About

A free calculator built on the US Foreign Service Institute's published class-hour figures, adjusted for the level you actually want, the languages you already speak, and how you really plan to study. It has no account, no paywall, and no upsell.

It will sometimes tell you your plan cannot reach your target at all. That is the most useful thing it does, and it is why the model has a ceiling per study method rather than only a multiplier.

Affiliate commissions on 5 of the 7 courses and tools we cover. Every page carrying such a link says so at the top, above the link, not in the footer.

Three things follow from that, and they are enforced in code rather than promised:

• The calculator and its results carry no affiliate links. Outbound links live only on course and comparison pages.
• The matching rules never read commission rates. An automated test rewrites every payout in our catalogue to $999 and to $1 and fails the build unless the recommendations come out identical.
• Every shortlist includes at least one option that pays us nothing — currently 2 of them.

We do not publish star ratings. We have not run these courses side by side for months, and inventing a score to win a rich snippet would be exactly the fabrication this site exists to argue against.

We do not present estimates as research. One number here is published by the State Department; the four coefficients applied on top of it are our own judgement, and each one is written up with its reasoning.

Kevin Fan. This is a one-person project — the model, the code, and every page on it.

I am a native Mandarin speaker, which is why Chinese was the first language written up here rather than the obvious high-traffic choice. Most English-language writing about learning Chinese repeats the same three claims about tones and characters; I can at least tell you which of them survive contact with the language.

I have studied one language seriously: English. Not a list, and I am not going to pad one. What that is worth here is specific: Mandarin and English sit in each other's most expensive band, sharing no vocabulary, no script and no grammar to lean on. It is the same shape of project this calculator prices at 2,200 hours, and I did it from the far side.

So I have not sampled thirty languages, and I have spent the kind of hours the model estimates — which is a different and, for this purpose, more useful thing. I know which parts of an estimate feel wrong from the inside: that the writing system is the small part, that the plateau arrives on schedule, that the gap between reading comfortably and following two people talking is much wider than an hour figure suggests.

You can judge the result without taking my word for it. You are reading it.

The claims above are meant to be verifiable rather than believed, so here is where to look:

• Every citation has been fetched. 21 sources, each recorded with the date its URL was last confirmed to load. A source that returns 403 to an automated request but works in a browser is marked as such rather than quietly dropped.
• Every constant carries its provenance. The 32 language baselines, the level factors and the method coefficients each record where they came from, how confident we are, and — where the underlying frameworks disagree — what the dispute is.
• The rules are enforced by tests, not by intention. Commission-blindness, the free-option guarantee, the absence of affiliate links on the calculator, and the requirement that every published hour figure matches the model — all of them fail the build when broken.
• Nothing is bulk-generated. 33 language write-ups and 8 tools, each with content specific to it. A test rejects a page whose questions repeat another page's.

Four errors have shipped here so far, and three of them passed a full green test suite. The calculator once returned 1,782 hours for elementary Chinese. A $260 course once displayed as "Free" and was recommended to people with no budget. A citation once pointed at a page that had never existed.

All four are written up with what produced them and what changed. A site that traces other people's unsourced numbers does not get to hide its own mistakes, and the errors are more informative about how much to trust the rest than any assurance would be.

This is a one-person project and that has limits worth stating plainly. Nobody here has taught these languages professionally. No course has been run side by side against another for a month. The coefficients on top of the FSI baseline are considered judgements, not measurements, and a specialist would improve several of them.

What a small project can do instead is show its working completely and correct quickly. That is the trade being made, and it is the honest reason there are no star ratings here.

If a figure here is wrong, or a source has moved, we want to know. Commercial terms in particular expire — every merchant record carries the date it was last checked, and anything over ninety days old is flagged on the page as unverified.

kevin.fan@large.net — corrections, and the address the privacy policy points at for seeing or deleting anything held about you. It is one person's inbox, so a reply may take a few days; a request is acted on whether or not it gets a reply first.