How it's made

The problem the show is built around

Philosophy is hard to break into pieces you can learn one at a time. Try, and you usually end up teaching the history of philosophy instead — Plato, then Aristotle, then everyone after them. That tells you who said what. It leaves you no better at thinking through a question of your own.

This show is built the other way round. It starts from questions people actually have — what do I owe a stranger, is there anyone in the machine, should I have children — and brings in a thinker only when the question needs them. Peter Singer arrives in the first series because the question about the stranger needs him. Descartes waits until the fifth. No series is "the Greeks" or "the moderns". Underneath each question there is a curriculum, worked out before a word of the episodes was written, but you never see it. You see the question.

The teaching idea

The way each series is built is adapted from a way of teaching called Teaching for Understanding, developed at Harvard's Project Zero. I had used it to build a course site, Learn to Drive AI, before applying it here. Its central distinction is between knowing about an idea and being able to use it. Knowing what Singer argued is the first. Being able to say which step of his argument you refuse, and what that refusal costs you, is the second. The show is built for the second.

In practice that means four things, and each series page shows them:

• A question worth the time. Each series starts from a question that matters outside a classroom.
• One thing you get better at. Each series is built to make you better at one specific thing. It is written at the top of the series page, under "What you practise".
• Doing it, not hearing about it. The episodes ask you to run the argument yourself, on cases. "Your turn", at the end of each series page, gives you a fresh one that the episodes never touched.
• Coming back to the question. Each series returns to the same question after every episode, harder each time, and does not settle it.

One honest limit. The framework assumes a teacher who reads your work and tells you how you did. A podcast can't do that. Nobody here marks your answer. That part is yours.

How the questions were chosen

Not by hand, and not to cover the field. The forty-four questions came out of a process that ranked candidate questions by how pressing they are for someone living now, and kept the most pressing. The aim was never a balanced tour of every thinker and tradition. The order runs from questions close to home — grief, loneliness, what you owe a stranger — outward to questions about the shared world, and finally to questions about the tools of thinking themselves.

Who makes it, and what

I'm a former teacher with a long interest in philosophy and in AI. What I built is the process: how a question is researched, turned into a curriculum, and then into a series; the checks every draft has to pass; and the standard the whole thing is held to. I did not write the episodes, and I did not pick the questions — the process did both, which is the point of having one.

Claude, Anthropic's AI model, does the research, designs each series, and writes every episode. A second model from a different company, OpenAI's Codex, reads the drafts without seeing the notes behind them and checks three things: that arguments are attributed to the right people; that the view you are least likely to hold is given its strongest words; and that no episode quietly hands down a verdict. Two model families on purpose. A model checking its own work shares its own blind spots.

The voice is synthetic: OpenAI's text-to-speech. Nobody reads the episodes aloud. I authorised every recording. No philosopher outside the project has reviewed the material.

Not neutral

We choose the questions, the order they come in, and which arguments get the most time. That is a standpoint, and we say so. What we don't do is tell you the answer. The view you are likeliest to reject is stated in the words its defenders would use, and every series ends with the question handed back, sharper, and open.

The sources, and what "checked" means

Fluent prose is not evidence. So each series page lists, under "Sources", the works the series was built from, and under "Where to start", a few places to go next. Every entry says in plain words what it gives you, and whether the link is the work itself or a summary of it.

Where an entry is marked "checked", a reader opened the text and found the passage the episode relies on, and a second AI model from a different company confirmed that the passage says what we say it says. Where an entry is not marked, we haven't done that yet, and we say so rather than pretend. The full record for each series — what was checked, when, and against what — is kept, and will be published once it is complete for all forty-four.

Independence

This show is independent. It has no connection to Harvard University, its Graduate School of Education, or Project Zero, whose teaching framework it adapts. It has none to Anthropic or OpenAI, whose models were used to make it. None of them has seen it, reviewed it, or endorsed it.

— Josh Torkington