Hansen Qian

6.4K posts

Hansen Qian

Hansen Qian

@Hansenq

@streamstraight (YC W23) @lightskis @statusrope @affinity

San Francisco, CA Katılım Kasım 2008
651 Takip Edilen754 Takipçiler
Sabitlenmiş Tweet
Hansen Qian
Hansen Qian@Hansenq·
🗺️ Can LLMs navigate San Francisco purely from memory, without Google Maps? I tested spatial reasoning of the 19 top LLMs. Honestly...they were all pretty bad. Best in cities: o3 (38%) Best on highways: grok-4, o3, o4-mini (62-64%)
Hansen Qian tweet media
English
3
2
25
4.7K
Hansen Qian retweetledi
KC-10 Driver ✈️ 👨‍✈️ B-737 Wrangler
If simultaneous approaches are prohibited, how about departures? SFO uses them, too…see below. If not banning those as well, why not? My theory: because the TCAS is inhibited below a certain altitude. By the time these jets reach that altitude, they are turned away from each other & TCAS doesn’t care. The jets turn away from each other after reaching 400’ (generally we don’t make turns below 400’ unless it’s critical for terrain), but what if one jet makes a mistake & turns the wrong way? It’s happened before…but there is no TCAS RA to report. The jets don’t get any closer on approach vs. departure. It’s essentially the same as the now banned parallel approaches, just in reverse – BUT – there is not a spreadsheet with nuisance TCAS RAs for a bureaucrat to point to. It’s not about safety. It doesn’t actually improve safety. It’s about the appearance of safety & appearing to “Do Something!”. If that’s what you want, fine, but as I said just a few days ago you must understand that everything has a cost. We can reduce arrivals to 6 an hour if that’s what we choose to do…then it’s really safe, with only one jet every 10 minutes. Maybe that’s too extreme, so 12 will work…hopefully you get what I’m saying. There is a line here where it’s unsafe, but defining that based on a spreadsheet of TCAS RAs is a terrible method. Everything has a cost. At SFO, that cost will be high in terms of delays, cancelations & ultimately ticket prices…and you won’t be any safer for it.
Air Safety #OTD by Francisco Cunha@OnDisasters

Most of the public doesn´t know, but once the flights are over, there is a pink slip exchange. I know a guy who built an airline just by winning these races.

