Zu Jian Lee (@JeanLuize) header
Zu Jian Lee (@JeanLuize) profile picture

Engineer who likes music, basketball and anime. I like thinking about things

Singapore
Joined September 2013
Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)Photo by Zu Jian Lee (@JeanLuize)
Zu Jian Lee retweeted
Real life, baby
19,197
96,266
45,751
773,963
99,428,902
Zu Jian Lee retweeted
shooting free throws in NBA 2K5 used to have people stressed 🎮😂 (2004)
29
153
101
1,677
2,730,201
Zu Jian Lee retweeted
Just came across the cutest ancient relic in Turpan, Xinjiang: a 5-meter-long 1300-year-old homework scroll from a 12-year-old boy in the Tang Dynasty (around 710 AD). His name was Bu Tianshou, and apparently Tang dynasty tiger moms were already a thing. Poor kid had to copy and interpret 178 lines of the Analects. But being a normal 12-year-old, he couldn’t resist sneaking in some complaints at the end. First poem: “Teacher, I finally finished today — please don’t roast me for being late. Tomorrow’s a holiday, so can you let us out early?” 写书今日了,先生莫嫌迟。明朝是假日,早放学生归。 Second one: “Everyone says sideways script is easy. I say it’s brutal.
You gotta tilt the paper and squint just to read the damn thing.” 他道侧书易,我道侧书难。侧书还侧读,还须侧眼看。 Scholars are still arguing over what “sideways script” actually means, but we all know what it really means: ancient Chinese homework was suffering too. And the wild part? For a kid who clearly didn’t want to be there, his calligraphy is actually pretty impressive. 1300 years later and kids still haven’t changed. Homework sucks in every era🤣
24
352
20
1,743
71,804
FSD Supervised is now available in Australia and New Zealand
652
848
306
5,436
13,894,569
Zu Jian Lee retweeted
Steve Nash in Games 4-6 of the 2005 WCSF 40.3 PPG 9.0 RPG 9.7 APG 1.3 SPG 60/58/100 Shooting 71.1% TS (+19.2 rTS%)
Replying to @pitlessball
Steve Nash 2005 WCSF bag work. Dude averaged 40 10 and 9 on 60-57-100 splits in games 4-6 to knock out the team that let him go the offseason prior.
69
605
184
4,450
748,893
Zu Jian Lee retweeted
Robotaxi app now available to all Download to join waitlist – expanding access soon
1,111
1,643
624
10,236
9,267,497
Zu Jian Lee retweeted
Avg price for a cup of coffee in the US is $3.50 Your car drives itself with FSD Supervised for ~$3.33 a day
824
840
108
7,288
730,825
Zu Jian Lee retweeted
Empty bottle caught in aerodynamic drag. As the truck moves, it creates a low-pressure volume of space behind it — roughly shaped like a 3D teardrop — inside of which is less air pressure, which pulls the bottle in a perpetual spin behind the truck.
51
264
20
2,524
133,101
🚨 PARALYZED MAN STANDS AGAIN AFTER STEM CELL BREAKTHROUGH IN JAPAN Japanese scientists have achieved what was once thought impossible: a paralyzed man can now stand and is relearning to walk after receiving an injection of lab-grown stem cells. The team at Keio University transplanted neural precursor cells into the spinal cords of four men. Two regained significant function—one now stands unaided, another moves arms and legs. Professor Hideyuki Okano: "This is a dramatic recovery." Source: ZME Science.
414
3,958
316
26,276
1,580,004
Zu Jian Lee retweeted
15,980
37,474
2,753
461,910
84,055,286
The chicken has figured out how her feeder works Now closes it in time so the alpacas don't eat all the food.
1,469
13,093
1,546
131,336
7,480,074
Next time I take a flight I will make sure my plane isn't Boeing 737-800.
