Language Log #6: One-Year Check-In on Learning 20 Languages

A quarterly progress update tracking study hours, review counts, and real speaking attempts across 20 languages, one year into a challenge to reach functional fluency in multiple languages simultaneously.

Challenge Recap (00:00:09 – 00:06:07)

• Original goal: functional fluency in 10 languages within 4 years, based on 1 hour of daily study. FSI estimates suggest ~40 years using traditional methods.

• Now studying 20 languages instead of 10, partly to stress-test the Phrasing app across every language family it supports.

• Last quarter averaged only ~30 minutes/day (45 hours over 90 days) due to a major backend rebuild of Phrasing's sync engine — the biggest technical project to date, enabling full offline support and better performance.

• Goal for the next quarter: ~1.5 hours/day to average out to the 1-hour/day target.

Tier 1 Languages: Heaviest Focus (00:06:07 – 00:26:14)

• Croatian (00:06:18) — Now 33% of all reviews (up from ~13%), 29 total hours, most progress of any language. Attempts a spoken demo.

• Greek (00:10:09) — Bumped to 17% of reviews ahead of a planned trip to Greece, 12 total hours, describes finally "breaking through" after a year of light exposure.

• Japanese (00:12:30) — 17% of reviews, 43 total hours; doubling review volume for two weeks produced an immediate, noticeable jump in progress.

• Turkish (00:14:06) — Reduced to 8.4%, 33 total hours; despite equal study time, progress lags far behind Croatian due to unrelated grammar and vocabulary.

• Arabic (00:18:10) — Also 8.4%, 41 total hours; used casually in real conversations while traveling in Marseille.

• French (00:22:38) — 7 total hours; already fluent, used as a benchmark language, records an update spoken in French.

Lower-Priority "Warm" Languages: Minimal but Real Progress (00:26:14 – 00:47:12)

• Cantonese, Lithuanian, Macedonian, Maltese, Estonian, Mandarin Chinese, Sanskrit, Italian, Spanish, Dutch, Welsh, and Portuguese, each getting roughly 1.1% of reviews (a couple minutes to an hour per quarter).

• Notable finds: Maltese sentence structure unexpectedly clarified Arabic grammar; Estonian's case system made grammatical cases click in a way German never did; passive exposure to Spanish let him correctly guess verb conjugations without ever studying them.

Two New Languages Added (00:43:23 – 00:47:12)

• Basque (00:45:11) — A pre-Indo-European language isolate, added for its linguistic uniqueness.

• Thai (00:46:31) — Added to stress-test Phrasing's handling of tonal languages and a script with no spaces between words.

Key Takeaways (00:50:04 – end)

• Small effort produces disproportionate results: doubling Japanese reviews for two weeks caused an overnight shift in comprehension.

• There appears to be a low "critical mass" threshold, roughly 5–10 reviews a day, below which progress feels invisible even though it's happening subconsciously.

• Doubling study time in a language (e.g. Estonian) can take it from passive exposure to genuinely feeling like active learning.

  
  

AI Transcript

00:00:02 All right, and welcome back to another episode of Phrasing FM, a podcast where I talk about languages and my experience building an app to learn them. Today I want to do a check-in on my language progress. This will be Language Log number six, part of a series where I'm learning a bunch of different languages and trying to get to a functional level of fluency in ten languages, and see how long that takes me, how many study hours it takes, and how much faster I can do it with phrasing than the FSI estimates would suggest. The FSI estimates would suggest it would take me forty years to learn these languages, and I'm trying to do it much, much faster than that.

00:00:48 The original goal when I started this challenge was to do it in four years. However, a couple of things have changed. One, that number was based on spending an hour every single day, and so far I have not maintained an hour of study every day, so it will take longer just because of that. Secondly, I originally planned this out in ten different languages, and I'm now up to twenty languages that I'm studying at the same time. That's naturally going to slow down this process, but hopefully not by much.

00:01:23 Of course, I'm not studying all languages evenly. A large reason I'm studying so many languages is to test them in phrasing. Phrasing supports around one hundred languages, ninety-ish languages plus forty dialects, and I want to make sure it works in all of these. I can't use it in every single one, that would be too crazy, but I'm trying to use it in every language family I support to make sure I'm making progress. If it works in one language in a family, it likely works for other languages in that family. And I tend to learn the more minor languages in a family, so if it works in the minor one, it should work in the more major ones as well.

