In short: what's the difference
Both names have long stood not for a single chat but for a whole lineup. At OpenAI it is ChatGPT, Codex, Work and the Dots assistant; at Anthropic it is Claude itself, Claude Code, Claude Duet and the Opus and Fable models. These are the two companies Alexander Volchek calls the main leaders, and he advises working with them directly, without aggregators and wrappers. The choice between their assistants, Dots and Claude Duet, he calls a choice of infrastructure, not of model.
Alexander's own split goes like this. Email and services are easier for him to connect on the OpenAI side, since Codex is built on ChatGPT, which Alexander values more than Anthropic's system. For development, though, he picks Claude Code — which, as he puts it, is “cooler than Codex”. In September he discussed a model ranking that, he said, Elon Musk had conditionally endorsed: Claude Opus 5.5 sits in class SSS, ChatGPT-6 Astra in SS and Fable in S, although Alexander personally rates Fable 5.1 higher.
When the hosts choose ChatGPT
Alexander Volchek's first argument for ChatGPT is the infrastructure around everyday life. Over thousands of conversations, the ChatGPT, Codex and Dots combination has learned his daily routine, and Dots comes to him with suggestions on its own and warns him about problems. In the US and a number of other countries, the Finance module connects ChatGPT to practically every financial institution. The host even shops past Amazon more and more often: ChatGPT finds him stores where things are cheaper, and items that can't be found on Amazon at all.
The second argument is limits. For a long time good models in ChatGPT ran practically without restrictions, and Alexander called the most expensive Pro plan, at five hundred dollars, unlimited even on Astra; his cases were built on exactly that — from tens of thousands of pages and a million calls to two hundred thousand tickets. Ilnar Shafigullin considered limits one of OpenAI's main trump cards against Anthropic — and so he was upset when they were cut in half at DevDay. In the episode on job hunting, Anna Volchek praises both systems equally for tailoring a resume to a role, but recommends debriefing a past interview with ChatGPT.
When the hosts choose Claude
Alexander Volchek turns to Claude when the work is complex. He considers Opus 5.5 the best model today for mass development and automation. Fable amazed him even before the model was blocked: launching it in Claude Code straight from his phone by the pool, he watched the model take a big task, described by voice, all the way to the end. And when Fable 5.2 came out, the host moved a task with too much money at stake onto it overnight, because he didn't fully trust Opus 5.5: in complex programming, he says, the new model wins many times over.
Fable, Alexander advises, is for cases where a mistake is costly and the result has to be verified: contracts, finance, finding errors in processes, analyzing advertising and strategy. It also needs a good plan: on the twenty-dollar one it is no use, and it would eat through his two-hundred-dollar subscription in four to five hours. Meanwhile Fable 5.1 costs five to ten times more than Opus 5.5, although it sometimes uses fewer tokens per task. For small everyday things, the host believes, there is probably no need to bring in a model like this.
Claude also has a controversial habit. Although the model has long known that Alexander is American, by default it adapts his Russian-language projects to Russian law, and a landing page for Belarus to Belarusian law. The hosts explain it differently: Ilnar Shafigullin sees the cause in the training data, Alexander thinks the model looks first of all at your location, and Anna Volchek gets all of the AI's advice from a Polish point of view. Alexander's conclusion: know the laws, but don't fall into paranoia.
What has changed in recent months
In the autumn of 2026, both companies reorganized. OpenAI wrote that within the next one to two weeks it would merge ChatGPT and Work and scrap unlimited chat, and Alexander Volchek already notices that ChatGPT no longer handles the data-analysis volumes it used to. Subscription limits were first cut and then, by November, doubled back: on the $200 plan it is once again twenty “x”. Anthropic is stingy with resets — Ilnar Shafigullin explains this by the company having bought capacity for a pessimistic scenario, while demand turned out to be far higher.
One-window assistants have appeared too. Claude Duet came out in preview: access went only to Pro subscribers in the US, and even then through a waitlist. Alexander was quickly disappointed: Duet doesn't see sessions already created in Claude, spends your tokens and, in his view, is aimed more at running a business than at everyday life; he soon stopped using it, calling it an unfinished copy of Dots. Anthropic, on the other hand, released Fable 5.2 and described a memory in which Claude double-checks and corrects its own notes during pauses.