If this impacted you - we are hiring at Magnetic (AI doc scanning and workflow automation for CPA firms). Cool technical problems, building a senior, co-located team in SF to have fun and build a great product from scratch
Air travel can be revolutionized with a vacuum-sealed passenger transport system, reducing total airport time to just 10 minutes while maximizing efficiency and safety. Upon arrival, passengers would undergo a rapid preparation process, where an oxygen-rich biofluid is introduced through both mouth and anus, ensuring full oxygenation without the need for active breathing. They would then be gently encased in a flexible, vacuum-sealed transport sheet, immobilizing them completely to eliminate security concerns and expedite boarding.
These passenger sheets would be automatically sorted and loaded onto aircraft via a conveyor system, optimizing cabin space with a precision-stacked configuration. Without the need for traditional seating, aisles, or security screening, aircraft capacity could increase significantly, lowering ticket prices while enhancing operational efficiency. Upon landing, passengers would be swiftly unloaded, revived from stasis, and ready to continue their journey within minutes.
This system would not only eliminate delays caused by baggage handling and TSA procedures but also enhance fuel efficiency by optimizing passenger distribution within the aircraft.
FYI this is a reference to the opening statement of the Unabomber Manifesto "Industrial Society and its Future". Don't think OP meant anything by the distinction of US/humans
> The Industrial Revolution and its consequences have been a disaster for the
human race.
Fairly obvious for those who've spent enough time online, I'd say most people would only recognize that first sentence. The Unabomber Manifesto has become something of a copypasta
Yeah this is a huge problem. AFAICT the "solution" currently is basically to just grind to the point you can understand Peppa Pig or the rough equivalent at ~40%-50% comprehension, and then sentence mine each episode painstakingly using something like Migaku + Anki flashcards until you can watch a brand new Peppa episode at ~80% comprehension. Its painful but after this you really "unlock" content and its a lot smoother and more interesting after that (the stone finally starts rolling _downhill")
I'm making an app to try to help with low level comprehensible input, posted elsewhere in this thread
I’ve been learning Mandarin via Comprehensible Input (CI) for about 9 months and really admire OP’s dedication and consistency. In the first 4-5 months of being truly consistent with ~1hr a day of Anki and Peppa pig I got to around 2,000 words and was able to have a great experience when I traveled to Taiwan, so I can vouch for the core methodology in this post. It’s not “easy”, but it’s definitely the most effective way to learn a foreign language that I know of.
The CI community has come a long way in the last ~5 or so years - the general consensus looks a lot like OP’s methods, which I would summarize as:
Self plug: I’ve been working on a way to generate Mandarin audio comprehensible input using LLMs/TTS models. The idea is that there aren’t many great CI options between 500 words and ~3k-5k words - OP himself mentions that when he started watching Scissor Seven 刺客伍六七 he barely understood anything, which is pretty hard to “push through” without some hardcore willpower. My project https://plusonechinese.com makes Mandarin audio stories that are 85% comprehensible at any level from 400 words all the way to 8k or more words and then auto-imports the audio snippets into SRS flashcards, which makes a CI workflow like this a lot easier to engage with at a lower level and without advanced willpower. Still working on making the content _truly_ interesting, but would love some feedback!
How do you pass the initial language barriers? I'm trying to learn Spanish with Netflix shows and the Language Reactor, but I find it extremely hard to understand spoke Spanish. The characters speak so fast that I couldn't discern individual words or phrases. For a lot of times, they speak as if they are murmuring. When I was learning English, my ability to understand spoken English grew with my ability to read, but in the case of Spanish, I can read a lot more advanced text than being able to listen...
There's tons of free content to keep you occupied for a while, and then there's a paid tier if you want more down the track. After choosing the level that's suitable, I recommend starting with Pablo's videos - he's the founder and I still like his style the most, particularly his older videos before things expanded. Just click on whatever looks interesting.
Thanks! I've been on it for some time. I can understand Beginner level most of time, even with 1.25x or 1.5x for most of the videos, but I find most of Intermediate can be significantly harder to understand. One possibility is that the there are more idiomatic phrases in Intermediate, like the use of reflexive verbs, such as hacerse, levarse, ponerse, irse and etc under different context. They are so different from English and I still need to translate them instead of comprehending them.