00:02:14 But I'm not learning every language evenly. In phrasing, you can configure the distribution of your reviews. For the last year, I've been studying a lot of Turkish, Arabic, Croatian, and Japanese, while some languages I haven't even seen every day, just a couple reviews per week. I change these around trying to find the best fit. Recently, about a week ago, I made the biggest change so far in my study distribution. So I'm going to go through all twenty languages I'll be studying over the next couple of months, talk about my progress, how much I'll be studying them, and how they're going. If I can speak in a language, I will, to demonstrate progress as a record. However, with so many languages, a lot of them I don't expect to speak, since I've spent a cumulative two or three hours on some of these.

00:03:26 Before jumping in, a quick recap: it's been nearly a year since I started this challenge, since last June, so this is a great time to check in. I try to check in every quarter or so, and a lot of these statistics are for the last three months. I did not do very well studying over the last quarter, averaging about forty-five hours over ninety-ish days, roughly half an hour a day. There are reasons for this: I had to move, but the bigger reason was I did the biggest update to phrasing yet. It didn't change anything visually, phrasing is still the same product, but I changed how all the data is stored, saved, and synced. This has huge benefits: phrasing now works entirely offline, page loads are near instant, performance is way better, and a bunch of bugs were fixed. But this was the biggest project I've ever worked on, brutal to build, and after struggling to get it launched, the last thing I wanted to do was use the application. I still checked in every day and did reviews, but there were some low-effort days, and it wasn't really until a week or two ago that I got back into a good routine. So not the best turnaround this quarter, but hopefully that means I can spend even more time studying next quarter, maybe around 135 hours, which would average out to about an hour a day across both quarters.

00:06:07 So let's jump into the languages, ordered by how much I'm studying them. Number one is Croatian. It's always been a tier-one language, on par with Turkish, Arabic, and Japanese, but I've made the most progress in it. It's the only Indo-European language in that top category, which I think is why. I'm now considering a trip to Croatia, and I've decided to lean into it: it's going from about 12–13% of my reviews to over 33%, one third of everything I study. I've done 1,000 reviews in the last quarter, six hours studying, for a total of 29 hours over the past year. I don't really speak Croatian yet, but I can pick out words in conversations, understand gist, and occasionally interject. Far from conversational, but making progress. (He then attempts a short, self-described "butchered" spoken passage in Croatian.)

00:10:05 The next language is Greek. Over the past quarter I did 373 reviews, about three hours, for a total of twelve hours on phrasing, more than expected given it's usually a low-priority language. For the next quarter it's jumping to 17% of my reviews, more than any tier-one language got before, partly because I might travel to Greece for a couple weeks and want to be able to order coffee and read signs. I'm not expecting fluency, just something like "disjointed fluency," an A-level. Interestingly, even with very few reviews, once I bumped up study time I started making real progress fast, like pouring gasoline on an already-existing fire. Past attempts at Greek felt like hitting a brick wall, but a year of light exposure broke through that wall brick by brick, and now it feels like smooth sailing.

00:12:28 Language number three is Japanese, also at 17%. I realized I was making good progress in Croatian, Arabic, and Turkish, but not Japanese, despite doing the same number of reviews. So for a couple weeks I doubled my Japanese reviews, almost 50% of all reviews for that stretch, and immediately felt progress kick in. Last quarter: 1,500 reviews, eleven hours, for a total of 43 hours. Japanese is just very foreign, requiring more raw exposure than my other tier-one languages. Keeping it elevated at 17%, up from 12–13%, in hopes it's enough to keep the progress going.

00:14:01 Language number four is Turkish. In the last 90 days: 750 reviews, four hours, for a total of 33 hours. Reduced down to 8.4% of reviews, from 12–13% before. Interestingly, Turkish and Croatian are rated the same difficulty by FSI estimates for English speakers, but despite equal study time, Croatian progress has far outpaced Turkish. A new Croatian sentence still feels like a sentence, just with unfamiliar words, while a new Turkish sentence feels like random letters on a screen, since Turkish is a Turkic language with a completely different sentence structure (subject-object-verb), vowel harmony, and no direct translatability the way Croatian and English share. Still happy with progress until realizing it "should" match Croatian; hoping to eventually break through.