English
25
20
177
15.9K
Hansen Qian retweetledi
KC-10 Driver ✈️ 👨‍✈️ B-737 Wrangler
Attention Aviation Journalists (& anyone interested)! I have a primer for you on the SFO “parallel approaches” story. It has the potential to be a big story. I imagine the airlines, and maybe even the city, will fight the FAA’s decision. Big impact. So, here’s a bit of history & a perspective on simultaneous close visual approaches to runways 28L & 28R from a pilot perspective. Summary: these are *not* dangerous approaches. You can find someone who will tell you they are, but I’d bet 95% or more of pilots who’ve flown them will tell you it’s not dangerous. In fact, it’s a fun approach & a treat for passengers w/ the added benefit of increasing the airport’s arrival rate capacity. In essence, these approaches – the “FMS Bridge Visual 28R”, the “Quiet Bridge Visual 28L” & the “Tip Toe Visual 28R” (list not all inclusive) – are used in Visual Meteorological Conditions (VMC) only. Pilots must be able to see a long way to do these approaches. When there are low clouds, fog, rain or something, these are not an option even offered. Visual approaches are not at all unusual at any airport. In many cases, the visual is the preferred method of approach. At SFO, we must be able to see a long way because we must be able to identify our “pairing traffic” & keep them in sight during the entire sequence. We know exactly where the other jet is at all times during this approach. I’ve included a sample of the approach to runway 28R, below. Somewhere around point “TRDOW”, approach control is pointing out your “pairing traffic”. They won’t allow you to get close, or continue the approach until you have them in sight. It can take a couple minutes to visually acquire them, but by “GAROW” or “JANYY” it’s very easy to spot them. If you still can’t spot them, approach won’t let you continue. Keep in mind that the example I included below is *only* for 28R. Runway 28L has a different approach & it’s not depicted here at all. There is space between the runways to begin with…but the 28R approach creates more space between aircraft by bringing you in at an angle…it does not line up with the runway until a very short final. The aircraft on 28L is also looking for you & must keep you in sight, as well. The problem with these approaches is that they used to set off a “TCAS RA”. I assume you know this term. A TCAS RA must be obeyed. So, we obeyed it & did a go around, even though we can see the other jet & know it’s not really a problem. Reports are filed & records kept. At one point, SFO was one of a few places that certain airlines were allowed to switch the TCAS to “TA Only” mode. The RAs were a nuisance rather than a real threat & everyone knew it, so we disabled the RAs. I don’t know of many other instances this was allowed. That may sound extreme, but (the majority of) pilots were happy to have a solution, even if it was only temporary as they tried to find a better fix. They did. They offset the approach track to 28R enough that a jet following it would be below the TCAS inhibition altitude before it got close enough to the other jet to set off an RA. It’s an elegant solution & works very well. Now, these approaches are famous for videos of jets landing simultaneously. That’s cool, but not how it works in practice most of the time. One jet usually ends up behind the other & at nearly the same speed. They are below the required distance for usual separation, but they are separated in trail. Even when it does end up with the jets landing together, it’s not dangerous. I’m speculating, but this decision seems to be someone looking at reports of TCAS RAs over several years, seeing a bunch at SFO & deciding we can’t do this anymore, not understanding that the RAs are not a real danger signal in this case. It’s a spreadsheet decision, not a wisdom decision. It doesn’t actually make anyone safer, but it looks like it does & that’s what matters. The disruptions this decision will cause? Not included on the spreadsheet. Hope that helps!
KC-10 Driver ✈️ 👨‍✈️ B-737 Wrangler tweet mediaKC-10 Driver ✈️ 👨‍✈️ B-737 Wrangler tweet media
English
74
84
713
37.4K
Hansen Qian retweetledi
Matthew Yglesias
Matthew Yglesias@mattyglesias·
There are no perfect models out there but this helps underscore why I'm glad Slow Boring is a for-profit business enterprise.
Matthew Yglesias tweet media
English
8
15
281
29.7K
Hansen Qian
Hansen Qian@Hansenq·
@__venki__ I love the United app but it keeps on freezing randomly for me requiring a force quit. My iPhone 14 isn’t that old!!
English
0
0
1
97
venki
venki@__venki__·
apps on my phone Id consider delightful: - Transit (#1 rn) - United Airlines - Waymo - Amazon (groceries) - Claude - Partiful close but no cigar: - Beeper (it stopped working) - Spotify - Google Home uniting theme: fluent utility that I’m still surprised by
English
4
0
40
2.2K
Hansen Qian retweetledi
dax
dax@thdxr·
they're not evil they're just caught in a flawed line of thinking that's historically very common 1. this is very powerful tech (imo it's over stated out of ego) 2. it should only be in the hands of the most moral people (like us) every single person who's ever had this line of thinking ultimately became corrupt there's no larger corrupting force than believing your team is morally superior, does not matter how good of a person you start out as you incrementally justify bigger and bigger transgressions until you're in a pretty dark place infinite examples from history that demonstrate this
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex

Apparently you get bullied and rejected from Anthropic for being insufficiently woke on risks of open source AI. Experience with Anth HR interview. 1point3acres, translated from Chinese. People were asking sometimes why I dislike Dario Amodei. I hope this clarifies matters.

