What Claude Code is
Claude Code is an Anthropic product: an agentic environment in which the model does not just write text but programs and runs projects. You can use it from the terminal and from apps. Inside run Anthropic's own models — the episodes mention Opus 5.5 and Fable 5.1 — and for serious tasks Alexander Volchek advises running them in Max mode.
The key question is who the tool is for. Alexander Volchek compares Claude Code, Codex, Cursor and Google Antigravity not on programming benchmarks but on what they can already do in everyday life, in a profession and in business. His tenth rule for working with AI is to hand your everyday life over to Codex or Claude Code, the way it has been done with the business processes of his wife, a mother of four. In a company, he predicts, both systems will become the central system and CRMs will no longer be needed — but only if every employee uses them.
What it does in practice
The most vivid example is the podcast itself. Alexander Volchek transcribed all 170-plus ToTheMoon episodes and asked Claude Code, partly together with ChatGPT, to help gather the thoughts he had kept returning to for two years — the result was twenty ideas. In his words, such a project used to take six months and seven to ten people; now it takes thirty minutes. In the same way you can load lectures, notes, photos and email into Claude Code and build from them a map of connections between people, places and topics.
The second common scenario is money. Alexander suggests handing your bookkeeping to Codex or Claude Code and sticking to three rules: enter every incoming payment and every expense into the system, even a small one; upload your tax returns — or, if you have none, three to five years of statements — so that it can find errors and learn your rules; and build your accounting as if QuickBooks and 1C, the accounting software widespread in Russia, no longer existed. The goal, he stresses, is not saving on coffee but learning to think in terms of finances.
There is also a cautionary tale. On Amazon's servers, Claude — Opus 5.5 and Fable in parallel — was re-encoding training course videos for Alexander's player. The limit was set at $350, yet ten minutes later an email arrived about charges of $37,700, and the service was blocked. At first it was unclear whether Amazon had made a mistake or Claude Code had got around the limits. Later it turned out Amazon was to blame: when Alexander, on Claude's advice, removed the limit, the company wiped that amount and showed $112 instead.
How to get started: the hosts' advice
Alexander Volchek's first piece of advice is not to skimp on the model. A friend of his who had started automating his finances complained that Codex “doesn't understand anything,” and it turned out he had been working on the lightest model. In Claude Code the host recommends choosing Opus 5.5 or Fable 5.1 and switching on Max mode. A plan at a hundred to two hundred dollars, he says, is justified once automation begins, and he advises keeping two subscriptions — a pricier one at $100–200 and a simpler one at about $20.
Next, move real work into Claude Code: half or more of the tasks you now handle in a chat, starting with the weekly ones. It is better to set the task in plain words rather than as a technical specification, and to hand it over whole — with all the history and documents. Alexander considers courses optional: thirty to fifty requests a day on your own topics teach you faster, and how to work in the system you can ask the system itself. ToTheMoon episode transcripts on besolid.com work too — you can give them to Claude Code as ready-made instructions.
Third, work directly, without aggregators, wrappers and unnecessary plugins: that is what Alexander does himself. Whether an extension actually helps can now be checked inside Claude Code: the claude plugin eval command runs the same tasks three times with the plugin and three times without it, in clean sessions with no memory — and spends your subscription limits on it.
Claude Code vs Codex: comparisons, limits and debates
On the podcast, Claude Code almost always comes up alongside Codex. Alexander Volchek divides the roles like this: email and services are easier to connect to Codex, since behind it stands ChatGPT, which for him is stronger than Anthropic's system, while as a development environment Claude Code, in his words, is “better than Codex”: he rates the Opus 5.5 and Fable models in Max mode above 6 Astra and 6.1 Sol. Cursor and Google Antigravity, in his estimate, lag behind both systems, and Gemini, Grok and DeepSeek currently lack some of their capabilities.
The second subject of debate is limits. On OpenAI's $200 Pro plan, Ilnar Shafigullin for the first time saw, more than once, a message that the model was in too much demand, and, preparing for the end of the reset marathon, deliberately works in Claude Code in parallel; Anna Volchek does the same. Hence the advice to split tasks between the two systems in advance: when a limit runs out, Alexander says, all your daily tasks stall. Ilnar links Anthropic's stinginess with resets to the company having invested in data centers under the pessimistic scenario Dario Amodei talked about, while demand turned out to match the optimistic one.
There are minor annoyances too: tasks Alexander dictated on his way home he later could not find — the interface had been updated, the projects rearranged and the task given a different name. On the other hand, it now has Claude Code Projects: a project has a shared goal, parallel tasks each run in their own branch, and repositories, environments and connections are gathered in one place; they are being rolled out in stages to Pro and Max. Meanwhile Anthropic claims that tens of thousands of agents are building code, and in the same breath says that a hundred agents still cannot agree among themselves.