2026-06-17 151 ¶

100 days of learning Chinese

Today marks 100 consecutive days of listening to Chinese, ranging anywhere between 30 minutes to 4 hours per day. I now have an adequate ear for delineating individual words and matching sounds to hanzi when subtitles are shown, though any native speech at full speed is, of course, completely incomprehensible. These two videos are a good representation of where my listening ability is right now:

1. Easily above 90% comprehension
2. Maybe 50%? This one is difficult to rate. I feel well grounded throughout the video, which I could not say of Annie's videos 50-75 hours ago. If I had to estimate, it's probably only one word or so per sentence that I actually recognize. While there is an occasional sentence I understand fully, most of the video I could not understand without visual assistance.

Following up from the post four days ago, there's nothing quite like stating one's intentions clearly to affect a complete reversal. I've spent some time watching more methodology videos and I no longer think sticking to ALG is right for me.

I'm dead certain that the Japanese sentence cards from the Kaishi 1.5k deck elevated my Japanese quickly, and I like the ideas behind Will Hart and Evildea's Anki method from some recent videos. From here on out I'm going to use Anki to mine sentences and drill "chunks" of Chinese. I'm also going to play with sound as Stuart Raj does to figure out how to pronounce each phoneme.

Here's some of what I've been perusing that made me reconsider my strategy:

• Should you speak from day 1?
• Analyzing Pablo's Thai
• Migaku founder gets fluent in Portuguese
• Evildea's grind
• How Will Hart uses Anki
• Why Matt v Japan changed his mind

There are countless examples of this method that still incorporates mass input, but reinforced with "chunked" SRS, and a whole lot of doubt lately about the late-game claims of ALG.

2026-06-12 148 ¶

My Implementation of ALG

I'm taking a lot of inspiration from the ALG school of thought, but I'm not totally convinced of all of their claims. I'll be modifying their method a bit, but I think it will still count as an implementation of ALG in the end. To start, here are my hypotheses, based on my journey through Japanese and hearing stories from other learners.

1. Reading early has a profoundly negative effect on pronunciation
2. Speaking potential is a pure function of input quantity, while speaking ability is an almost fully disjoint skill.
3. Adult learners cannot identify tongue position from aural acquisition and self-correction alone.
4. Bilingual definitions have no negative effect on vocabulary acquisition in the target language

David Long and Dreaming Spanish's ALG both agree with me on point #1. It's on the rest where we differ. In Matt v. The World's new channel, he argues from experience with his students and other YouTubers that reached high levels of fluency that delaying speech to the 2,000 hour mark is not only unnecessary, but actively detrimental to developing self confidence in the early output stage. I began speaking Japanese at the 1,000 hour mark via lessons once a week on iTalki, and it was, predictably, awful. It felt like the whole language was on the tip of my tongue, even very common words that I had fully acquired and assimilated such as "cat" or "green" were somewhere in the back of my head, but since I had never made the effort to actually say them out loud on demand, there was a long delay before it came out. When a partner is waiting for you, this can be pretty overwhelming. I liked Matt's explanation: earlier on in the language journey, you just expect to sound awful because you're new -- it's just a fact. Delaying speech until you have strong listening can be a big ego hit, because you expect more of yourself.

Point #3 is what I'm going to try and disprove while learning Mandarin. For Japanese, I made it to 1,000 hours, attempted speaking, and knew that my "shi", "wa", and "u" all sounded unbelievably wrong, but I could not for the life of me figure out why. It never occurred to me to put the tip of my tongue behind my lower teeth for "shi" until I looked up a phoneme chart and watched a pronunciation video; I simply couldn't conceive the possibility of a different tongue position that produces so similar a sound to the English "sh".

Here's the rough breakdown I'll be using for Mandarin. Throughout all stages I'll be using video/audio comprehensible input, but with the following unlock stages:

• 0-600 hours: Silent period with crosstalk
• 600-2,000 hours: Code switching during crosstalk. Shadowing for pronunciation self-correction.
• 2,000-3,000 hours: SRS for reading

In this implementation speech unlocks at 600 hours, which still aligns with David Long's original estimate of when students begin speaking Thai. The key difference is that during crosstalk I will attempt code switching for words I've acquired. This will be my method for actively prodding my mind for active recall of Chinese words without using SRS and without the pressure of staying 100% in target language. By the end of the 2,000 hours I should be able to speak entirely in Mandarin.