English
35
71
864
70.3K
Hansen Qian retweetledi
Alex Johnson
Alex Johnson@AlexH_Johnson·
In 1977, after Citi spent $50 million to install ATMs in a bunch of its branches, people thought the idea that ATMs would replace bank tellers was absurd. Lines in Citi branches were longer, as customers avoided the new machines queued up to talk to a human. They were wrong. Between 1980 and 2000, the number of ATMs installed in U.S. bank branches exploded and usage soared, as consumers realized how convenient they are for routine tasks. Market observers changed their minds and predicted the imminent death of the bank teller as a profession. They were wrong. By 2010, economists had started to notice that the number of bank tellers had actually risen over prior three decades. Their explanation was that ATMs were a complementary technology that freed up tellers to focus on higher-value relationship banking tasks and that bank tellers would be safe. They were wrong. Since 2010, the total number of bank tellers in the U.S. has fallen off a cliff. Market observers now explain the growth of the bank teller profession between 1980 and 2010 as a function of deregulation (interstate banking, specifically) which drove a massive increase in the total number of branches in the U.S. (even though each branch could get by with fewer tellers thanks to ATMs). These observers expect that smartphones, which ushered in a completely different competitive paradigm in financial services in the 2010s and 2020s, will eventually lead to the death of the bank teller as a profession. Maybe! Or, as has happened repeatedly in this story, maybe they're wrong! In 2024, JPMorgan Chase — a company that fully understands the potential of ATMs and mobile banking and that, I promise you, is not soft-hearted when it comes to the topic of automation and job destruction — embarked on a plan to open more than 500 new branches, renovate 1,700 existing branches, and hire 3,500 employees to work in them. I share this story because, at a time when everyone is rightly worried about the impact that AI will have on the job market, it's important to remember that old William Goldman line (which @DKThomp recently introduced into the AI/jobs debate): Nobody knows anything.
English
19
57
419
101.5K
Hansen Qian retweetledi
Hansen Qian retweetledi
Cheng Lou
Cheng Lou@_chenglou·
My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow
English
1.3K
8.1K
63.9K
22.4M
Hansen Qian retweetledi
Casey Handmer
Casey Handmer@CJHandmer·
SoCal has the best climate on Earth and for the last 15 years invasive mosquitos have been spreading, making it very difficult to actually be outside, risking disease. Official guidance is "err, eliminate standing water". It's 2026. We have crispr, gene drive, and a local pest control budget that's high though to run a space program. Who do I have to fuck to get these mosquitoes extincted? They aren't native, they're displacing local species, they spread disease, they don't have rights. Kill them all!
English
205
58
2.2K
179.5K
Hansen Qian retweetledi
StripMallGuy
StripMallGuy@realEstateTrent·
Health insurance for a family of five can easily cost $30,000–$40,000 per year. So -- If you run a small business and want to recruit an employee who expects fully paid health insurance, that’s a really big number. If that same candidate happens to have a spouse whose employer covers the family's health insurance, it eliminates that cost for you and allows you to make a competitive offer. You literally save $40k just based on what that candidate's situation happens to be. Such a crazy system in so many ways.
English
99
34
964
87.6K
Hansen Qian retweetledi
Andrew Weiss
Andrew Weiss@BayWestInvest·
One more time for the people (and mayorial candidates) in the cheap seats: The cities with the MOST amount of renter protections are the LEAST affordable. The cities with the LEAST amount of renter protections are the MOST affordable. By allowing a subset of renters to stay in their apartment for months on end without paying rent, you create a scarcity of vacant apartments. Or to put it more succintly... we should do what these cities are doing 👇
Andrew Weiss tweet media
Nithya Raman@nithyavraman

We’ve already reduced rent increases and strengthened tenant protections. This issue is that they are not being enforced. As mayor, I will create a new Office of Tenant Protections to stop violations of existing renter’s rights.

English
13
92
520
20.1K
Hansen Qian retweetledi
yatharth ༺༒༻
yatharth ༺༒༻@AskYatharth·
me: "i visited stanford today. it was beautiful" @devonzuegel: "it's because they don't have cars" me: "sf has beautiful parks too, but" devon: "cars" me: "something about the campus just felt… human-sized" devon: "they don't have cars" me: "even though it has taller buildings than sf. isn't that interesting?" devon: "ca-" me: "i guess you can't explain these things" devon: ... me: "a mystery…"
English
12
4
201
13.9K
Hansen Qian retweetledi
Lee Edwards
Lee Edwards@terronk·
The story here should actually be that 3 school board members voted against middle school algebra. This is after a BoS resolution supporting middle school algebra, and support from the Mayor. SF is still on the knife’s edge of clowns vs normal people. Don’t sleep on this.
Katie Miller@KatieMiller

Twelve years ago, San Francisco killed 8th-grade Algebra in the name of equity. Advanced math enrollment (including AP Calculus) dropped, and racial gaps expanded. China graduates ~1.3 million engineers yearly vs. just 130,000 in the U.S. Last night, the SF school board finally voted 4-3 to bring math back. When DEI overrides academic standards, America falls behind. nytimes.com/2026/03/24/us/…

English
7
14
206
17.4K
Hansen Qian retweetledi
Udi Wertheimer
Udi Wertheimer@udiWertheimer·
everyone is wrong about why sora is cancelled sora 2 is last-gen tech. seedance2 is 10x better and will be rolled out in the US in the next few days trained on tiktok which has more video data than anyone else on earth. openai can’t compete
Sora@soraofficialapp

We’re saying goodbye to the Sora app. To everyone who created with Sora, shared it, and built community around it: thank you. What you made with Sora mattered, and we know this news is disappointing. We’ll share more soon, including timelines for the app and API and details on preserving your work. – The Sora Team

English
70
24
489
58.3K