Keep at it, consistently doing a bit every day, and you'll get there. I went through all the free beginner videos (at the time) before moving onto intermediate, then did the same before moving onto advanced, and only then was I really able to start watching stuff made for native speakers without feeling too lost. I'm not a CI purist though, I looked up words or grammar points occasionally, but I didn't actively try to study or memorize anything.
this is akin to incomprehensible input and you need it, after listening to a lot of it your brain will start recognizing patterns and decompose things on its own, for now just brute force through as much of it as you can.
I tried but all the content was sooo boring! The only show I could sort of watch was aiqing gongyu back then, because it was a pure rip off of Friends but at least it had some humor, but most of the content on Netflix is pure drama and is really boring to watch.
Your app is good, I'm between HSK1-HSK2 and there are often scenarios I want to practice that aren't covered in the standard dialog options in other popular Mandarin apps, so I see the appeal.
Some feedback:
- It took a while to generate my first session, I almost lost interest before I even knew what the app was.
- I tried to click on the buttons in the splash screen screenshot, mistaking it for the app itself.
- I would have liked to not to have to enter the prompt myself the first time, and just click something pre-made, only because the first time I use an app I have very little patience and just want to see what it actually does, so if it makes me think (or wait a long time) I may churn.
Maybe if you had a default prompt's audio baked into the APK it would feel more snappy (for those who select the default prompt.)
I remember watching haimian baobao (spongebob squarepants). Back then you could download this pptv app and watch a bunch of shows for free, I'm not sure what's the equivalent today.
I wish more people would talk about incomprehensible input. Just feeding input in your brain, like an LLM, so that with time you can detect pattern. Like playing the radio in the background, or watching Chinese movies without understanding much, or watching educational chinese youtube videos while you sleep, etc.
DuChinese is a great app for reading stories at beginner levels of vocabulary. It also supports tap-to-lookup and saving words to flashcards, but unfortunately they don't integrate with Anki, only their own app's system.
DuChinese does have an integration with HackChinese, which is basically a Mandarin only paid version of Anki with a sleek interface. I use it for convenience because I find managing Anki decks too tedious.
Yeah DuChinese is I think the premier Chinese graded reader app right now. We also have the tap-to-lookup and make-flashcard-from-content flows (unfortunately only in our system for now, haven't build flashcard import/export yet). The thing we have that they don't is the ability to generate content about whatever subject you want (which can help make it much more personally interesting)
Also while naming quality resources I should also mention [Pleco](https://www.pleco.com/) - it's _definitely_ the best Chinese dictionary app - highly recommended.
Just to offer you another perspective: In all of the (B2C) companies in which I have worked as a PM (three, across two different industries), I have been closely involved in shaping and controlling the evolution of the product.
My normal month has always consisted of creating anywhere from four to twelve unique proposals (PRDs) for new features for the company to build, with the number created varying based on the size of the company and the amount of process needed as a result. But regardless, it has always looked like: "Our users currently have X problem, and we should build Y to solve it. Y looks like this...". Yes this comes with some coordination and communication challenges, but those are always in service of figuring out what to build next.
I understand not all PM roles are like this, but I do disagree that PM never gets better than "sorting/filtering/decision maturity theatre". I am curious: if you were simply ordering an existing backlog, who was defining the new items to add to the backlog?
My experience was B2B. I was certainly expecting a lot more of the problem identification and solution generation work that you described.
The difficulty was that the owners at the top had unshakeable ideas about what the product should be. That's ok, single minded vision can be good and all that. In my very hands-on sales engineering role I'd make things that my prospects were asking for, put them in the product, they'd buy it, and my prototypes and hacks and tools would end up getting refined, hardened, and supported by the implementation team. It worked well because everything I did had big dollars attached to it (so the org itself didn't mind) and the product advanced in a way that the market wanted.
The problems arose when we put in the product management process - the whole committee/requirement/signoff thing the article describes and I mentioned in my earlier comment. That created a formal bureaucratic mechanism for various players to stop all the ideas I had. Before, they'd be fait accomplis necessary for winning a big deal. Now they were just ideas divorced from value from some guy who should've known his place. Back then I was hopeless at understanding how to get a group of people with different agendas to agree on something (I think it's called 'politics').
> if you were simply ordering an existing backlog, who was defining the new items to add to the backlog?