From 2,000 hours to 3,000 hours I'll use an SRS to learn how to read Chinese characters for words I've aurally acquired. My #4 hypothesis is a direct contradiction to David Long's position. I have yet to see a good argument for how looking up words like "quartz", "arsenic", and "manatee" via a Japanese to English dictionary could possibly be any different than learning that "somnambulism" is the same as "sleepwalking." It's obvious that there are subtleties and nuance to each individual word, including across languages, but when acting as an anchor for specific concepts where a one-word definition suffices, I'm not convinced there is any benefit from remaining solely monolingual.

Despite my lenient position on bilinugal lookups, I'm planning to stay completely monolingual in Mandarin. This is in keeping with ALG's recommendation, so the only real change I'm making from David Long and DS is earlier output via code switching and shadowing. Both Long and DS are opposed to shadowing, which I find surprising. Taking Long's assertion that "we can learn how a child learns" at face value, then one would think mimicry a critical part of the beginning output stages; children love to repeat what they hear.

By taking what I still think is a "pure" ALG approach to a language that's fully unrelated to both Japanese and English, I'll at least have some idea of how it compares to my previous CI-focused but traditional-supplemented method for Japanese.

References

1. David Long Explains ALG
2. Matt v. Japan advocates earlier output
3. Evildea after 1,000 hours of ALG in Spanish

2026-05-11 132 ¶

Dark night of the soul

Almost entirely gibberish.

Fully comprehensible, despite several new words.

I am rapidly approaching the worst section of the journey. When the language is still new, it's engaging just through that trait alone, but it's now dipping into the big trough of disillusionment that happens through most of level 2 (100+ hr) and all of level 3 (300+ hr). The language no longer feels new or exciting, and I still have a long grind until everyday content starts to make sense. Level 4 (600+ hr) will be about where the "fun" factor starts to dip back up, and, based on my experience with Japanese, I think the start of level 6 (2000+ hr) is the most enjoyable section.

Posts here will slow down for a bit as my daily intake is reduced to around 30-40 min/day. This is to stave off burnout with Chinese while still making progress to level 3 by the end of the year, and leave enough time to hit my level 6 goal for Japanese by August.

2026-04-24 103 ¶

Week 6: Level 2

Assuming the Dreaming Spanish roadmap (x2) is accurate for learning Mandarin, I've now moved on to Level 2! I like the lengualytics description of this stage:

100 - 300 hours

You can recognize some common words and short phrases, especially when spoken slowly and clearly. You understand the language in chunks of meaning as opposed to singular words.

You're beginning to recognize patterns in the language. Familiar words and phrases start to pop out--like common phrases used in intros and outros, common nouns, and common verbs. You're still not fully accustomed to all the sounds of the language, but by the end of this level you will be. You should be able to start to look back at some of the earliest content you viewed and notice it's significantly easier to understand than it once felt.

This feels extremely accurate, and also I'm at the low end of this ability. Right now the main challenge seems to be that when watching even very simple material designed for learners or toddlers, I can pick out questions about colors or simple verbs (stand/sit/walk/run) quickly, but everything else is opaque. For example, in this "elephants: true or false?" video there are long passages, such as when she asks "do elephants like to eat meat?", that are clear to me, but the vast majority of what she's saying is just a stream of familiar sounds with no meaning. I have no idea what question 4 in this video was asking, but I know that the answer was "true". It might be worth noting that I'm covering all text using bennycopter's subtitle hider, so there may have been text in the box that would give a clue, especially if it uses numbers or characters that are the same between Chinese and Japanese.

2026-04-17 83 ¶

Week 5: First fully comprehensible video

Blabla Chinese's Father and Son series is probably the most language-dense thing I've found at the superbeginner level. Even when first starting out, it was easily comprehensible via Amber's excellent use of visual aid, but none of the comprehension came from her words. Today was different, though. Long passages such as "whose hair is this? Is it my hair? No. Is it Lady Gaga's hair? No. It's Father's hair." I could fully understand just through audio alone. Last week I was picking out just the most common sentences, but this time I understood a full ~6 minute video at the superbeginner level.

Anything beyond a very simple, visually-aided video, though, is still completely out of reach. The Mandarin Click newcomer videos, while slow and simple, are still tough to fully comprehend, since I can't discern much from the pictures. The Story learning with Annie spot the difference series, aside from where she points at specific things in the image, are almost completely incomprehensible. I can only pick out a word here or there.

I'm also keeping up weekly crosstalk with my coworkers to mixed success. I still need to figure out a good way to do this over video call, where we're missing the speed, accuracy, and fun of pen and paper. We've covered some simple topics like where we're from, pets we've owned, and our family structure, so maybe I'll try out easy games like "spot the difference" next.