00:18:10 Language number five is Arabic (specifically Lebanese Arabic), also at 8.4%. Last 90 days: 633 reviews, four hours, for a total of 41 hours, the second-most studied language overall. Arabic is the language that originally motivated building Phrasing, since existing resources were so inadequate. It no longer feels more or less difficult than any other language in rotation. While in Marseille, he had several real Arabic exchanges with native and heritage speakers, small talk about where people are from, understanding and forming responses using vocabulary directly learned from phrasing. Both Arabic and Japanese are being kept as long-term exposure languages, the idea being that years of light exposure now will make serious study later feel much less foreign, more like learning Spanish than starting from zero on Chinese or Arabic.

00:21:51 The rest of the languages are each 1.1% of reviews, meaning roughly two reviews a day on average if doing 200 reviews daily. Not much, but enough to keep them "warm" while focus stays on getting Croatian to a B1 level.

00:22:34 Language six is French. Last quarter: 522 reviews, two hours, for a total of seven hours, notable because French is already spoken fluently. (He records an update spoken entirely in French, explaining that using words learned through phrasing translated immediately into real conversations, unlike past experience with Anki, where he'd recognize a word in review but blank on it in real conversation.)

00:26:14 Language seven is Cantonese: 269 reviews, two hours, nineteen total hours. Originally a tier-one language, swapped out for Japanese since he doesn't consume much Cantonese media. Still loves the sound of it, especially the tones, more fun to speak than Mandarin even if Mandarin "sounds cooler." Treating it the same as Arabic and Japanese: building exposure now for serious study later.

00:27:29 Language eight is Lithuanian: 259 reviews, about one hour, twelve total hours. Very little done recently, feels like backsliding from a previously higher priority. Has personal heritage ties to Lithuania and a soft spot for Baltic history, but isn't getting enough exposure to make real progress right now.

00:28:08 Language nine is Macedonian: 242 reviews, about one hour, thirteen total hours. Benefits heavily from Croatian knowledge due to shared Balkan sprachbund features, comprehension transfers well (similar to Portuguese and Spanish), though speaking mistakes happen from assuming false equivalence, e.g. "dobro jutro" ("good morning") sounds similar but isn't pronounced the same way. Plan is to focus on Croatian first, then return to properly distinguish the two.

00:29:37 Language ten is Maltese: 223 reviews, about one hour, ten total hours. Discovered by accident via a YouTube video, a Semitic language written in the Latin alphabet with heavy Romance loanwords from Sicilian and Italian contact, an EU language. Went from not knowing it existed to studying it on phrasing within 30 minutes. Learning Maltese unexpectedly clarified a lot of Arabic grammar, since seeing it in consistent Latin script makes sentences easier to parse at a glance than Arabic script. Also doubles as a stress test for phrasing's handling of extremely low-resource languages; sentence generation is less accurate for minor languages, so bringing your own sentences is recommended.

00:32:00 Language eleven is Estonian: 198 reviews, one hour, six total hours. A heavily case-inflected, agglutinative language similar in structure to Turkish. The combination of agglutination and many grammatical cases actually made the concept of "cases" click far more than German's four cases ever did, cases in Estonian feel semantic rather than arbitrarily grammatical.

00:33:26 Language twelve is Mandarin Chinese: 159 reviews, one hour, eight total hours. Minimal progress, it's never been prioritized highly enough, and review volume is too low relative to how foreign the language is to retain what's learned. Cantonese has consistently gotten more attention. Recommends spending significantly more than six hours a year on any language to see real progress.

00:34:33 Language thirteen is Sanskrit: 159 reviews, one hour, eight total hours (though effectively closer to two, after a recent reset). Originally built from memorized yoga mantras already known by sound but not meaning, this turned out to be a flawed method, too easy to "cheat" through sentences already memorized without real comprehension. Those cards were removed and study restarted from scratch. Sanskrit is Indo-European but about as foreign as that family gets. Lesson learned: don't build flashcards from sentences you already have memorized to try to learn their meaning; it doesn't work well.

00:36:03 Language fourteen is Italian: 118 reviews, half an hour, three total hours. Minimal but Italian feels fresher lately, possibly due to overlapping French exposure. Would like to properly refresh Italian and Portuguese at some point, but not this quarter or next.