763
11,428
2,406
55,148
8,907,218
Zu Jian Lee retweeted
Tell a friend
V13 is MINDBLOWING. Here’s the first drive video with FSD v13.2.1. 31 mins from parking lot to parking spot with ZERO disengagements or interventions. I can’t believe how smooth it is- it’s incredible. This is it folks, it’s here. @elonmusk @teslasi
880
1,400
153
10,047
2,820,574
Zu Jian Lee retweeted
FIRST IMPRESSIONS WITH FSD v13.2.1 Summary I approached my first drives with FSD v13.2.1 with pretty high expectations, especially after logging over 2,000 miles with FSD v12.5.6.4, which only had two disengagements. Tonight, I drove about 70 miles with the new version, and I came away even more impressed than I was with v12.5.6.3 and .4! The Route Our route took us from my hotel near LAX to The Stand restaurant in Calabasas and back. We decided to venture up into the hills to test how 13.2.1 would handle the twisty roads, along with other challenges like road construction, lane closures, and various other conditions. Traffic & Road Conditions The roads were fairly busy tonight which meant more traffic and even more erratic LA drivers sharing the road, especially near LAX. As I mentioned above, we encountered a lane closure up in the hills near Calabasas which made things even more interesting. What's New & General Commentary - CRAZY HUMAN DRIVERS: 13.2.1 did a fantastic job holding it's own by keeping its composure, even when it was cut off numerous times by surrounding human drivers. No excessive slowing, no phantom braking, just smooth yet calm & collected throttle / brake control. - PRECISE CURB HANDLING: One of my big gripes with 12.5.6.x on city streets was that it would sometimes get a little close to curbs and then not correct it's trajectory. Tonight, I noticed several instances with 13.2.1 where it would correct itself if it thought it was getting too close to a curb. It did so smoothly, without any seesawing on the wheel. As someone who doesn't want to rash any wheel, this behavior made me more confident. It mimicked what a human driver would do in similar situations while navigating tight turns. - SPEED PROFILES FOR ALL SPEED RANGES: Having speed profiles at all speeds made for a smoother experience all around. FSD did seem to do a better job at accelerating up to the speed limit on city streets compared to before however I still think there is room for improvement here (especially on city streets where the speed limit exceeds 40 mph). - STANDARD SPEED PROFILE UPDATES: I drove solely with the "Standard" FSD profile which seemed noticeably more assertive than what I was used to with 12.5.6.3 & .4. By this I mean it was more eager to pass slower traffic and increase its speed >5mph over the limit to better fit in with surrounding traffic. I never felt the lane changes were excessive and instead almost perfectly mirrored how I would have driven the vehicle myself. Unlike with v12.5.6.3 & .4 I never felt the need to switch to the "Hurry" speed profile. - USE OF PASSING LANE: FSD 13.2.1 did exhibit some similar behavior to what I was used to with 12.5.6.x in that it would still stay in the leftmost lane a bit too long, even with traffic directly behind us. I would like to see more work done to increase FSD's awareness of traffic approaching from the rear. Additionally, it should do a better job at proactively moving out of the passing lane when it is safe to do so. Keep in mind this critique is focused on the "Standard" speed profile as that is what I used for tonight's drives. - SLOWING AROUND SEMI TRUCKS: One issue I repeatedly encountered on my trip from Minneapolis to Los Angeles was that 12.5.6.4 (and .3) would noticeably reduce speed when coming alongside Semi Trucks, regardless of the speed differential. This unnecessarily increased the time we spent next to the Semi, which I felt put us more at risk compared to maintaining our speed and quickly passing them. I encountered a few Semi Trucks tonight with 13.2.1, but it wasn't enough to conclude this issue has been resolved just yet. More testing is needed on this one. - PAUSING AT GREEN LIGHTS: At the start of my first drive with 13.2.1 tonight on city streets, there was one occurrence where the light changed and the intersection seemed clear, yet FSD refused to proceed. After waiting a few seconds with no movement from the car, I intervened by pressing the throttle. I reported this incident, hoping the Tesla AI team can investigate what the car might have perceived. - SMOOTHNESS: Overall 13.2.1 seemed even more smooth with brake and throttle inputs compared to v12.5.6.x, especially when decelerating after exiting the freeway. As you may recall, 12.5.6.x would often come in a little hot then have to use a bunch of friction brake to complete the stop. With 13.2.1, the car seemed to almost always rely solely on regenerative braking, which was much smoother. More braking input was applied earlier on the exit ramp, resulting in a smoother experience. - CONSTRUCTION & LANE CLOSURE: As we ascended towards the Top of Topanga Overlook, the right lane was closed, forcing traffic to alternate using the left lane. A construction worker was there, managing traffic with a stop/slow sign. FSD not only detected the hand held stop sign, but immediately adjusted its speed when the worker flipped the sign to "slow". The latency in its decision-making was almost real-time. I was certain was going to have to disengage, but when FSD managed the situation flawlessly, I was genuinely speechless. For the entire journey to the overlook, FSD navigated the single open lane without attempting to veer into the closed lane, correctly following the cones. Seriously impressive stuff! - VOICE COMMANDS: This isn't directly related to FSD but pertains to Tesla OS v2024.39, which seems to include some enhancements to voice commands. Common commands like "set ride height to medium" which were often missed, are now recognized on the first attempt. Additionally, the transcription speed has noticeably improved, happening almost instantly compared to the previous awkward pauses. I hope this is a step towards integrating @grok into our Teslas, and if so, it's an encouraging beginning. Closing Thoughts It is now apparent that driving manually is no longer a necessity when you get into a Tesla equipped with FSD v13. Even long-time Tesla owners familiar with FSD Beta v10 & v11 might underestimate how capable v13 has become and how closely it resembles a safe human driver. We're at a stage where we're fine tuning comfort levels and seeing hundreds of miles between disengagements. While my critiques aim to further enhance FSD, I believe this version is ready for a broader release, and I hope the Tesla AI team will roll it out to the wider AI4-equipped fleet soon. Great work @teslasi team, you all never cease to amaze me! @elonmusk @aelluswamy @teslasi @tesla @tesla_na @pduan @YunTaTsai1 @KalenaKBrown @ArekSredzki @TeslaOwnersMN @wholemars
263
401
109
3,375
18,995,004
Zu Jian Lee retweeted
Recycling, Incineration, or Landfill? The traditional ‘waste hierarchy’ says recycling is best, followed by incineration, with landfill as the last resort.  However, I’m not too sure about this. I already discussed the problems of recycling and why it hardly happens. Incineration burns trash for energy. Since plastic is made from oil, it’s essentially a roundabout way of burning oil. Air pollution can be minimal with modern incinerators (thanks to high temps + filters). CO₂ emissions are arguably not an issue as it substitutes regular fossil fuel power generation. Landfills don’t provide further use for plastic. But they are a carbon sink (you’re putting the carbon back into the ground) and, when properly designed, are a simple way of safely containing the trash away from the biosphere.  My read of all this is that recycling, incineration and landfilling are all fine — the key thing is to make sure the plastic does not end up in the environment! This is where landfilling has an edge, because it’s affordable in places where most plastic is leaking into the environment. We should IMO care way more about plastic not being collected at all in poorer countries than about plastic ending up on landfills in richer ones. The step from dumping to landfills is much more important than from landfill to recycling.
