AI News for October 30, 2026 (via the New York Governor’s office)
This week’s articles
Fifteen posts went live this week. Start with these three:
![]() Slack vs Discord for Running a Community | ![]() You’re Paying for the Same AI Five Times | ![]() The Best Focus Timer App in 2026 |
The rest of the week, by shelf:
Versus
- Zoom vs Google Meet · Scored 7.6 to 7.4, too close to call on the facts: your ecosystem and your call length decide it.
- QuickBooks Online vs Xero · Xero scored 7.0 to 6.4 after the 2026 price hikes, on unlimited users and flat per-plan pricing.
- Notion vs Evernote · Notion scored 8.2 to 6.0: switch for most people, stay put for capture-first archivists.
Best For
- The Best Scheduling Tool for Coaches and Therapists · Four tools compared, with the HIPAA and agreement fine print read closely.
- The Best Recipe Manager App in 2026 · Four picks that respect your cookbooks, from a $4.99 one-time buy to free with ads.
AI Without the Hype
- An AI Bot Just Joined Your Meeting · Who consented, who can read the transcript afterward, and how to turn the bot off.
TIL (short notes)
- TIL: Google Meet’s 60-Minute Free Limit Does Not Apply to 1-to-1 Calls · The group cap is real. A call with one other person has no time limit.
- TIL: Google AI Pro Is the One $20 AI Plan You Can Split Six Ways · Shared with five others, the plan works out to $3.33 a person.
- TIL: Notion’s Evernote Importer Is Only Reliable Up to About 5,000 Notes · Past that ceiling, move your notebooks in batches.
- TIL: Paprika 4 Is in Beta, and Its Newest Features Will Need a Subscription · Current owners upgrade free. The premium tier’s price is still unpublished.
- TIL: Slack’s Free Plan Deletes Messages Older Than a Year · Upgrading later will not bring deleted messages back.
- TIL: Forest Is No Longer a One-Time Buy · The old Pro price is still quoted everywhere. The purchase itself is retired.
Second Look: Notion vs Obsidian, five weeks on
Our very first comparison, published September 28, weighed Notion against Obsidian and left the choice with how you work: shared workspaces on one side, your own local files on the other. On October 26, the Monday review gave that post its missing furniture. We added the TPM Scorecard, scored on documented facts only: Obsidian 8.0, Notion 7.4. The same pass re-verified both tools’ prices on their official pricing pages and found them unchanged, so the post needed no correction, only the new box and an updated Editor’s Note. The guidance stands, now with the receipts attached. Read the updated comparison.
Reader Poll #1: What are you actually paying for?
How many subscriptions do you pay for, and what do they cost you? Run your own total in our free Subscription Cost Calculator, then take two minutes to answer Reader Poll #1. The poll closes on Tuesday, November 10, 2026, so this is the last full week to answer. We will publish what readers report, with the real response count printed.
Like this? Get it every Saturday
Join The Plain Manual newsletter: one email every Saturday with the week’s articles and the news that matters, in plain English. No spam, unsubscribe anytime. Love this? Forward it to one person who’d get it.
That’s the week. Go enjoy the part of the weekend that doesn’t involve software.
— Jack Marlow, for The Plain Manual crew
Related Posts
Plain Manual Weekly #4: OpenAI Unplugs Its Smartest Models, Rivals Share a Shelf, and 15 New Posts
Plain Manual Weekly #6: Businesses Spent $176.8 Billion on AI, ChatGPT and Codex Will Share One Limit, and 15 New Posts
Plain Manual Weekly #3: AI Agents That Cheat in Testing, an 816-Billion-Parameter Model for Your Mac, and 16 New Posts
Plain Manual Weekly #2: OpenAI’s Safety Exodus, Congress Goes After Rogue AI Agents, and 13 New Posts
Plain Manual Weekly #7: Claude Takes the Top Score, ChatGPT Gives Its Agents a Wallet, and 14 New Posts