2026-04-09 71 ¶

Week 4: One month down

Anybody who's worked on learning a second language, be it through CI methods or not, knows that feeling prior to acquisition where there's a delay between hearing something like like "I'm going to drink this water" and actually comprehending the message through visual assistance. I think it was yesterday or today that I could encounter this sentence from Blabla Chinese and understand with very little delay that Amber was about to take a sip of water.

This is pretty different from "knowing" what a sentence means after hearing it. This second category is actually getting noticeably broader by the day for me. There are lots of sentences that with CI visuals I recognize as long as it sticks to simple nouns, verbs, and adjectives, but there's a lag between sound and meaning. If I had to guess, I'm probably about 100 hours away from all the content marked as "beginner" being comprehensible via audio only. The only content I can use as audio-only right now are the baby immersion videos by You Can Chinese, which is about as exciting as standing in an windowless room for an hour holding your thumb in a glass of lukewarm water. But it's comprehensible! That's progress.

2026-03-28 63 ¶

Week 3: Panda faces everywhere

I'm talking about these. They freak me out and they are in so many videos. At this point I've reviewed about 15 YouTube channels from the ALG Level 1 list, and I'm settling into three main contributors.

• Lazy Chinese
• Linguaflow
• (Recent) Blabla Chinese

Older Blabla Chinese videos have a lot of the seizure-inducing flashing visuals and text, so I've been sticking to premium videos posted in last 6 months or so. Linguaflow has long-form let's plays that are easy to set-and-forget when I have time to sit and watch for a while.

Of everything, though, Lazy Chinese is the most ALG-friendly with enough content to get a good number of hours under my belt. The website is constantly getting upgraded, although it's not at the CIJ or Dreaming level yet. What I like here is the focus on ALG basics: pantomime, images, stories. Also the (mostly) lack of AI imagery and flashing or disturbing materials means it's easier for me to focus on the what's being spoken. Some of the earlier beginner/super-beginner videos used AI for visual aids, but the AI usage seems to have worn off recently.

2026-03-22 50 ¶

Week 2: Setting a daily hour goal

The rhythm of Mandarin is becoming familiar and common words are easy to predict, although what I can actively recall is still extremely small. If you asked me to produce "green t-shirt", or "this man is standing" I'm confident I could sound it out in my head, but physically producing the sounds with accuracy would be impossible. I also am completely unaware of any grammatical nuance between simple sentences that can look outwardly the same. In English, for example, I can tell there's a difference between "a man stands" and "this man is standing." I have no idea whether this kind of distinction will show up in Mandarin yet.

Lately I've spent a lot of time assessing how much progress I can realistically make over the next year. I will be traveling to Japan in about 9 months and I want to bring Japanese from the ~1500 hours I currently hold to as close to that 3000 (Dreaming Spanish level 7) mark as I can prior to arriving in Tokyo. I don't want Mandarin to fall by the side either, though.

So how many hours does it take to progress at a decent clip using the ALG approach? At what point does dedicating more hours in a day not matter? I drafted up a little table based on page 2 of the Dreaming Spanish roadmap to answer these questions for myself.

The two important milestones are low fluency at level 6 (2000 hours), and high fluency at level 7(3000 hours).

As a native English speaker, that sweet spot for a language like Mandarin seems to be somewhere around the 3 hour mark. At one hour per day it takes over 8 years to fluency. Much too slow for me. Go up to 2 hours per day and we shave a full four years off the ETA. Just one more hour above that and we finish in around 2.75 years.

It's beyond 4 hours where I'm not sure it becomes worth it. For those with hard deadlines and plenty of free time, getting to level 7 in around a year might look appealing. I'm pursuing this as a hobby, though, so I really just need to balance my interest level against results. If I'm looking to create a daily habit where the finish line is more than a year off anyway, how different is 2.1 years vs 1.6 years really? Moreover, the time to low fluency (level 6) is almost identical, so sitting somewhere in the 2.5-3.5 hour range seems ideal to me.

For now, if I stick with one hour per day of Mandarin, I would just cross the threshold for level 3 (300 hours) at the start of December. Once I'm back from Japan in mid December, I plan to ramp up the Mandarin to match this ~3 hours/day goal. Until then, check out the Japanese page for regular updates on my journey from low functional proficiency to (hopefully) near-native fluency!