00:36:55 Language fifteen is Spanish (Castilian): 105 reviews, half an hour, two total hours. Never been a priority to actively learn, expected to come "for free" through exposure given existing French, Italian, and Portuguese. This has held true: in conversations with Spanish learners, guesses about verb conjugation turned out correct despite no dedicated study, category induction from related languages working automatically. Also warming up to the Castilian accent specifically.

00:38:29 Language sixteen is Dutch. (He attempts an extended, admittedly rough spoken update in Dutch, mixing in French and describing his review stats and app terminology like "language contouring" and "excelled" cards along the way, acknowledging the language flow keeps breaking.) Despite very little study, roughly two hours in the past year, he notes the app still surfaces the right words, which he finds encouraging even at minimal exposure.

00:42:04 Language seventeen is Welsh: 87 reviews, thirty minutes, four total hours. Historically felt like zero connection between written and spoken Welsh, similar to the disorientation of first learning Portuguese pronunciation after other Romance languages. That's starting to change: sentence flow and small bits of grammar are starting to click, though expectations remain modest given only four hours total.

00:43:23 Language eighteen is Portuguese: 70 reviews, fifteen minutes, two total hours. (He attempts a short spoken update in Portuguese.) Despite minimal recent study, he notes this is actually better spoken Portuguese than in prior updates, remarkable given the tiny time investment, though this is helped by having previously spoken the language.

00:44:33 The final two languages are new additions, added about a month ago after long being on a waitlist of roughly 25 languages meant to eventually cover every language family. Language nineteen is Basque: 41 reviews, fifteen minutes total. A pre-Indo-European language isolate spoken in parts of Spain and France, notable for having survived surrounded by Indo-European languages for millennia while nearly every other pre-Indo-European language family in Europe was absorbed or displaced. Added purely out of fascination with its linguistic uniqueness.

00:46:31 Language twenty is Thai: 26 reviews, about thirty minutes total. Added partly for the script, one of his favorite orthographies, and partly to stress-test phrasing's handling of tonal languages and languages that don't use spaces between words. Fixed several bugs specific to Thai in the process and wants to make sure the app offers a solid experience for the Thai-learning community.

00:48:17 So those are the twenty languages and their new distribution for the next three months: Croatian at one-third of all reviews, Japanese and Greek at the second tier (17% each), Turkish and Arabic at the third tier (8.4% each), and the remaining fourteen languages splitting the rest. The goal is to go from half an hour a day to roughly an hour and a half a day, averaging out to the target one hour a day. Once Croatian reaches a B1 level, reviews can shift toward bringing Italian and Portuguese back to conversational levels.

00:50:04 Wrapping up with some takeaways: First, it's remarkable how much progress is possible with an hour or less of total study over ninety days, real, felt progress despite the small numbers. Second, bumping up Japanese reviews produced an overnight, noticeable shift in comprehension, having a "knob" you can turn at any time to feel immediate results is genuinely useful. Third, and somewhat in tension with the first point, there does seem to be a real critical mass needed to make meaningful, felt progress, and that threshold is lower than expected, maybe five to ten reviews a day. Below that threshold, progress still happens, but subconsciously and invisibly; above it, gains start to compound in a way you can actually notice and that scales roughly linearly from there. Comparing Estonian and Lithuanian illustrates this: doubling study time in Estonian took it from passive exposure to something that actually felt like active learning.

00:53:21 Also very exciting: after a year of just twelve hours of exposure to Greek, watching an interview with the mayor of Thessaloniki (with subtitles) revealed he could pick out several words per sentence and hear them as distinct words rather than a wall of unfamiliar sound, real comprehension from very little formal study, and a strong signal that "keeping languages warm" for later serious study is genuinely paying off.

00:55:30 So that's it for today. If anyone's still listening at the end of this update, thank you so much. Consider liking, subscribing, and sharing. If you're also interested in learning a bunch of different languages, consider giving phrasing a try, I've been using it every day for over a year now, across what's now twenty languages. You can make good progress in a lot of languages at once and mix it up, it's just a lot of fun. Consider giving it a try, consider subscribing. And if anybody's still here: try guessing in the comments which five languages I want to add next, there are another five in my head I know I want to add eventually. Catch you all in the next one. Ciao!

Listen to this podcast episode on youtube: