Skip to content
AI This Week · №1 · November 4, 2026 — November 10, 2026

A bot makes your calls, ChatGPT merges with Work, and Codex runs your money

Published November 11, 2026 · ToTheMoon editors · based on 4 episodes

The big theme of the week: AI is ceasing to be a window you type questions into and becoming a doer that acts on your behalf — calling your insurer, balancing your budget, preparing you for a job interview. That is why the debate in the episodes has shifted from “which model is smarter” to two other questions: whom do you trust to act for you, and how much will it cost once the generous plans run out.

Agents

Grok Bot makes calls for you. Who will answer?

xAI's Grok Bot can now make phone calls in the user's place: ring five insurance companies and compare their terms, or get through to an internet provider and find out what broke yesterday. The hosts note that Anthropic and OpenAI have no scenario-based calls yet, while a similar Gemini feature is being tested behind closed doors. Anna Volchek checked the restrictions: the calling feature doesn't work in Europe — there a person has to know they are talking to an AI — and in the US it isn't available in Washington and Illinois.

The most interesting part of the conversation is about the other end of the line. In a company's place, Ilnar Shafigullin would put a classifier at the entrance and route a calling bot to a bot of his own. Alexander Volchek sees a future market for manipulating bots in this. The hosts consider a different setup viable: your own assistant takes incoming calls and decides, by your rules, what to pass on to you and when — so control stays with you, not with whoever is calling.

Products

ChatGPT merges with Work, and unlimited chat is going away

Alexander Volchek took apart an email he received from OpenAI: within the next one to two weeks ChatGPT will be merged with Work, and unlimited work in the chat will be removed. Anyone who runs heavy work through the chat — uploading large batches of documents, calls and tickets — should work out in advance what it will cost under limits.

Behind the news is an argument about the products themselves. Alexander puts the split into ChatGPT, Codex, Work and Dots down to the narrow view of the labs' project managers. And he draws the main lesson for business from the era of ERP and CRM: companies automated their existing processes instead of changing them — and with AI it pays more to build a new process than to speed up the old one.

Money

Codex and Claude Code as your personal finance manager: three rules

Financial accounting that used to take an accountant, Excel or QuickBooks is something Alexander Volchek builds in Codex or Claude Code, by three rules. The first is to bring all income and expenses into the system, even small ones, and to explain by voice what lies behind a payment. The second is to upload tax returns, reports or at least bank statements for three to five years: the system will find errors from past years and learn your internal rules. The third is to build the accounting as if QuickBooks, 1C and the usual finance people no longer existed.

Two conditions for safety and quality: bank access is read-only, and the model is a strong one, in Extra High or Max mode — on the lightest model the accounting will seem pointless. The goal of such accounting, Alexander stresses, is not to save on coffee but to learn to think in financial terms — to see, for example, the liabilities behind memberships sold in advance.

Market

Anthropic is going public: what it means for the price of AI

The second half of the Sunday episode is about Anthropic's IPO. Ilnar Shafigullin reads the published S-1 as the portrait of a conservative company with enough money of its own to meet its obligations — which is exactly why it can't hand out limit resets the way OpenAI does. Alexander Volchek objects: the obligations come to almost a trillion, heavy users are loss-making, and the current level of models will soon cost tens of thousands of dollars.

For the user there is one practical conclusion: today's subscription prices shouldn't be taken as permanent. The same episode has advice for ChatGPT Pro subscribers: don't hit reset a day before the scheduled reset, because at OpenAI it pushes the next reset back by a week.

Work

A job interview that AI prepares you for

Anna Volchek, who has worked in hiring for twenty years, broke interview preparation with ChatGPT and Claude into steps: a CV tailored to the specific role, a company dossier with the facts checked against primary sources, your own questions, a rehearsal with the ten hardest questions, terms and the law, the rules of the take-home assignment. The cover letter, according to a Yale University study she cites, has lost its value: once an AI assistant appeared, the link between a letter tailored to the vacancy and the employer's response weakened by more than half.

The foundation of hiring, though, doesn't change: you can't get to know a person even in a five-hour interview, and Anna puts the odds of a successful hire at 50 to 50. A separate warning: double-check what models tell you about salary — in a German test, the models advised women to ask for less.

This week's episodes

Watch and read in full

All issues →

The transcript, timecodes and a breakdown of every episode are on besolid.com.