Is Recycling a Scam? Yes and No. Recycling is usually the best thing you can do with plastic waste. Every ton of plastic you recycle is a ton of new plastic you don’t have to produce from oil. The problem is: it hardly happens. Why? Because the economics don’t work. Plastic is so cheap to manufacture that it’s often cheaper to make new plastic than to collect, sort, clean, and recycle existing plastic. There are two kinds of plastic recycling: mechanical and chemical. Mechanical recycling essentially involves remelting plastic. However, plastic degrades over time, and contaminants build up, meaning you can only recycle it a limited number of times. Even this relatively simple method is bad business—it’s only marginally profitable when oil prices are high. Chemical recycling, on the other hand, uses solvents to break the plastic down to its building blocks, removing contaminants and circumventing degradation problems. This allows the plastic to be recycled infinitely without any loss in quality. However, it is far more expensive than mechanical recycling, so once again, the economics don’t add up. This is why today only 9-13% of plastic is recycled. Unsurprisingly, most of this is PET—a more valuable type of plastic. This stands in stark contrast to many metals. Mining and refining metals require a lot of energy, which acts as a ‘natural tax’ on producing new metals. This makes recycling metals financially viable, which is why it happens much more frequently. Unless the fundamental economics change, plastic recycling rates will remain low in the future.
30
68
9
374
28,487
Zu Jian Lee retweeted
Model Y is the best-selling electric car for the last 12 months in Harbin, one of the coldest cities in China famous for its icy winters and stunning snow landscapes❄️
70
210
46
1,952
157,052
Zu Jian Lee retweeted
FSD Supervised avoids wildlife collision
When SECONDS matter… FSD matters. Would you SEE in time? Could you STOP in time? Tesla FSD 12.5.4.2. @Tesla @elonmusk @aelluswamy
936
1,698
145
14,173
3,841,941
Zu Jian Lee retweeted
If you're a fan of Hayao Miyazaki and Studio Ghibli and saw Nausicaa of the Valley of the Wind, you still may not know her glider, the Mehve, has been recreated in real life by artist Kazuhiko Hachiya. It's called M-02J and has a 963 cm wingspan.
29
594
35
5,444
170,399
Yes, obviously 23% is not a new record. We have had higher efficiency cells for a long time. The difference is that 23% efficient modules are now cheap. 10 years ago, 16% was standard. It wasn’t that we couldn’t make high efficiency cells, we just couldn’t make them affordably. You can buy a 700 W panel today for less than a 250 W panel 10 years ago. That panel is both larger and more powerful per unit area. Efficiencies are going up because we’re getting better at making higher efficiency cells cheaply. We’ve gotten so much better that now you save money overall with higher efficiency cells because you save on other components. We’re phasing out older tech and expanding manufacturing of new tech. Polycrystalline cells have been basically pushed out of the market by monocrystalline PERC and TOPCon. N-type silicon is growing. Back contact cells are just starting. There’s plenty of tech in the pipeline that will push further improvements in power density, cost, and longevity.
I don't think people realize how much power you can generate with a single solar panel. You can get panels that are over 700 watts now, and that's before bifacial gain. Cell efficiencies have increased substantially and panels are getting larger. More power per panel means lower balance of system costs. Bifacial glass/glass modules and n-type silicon are a big deal for both power and longevity. The panel I'm using for this example is warrantied to 87% of its initial output after 30 years. After 30 years, it will still be more efficient than a brand new panel from just a few years ago. Higher power and better longevity mean a lot more lifetime production. A single panel can be a life-changing amount of electricity.
7
62
3
432
34,518
Zu Jian Lee retweeted
INTRODUCING THE UNIVERSAL REAL TIME TRANSLATION TOOL. REVOLUTIONIZING THE GLOBAL CONTENT CREATION SPACE, FOREVER. SPEAK ANY LANGUAGE, WITH ANYBODY, ANYWHERE.
7,234
56,396
9,137
320,095
20,059,525