It's over 15 years ago but what I remember is that there were thousands of semi-structured documents which represented things that, with the new product org, got turned into backlog items (stories, epics). This was part of an attempt to move engineering to a more agile way of working while creating a product function to support it. So it's more that we didn't have a backlog as such, then we stood it up and triaged everything we already had into it. All this was happening while the company was trying to learn about agile. I remember day long meetings where we all tried to argue about whether technical tasks belonged on a backlog, and if not how you could write a user story for remediating a piece of technical debt. Basic stuff but none of us had a clue.
>The problems arose when we put in the product management process - the whole committee/requirement/signoff thing the article describes and I mentioned in my earlier comment. That created a formal bureaucratic mechanism for various players to stop all the ideas I had. Before, they'd be fait accomplis necessary for winning a big deal. Now they were just ideas divorced from value from some guy who should've known his place.
precisely, the introduction of a PM means that the people who actually implement the product are put below pure talkers. You suddenly go from being able to influence the product daily to everything being up for a vote and every vote being overriden by the PM. The PM gets to decide which issues are important enough to be included even though he really doesn't know better 99.9% of the time. 99.9% of the time he didn't talk to customers about it, he just has his own opinion and was put into this artificial leadership position that ruins the fun of the job for everyone else. Oh sure he talked to leadership but that ends up being about how to achieve his own dreams and goals, certainly not to help make the ideas of engineers a reality. I've literally never seen a PM that goes and asks what ideas of the engineers we can help make reality. Just let that sink in for a moment: Isn't that what a product is? Ideas of engineers? I mean it really is. The whole agile movement is one of the strangest gaslighting ventures that human psychology has ever produced. We are told that there is no boss while actively being micro managed issue by issue, hour by hour almost, while literally being asked every single day in the standup when X will be finished.
Is that not a straight up engineering job? A good chunk of "engineer school" is practicing "customer has an issue, design a solution" type problems, in-between all the math.
Customers are rarely able to express their underlying issue. They usually communicate what they think the solution is.
A good product person is able to drill down to the fundamental problem the customer really has and articulate that to engineers.
Also, good product people should be evaluating the customers problem in terms of the whole market.
After all, engineering is a limited resource. So you need to solve the problems that exist for your target market, not just a single customer.
Of course, I’ve seen good engineers who can do this, but it takes time and effort to sift through all of the customers issues and work out which ones to solve. So product managers are focused on this to free up the good engineers to design solutions to those problems.
>Customers are rarely able to express their underlying issue. They usually communicate what they think the solution is.
In my experience the vast majority of PMs do exactly this. Half of my job since I became a staff+ engineer almost a decade ago is taking “solutions” from product managers and trying to figure out what they really want.
A lot of traditional engineering (civil, mechanical, etc) are non-competitive endeavors. For example, if you are supposed to make a bridge, you just design what you think a good bridge to be.
In the modern consumer application, you have to understand business concepts like differentiation. If your product is very good, but it’s strictly worse than another product in every dimension, you don’t get any credit for second place. You actually get close to zero sales because no consumer chooses your product over the alternative, in contrast to a “bad” product that does at least something very well in a niche.
Clearly there’s a difference between building a bridge and getting customers to use your free app, but the vast majority of Engineers aren’t working in big infrastructure projects. Even the ones who do are competing with everyone else who bids on the project.
And in software there are plenty of cases where the 2nd or 3rd place product gets tons of sales even if it’s strictly worse in every category. There are so many things that impact market success that are completely out of the hands of product or engineering.
This doesn’t sound like any university CS degree I know of, apart from maybe a single class actually called “software engineering”, which in many cases isn’t even required
And even if it was, no, engineer personalities are horrible at designing solutions that actually meet customers’ and the business’s needs
The engineering classes I took were full of stories and problems centered around solving erroneous customer reports. A good example is the ice cream story [0], where a customer reports their car malfunctioning when they buy the wrong kind of ice cream, which was used as an introduction to root causing techniques. That specific story is apocryphal and the manufacturer varies, but there's an entire mythology of such stories being used to introduce topics / provide "realistic" problems / do design projects for students.
This sort of stuff is a not-insignificant portion of my job as well, since PMs rarely have the technical skills to know what's feasible as a solution anyway in my experience.