2026-03-15 37 ¶

Week 1: Watching Grass Grow

It's easy to forget how difficult this step is. I used to think it was pointless for people to post updates to Reddit and Discord at low hour counts, but the opportunity cost at these early audiovisual stages is painfully apparent. I still have 2963 hours to go, meaning I'm less than 2% of the way there, yet in these 37 hours I could have done any one of:

• Read 6 books in English
• Finish the main story and side quests in Hades 2
• Watched 10-15 movies
• Finished Death Note

Still, in this amount of time with absolute beginner content in Mandarin, only the very most frequent words in simple sentences feel familiar when paired with visual aid.

"This shirt is blue." "He is sitting." "There are birds."

Few, if any, Mandarin words are in my active recall yet. I found this phenomenon of recognizing sound but not being able to recall it incredibly distressing in the early phases of Japanese. Now I realize that's just the normal progression of acquiring words.

1. Noticing
2. Recognizing
3. Automatic, correct usage with context and reminder
4. Active recall without context

Finding high quality content at the super beginner stage has been difficult, to say the least. Very few creators come close to what Yuki at CI Japanese can do with static visuals, drawings, and pantomime. Content that really sticks to the ALG style of clear messages delivered via good audiovisual quality seems to be limited to the last 6 months or so. Prior to this, the field is just littered with videos that use visual aids ranging between distracting and outright disturbing, echoing or poorly recorded audio, and even lack of skill with presentation and markup. Despite all this, though, I'm incredibly grateful that so many people have put the effort into making this niche method of learning Mandarin possible.

Based on YouTube upload timestamps, the bar for beginner CI content has been rising rapidly over the last few months. By far the best channel I've found is Stickynote Chinese. Jun is a textbook example of storytelling via setting, props, and body language. The video and audio quality are excellent, the stories are relatable, and I find staying engaged with her videos effortless. The only issue here is that it's a pretty new channel, so there's not a ton of content yet. I'm eagerly awaiting more.

Until then, a new website called Lengualytics is proving invaluable for hour tracking and finding level-appropriate content. It's still looking pretty beta-ey, but for anyone willing to overlook that I think it's a fantastic resource.

2026-03-09 4 ¶

Day 1!

Today is just exploring the different CI channels on the Mandarin input super sheet from the MandarinALG subreddit.

2026-03-08 0 ¶

Why did I choose to learn Mandarin?

After putting roughly two years and two thousand hours into learning Japanese, I'd like to give Mandarin a try (shocking, I know). There are plenty of stories, opinions, and advice for people who want to take a massive-input approach to Japanese, but I'm not sure I can say the same for Mandarin. Moreover, I took a hybrid approach with Japanese; I study bilingual Anki cards every day and occasionally reference a textbook for grammar points.

For Japanese, my primary motivation was simply a better understanding of media I consume on a daily basis, and, if possible, consuming it entirely in its native language. I don't have any native speakers at my disposal aside from iTalki, and while trips to Japan are nice, they aren't a luxury I can afford regularly enough to say that fluency in speaking would give me a good return on the time invested. If I can keep consuming books, games, and TV in Japanese, then I've hit my goal.

Chinese is a different beast, though. At work, I have more than eighteen native speakers among my coworkers, representing many different regions of China. I want to speak their language with them, even if it's just a limited subset of a standard dialect. As much as possible, I want a better appreciation for the varied cultures they represent without going through an English translation. With that in mind, I'm planning to take an augmented ALG approach to Chinese.

A 2025 survey brought the effectiveness of methods that rely solely on CI into question. Nguyen and Doan argue that an ecology that affords "negotiating meaning" in an individualized way fosters the greatest second language growth. They go on to recommend students play an active role engaging with their second language rather than simply being a vessel for passive intake. In particular, they point to the rise of AI tools like ChatGPT for systems that can adjust their language level and provide immediate feedback at precisely the level an individual learner requires.

My method with Chinese will probably be mostly CI, much like I did for Japanese, but delaying SRS use until it's needed specifically for memorizing Hanzi, and limiting its use mainly to proper nouns. I'm also planning to utilize Crosstalk from day one, since it provides exactly the type of "neuro-ecology" Nguyen et al. describe.

I regret not writing about my comprehension levels as I traveled through the stages of acquiring Japanese. Of course I remember there was a "complete gibberish" stage of Japanese, but it's not really until I was getting help planning a trip to Kyoto with a native speaker in Japanese that I thought to reflect on my progress. So I'll be using this site to keep myself accountable and as a living record for my approach to ALG in a post-LLM world.