Infoman

Member since 2017-07-15T03:50:57Z. Last seen 2025-01-12T00:12:02Z.

2732 blog posts. 128 comments.

Followers: 0 Following: 2

Posts by Infoman

• 
  

  Infoman

  周一美股表現先高後低，美國總統特朗普於周日向中國示好，以及美國財長努欽同日說中美爆發貿易戰的可能性不大，緩和了投資者緊張情緒，令美股周一愈升愈有，道指一度飆440點。然而，收市前聯邦調查局（FBI）搜查特朗普私人律師科恩（Michael Cohen）辦公室的消息傳出後，美股三大指數隨即拾級向下，更以貼近全日最低位收市，道指最終僅升46點（0.19%）；標普500指數由最多漲1.88%縮窄至0.33%；納指由上升2.31%減至僅漲0.51%。

  過去一個多月來，政治消息充斥市場，不論中美貿易戰還是「通俄門」事件，一直左右美股及港股的走勢。當政治因素逐漸淡化後，市場焦點將重回企業基本因素。

  昨天提到美股市場焦點，將在本周稍後時間轉移至企業季度業績上，一眾大型銀行如摩根大通、花旗及富國等本周陸續派發首季成績表。市場研究機構Bespoke整合分析師對標普1500指數成分股的盈利預測，把獲分析師上調盈利預估的企業數目減去被下調盈利預估的企業數目，再除以相關企業總數，所得「淨盈利修訂比率」用來推測美股整體業績期的表現。

  按過去4星期美股分析師的預估，標普1500指數的淨盈利修訂比率由上季的+27.2%跌至-1.3%，意味目前被下調盈利預估的企業數目多於獲上調的企業。

  不過，值得注意的是，淨盈利修訂比率出現負數未必是壞事。如【圖】所示，從2009年至2017年間，統計標普500指數在業績期的表現，逾八成時間都與淨盈利修訂比率的方向背馳，即該比率出現負數時，標普500指數於業績期內多數報升，反之亦然。例如對上一季淨盈利修訂比率急升至+27.2%，標指就錄得0.4%跌幅。此外，於這36個季度中，只有7個季度出現例外（圖中紅點）。淨盈利修訂比率出現負數時，某程度反映分析員下調對企業盈利的預測水平，這意味企業實質業績表現較容易beat贏分析員預測（即好過市場預期），因而對股價產生刺激作用。

  總括來說，當市場焦點重新聚焦企業基本因素後，美股盈利增長將主導後市，尤其目前市場對企業盈利增長期望較低，美股較容易「意外地」造好。

  雖然周一美股升幅幾近全數蒸發，卻無礙恒生指數連續3個交易日上揚。恒指昨早低開24點後，早段一度倒跌66點，見30163點全日低，其後反覆走高，中午收市前成功突破100天線（約30515點）。港股午後走勢續強，A股高收後，恒指最多飆577點，高見30806點，逼近20天線（約30824點），最終收報30728點，大漲499點（1.65%）。主板成交額增至1383億元，上升股份佔59%。

  技術走勢上，恒指周初重新進入長方形形態後，料再度進入橫行整固的局面。現水平已突破100天線，下一個較大阻力將會是50天線，約31081點，這亦相當於2017年上升通道的底線。若能升穿此底線，後市將會較為樂觀，這期間宜觀察多項大市指標走勢，尤其強勢股買盤有否配合增強，作為指數反彈持久力的參考。

  2018/04/11
• 
  

  Infoman

  周一美股表現先高後低，美國總統特朗普於周日向中國示好，以及美國財長努欽同日說中美爆發貿易戰的可能性不大，緩和了投資者緊張情緒，令美股周一愈升愈有，道指一度飆440點。然而，收市前聯邦調查局（FBI）搜查特朗普私人律師科恩（Michael Cohen）辦公室的消息傳出後，美股三大指數隨即拾級向下，更以貼近全日最低位收市，道指最終僅升46點（0.19%）；標普500指數由最多漲1.88%縮窄至0.33%；納指由上升2.31%減至僅漲0.51%。

  過去一個多月來，政治消息充斥市場，不論中美貿易戰還是「通俄門」事件，一直左右美股及港股的走勢。當政治因素逐漸淡化後，市場焦點將重回企業基本因素。

  昨天提到美股市場焦點，將在本周稍後時間轉移至企業季度業績上，一眾大型銀行如摩根大通、花旗及富國等本周陸續派發首季成績表。市場研究機構Bespoke整合分析師對標普1500指數成分股的盈利預測，把獲分析師上調盈利預估的企業數目減去被下調盈利預估的企業數目，再除以相關企業總數，所得「淨盈利修訂比率」用來推測美股整體業績期的表現。