Good Product people don't just build exactly what the customer described as their problem. They understand that sometimes there's a better way to solve it. In "engineering school", you can't tell your professor "I understand what you're actually trying to achieve, so I built X instead of Y because I think it'll be better for you".
Product figures out what to build, and Engineering figures out how.
Sure you can. I’ve literally had classes that worked exactly like that. I had classes where I had to interview and observe actual customers using our product.
What you’re describing is a huge part of engineering. Engineers have been doing this long before Product Managers existed. When I started, software engineers used to just talk directly to customers, domain experts, business analysts, and business leaders.
At “product led companies” the engineers just end up trying to figure out what the product manager actually wants because it’s not really possible in most cases to separate the what from the how.
Having worked with both systems, I don’t think the way it works now is better.
not that this is real "data" or anything, but I have lived in San Francisco (in various parts of the Mission and Soma) since 2018. My vehicles have been stolen 4 times, and recovered all four times. In all cases there was some kind of police report, but no person was ever charged with a crime. In 3 cases they actually had the person in custody and let them go (one time he ran away while being questioned in the stolen vehicle, so maybe that doesn't count, but they told me they couldn't have prosecuted anyway because he was just "in the vehicle, that doens't prove he stole it)
Separately, I have personally witnessed 3 instances of store shoplifting (like filling a whole bag with anything that you can and just walking out) where the police weren't even called because it was useless. They were just yelled at until they casually walked away.
I have watched a car window get smashed and a bag stolen from the car on Haight St. near Ameoba at 1PM on a Saturday. Everyone just watched, no one called the cops from what I could see (I didn't).
Last night I watched two guys use an angle grinder to cut a bike lock and steal a bike on Mission st. I didn't call the police.
I'm probably part of the problem now since I don't even call the cops anymore like I used to, but its hard to feel like you can do anything about it when I have seen them just let people go time and time again.
Did you try this recently with GPT-4? Here's my attempt to replicate your test:
---Prompt---
I show you a set of four magisterial unicorns in a cute pink stable, each of them has a letter painted on one side of their body and a shape painted on the other. Right now you see all four unicorns, two of which are showing the shape side of their body and two of which show the letter side. You see that the unicorns read: 'A', 'S', Triangle, Square
Which unicorns do you need to spin around to test the validity of the following statement:
"if a unicorn shows a consonant on one side of its body, then the opposite side of its body is a triangle?"
---GPT-4---
To test the validity of the statement "if a unicorn shows a consonant on one side of its body, then the opposite side of its body is a triangle," you need to look for potential counterexamples to the rule. The two cases to check are:
* Unicorns with a consonant showing.
* Unicorns with a triangle showing.
You do not need to check unicorns with a vowel (like 'A') or a different shape (like Square), as they don't violate the statement.
So, you need to spin the following unicorns:
The unicorn with the letter 'S' (to check if it has a triangle on the other side)
The unicorn with the Triangle shape (to check if it has a consonant on the other side)
---
So it got the answer wrong, but in a way that I think a lot of humans would (you flip the square unicorn, not the triangle). Still seriously impressive in my book and definitely not falling into incoherence
I used a slight reworking and got a similar response to you with GPT4. I tried to prime it to think through the possibilities by giving it the context:
"This is a difficult problem that many people get wrong. Start by reminding yourself of basic logic rules. Then apply the logic rules to the unicorn situation, considering each unicorn in turn and understanding what it would mean for the rule if the unicorn is turned around. Only after doing that conclude with the unicorns that Tom should turn to have a chance of proving Paul wrong."
I gave it this instruction because of other articles I've read where forcing it to give the answer before the reasoning means it gets it wrong more often. It correctly identified that it should use the contrapositive, but still misapplied it, so I gave it that feedback:
"your third consideration is a misapplication of the contrapositive. Can you try that case again?"
Then it hadn't generated a consideration of the last unicorn (it's possible I was being throttled), so I said:
"Consider Unicorn 4 with the contrapositive rule"
With those extra pieces of guidance it gave the right answer and for the right reasons. While I was hoping for better, this is still a meaningful improvement over GPT3.5s performance on the same prompt - its answer was so muddled I couldn't see how to coach it.
Going from crazy nonsense to wrong, but arguably human level performance (80% of humans are bad at this task), is still a nice improvement. I'll have to give it some of my logic tests and see how it does.
https://bookface.ycombinator.com/company/30776/jobs
reply