  按過去4星期美股分析師的預估，標普1500指數的淨盈利修訂比率由上季的+27.2%跌至-1.3%，意味目前被下調盈利預估的企業數目多於獲上調的企業。

  不過，值得注意的是，淨盈利修訂比率出現負數未必是壞事。如【圖】所示，從2009年至2017年間，統計標普500指數在業績期的表現，逾八成時間都與淨盈利修訂比率的方向背馳，即該比率出現負數時，標普500指數於業績期內多數報升，反之亦然。例如對上一季淨盈利修訂比率急升至+27.2%，標指就錄得0.4%跌幅。此外，於這36個季度中，只有7個季度出現例外（圖中紅點）。淨盈利修訂比率出現負數時，某程度反映分析員下調對企業盈利的預測水平，這意味企業實質業績表現較容易beat贏分析員預測（即好過市場預期），因而對股價產生刺激作用。

  總括來說，當市場焦點重新聚焦企業基本因素後，美股盈利增長將主導後市，尤其目前市場對企業盈利增長期望較低，美股較容易「意外地」造好。

  雖然周一美股升幅幾近全數蒸發，卻無礙恒生指數連續3個交易日上揚。恒指昨早低開24點後，早段一度倒跌66點，見30163點全日低，其後反覆走高，中午收市前成功突破100天線（約30515點）。港股午後走勢續強，A股高收後，恒指最多飆577點，高見30806點，逼近20天線（約30824點），最終收報30728點，大漲499點（1.65%）。主板成交額增至1383億元，上升股份佔59%。

  技術走勢上，恒指周初重新進入長方形形態後，料再度進入橫行整固的局面。現水平已突破100天線，下一個較大阻力將會是50天線，約31081點，這亦相當於2017年上升通道的底線。若能升穿此底線，後市將會較為樂觀，這期間宜觀察多項大市指標走勢，尤其強勢股買盤有否配合增強，作為指數反彈持久力的參考。

  2018/04/11
• 
  

  Infoman

  Lessons on Bubbles from Bitcoin

  46 kgwgk 8 hrs 13

  https://www.bloomberg.com/view/articles/2018-04-10/bitcoin-was-prone-to-bubbles-until-bears-could-bet-against-it

  news.ycombinator.com/item?id=16805125

  Unable to load the content https://www.bloomberg.com/view/articles/2018-04-10/bitcoin-was-prone-to-bubbles-until-bears-could-bet-against-it

  2018/04/11
• 
  

  Infoman

  Why I’m Using a 30 Year Old Development Method in 2018

  https://news.ycombinator.com/item?id=16787247

  Making games is hard.

  And by far the hardest part of making games is pre-production. That statement may be confusing. We’ve all heard about very difficult production periods for games, and we’ve also observed light, easy, fun pre-production periods. So why am I claiming that pre production is harder? Because one of the things that can poison production is that pre-production is happening during it. As hard as pre-production is, it’s even harder (and way more expensive) to do while you’re also in production. Allow me to explain: In an ideal world, you’d never go into production on a commercial game that’s going to fail. If you were making a game with the intent of turning a profit, and you knew that game would not do so, you wouldn’t go into production. In spite of the obviousness of that statement, many projects, including projects of which I have been a part, do not see their look, mechanics, sound, and level design come together until the game is almost finished. The tail end of production is a terrible time to find out whether or not a game works. Frustrating or underwhelming mechanics will need to be thrown out or reworked, which will have a hugely expensive cascading effect on the technology, art, and sound. This synthesis of all the parts (aka first playable) is also the first time you find out whether or not your team is up to the original task. This is less than ideal if they’ve already been working on it for years. “But game development is inherently chaotic,” the strawman in my head complains. Well he’s right. It is. Innovation requires time, talent, discarded work and endless iteration. How do we accommodate that chaos and allow for innovation without risking the whole farm? Remember when I said pre-production is harder than production? The reason so many projects have such nightmarish development is that the pre-production was never truly finished. They may have put together some concept art, a few documents, maybe even a white box prototype. But doing this is a far distant thing from completing pre-production, at least according to Mark Cerny. In Cerny Method, by the end of pre-production, you should have one level of your game that is indistinguishable from a completed professional product. And there can be no cut corners. A demo is not a first playable. An alpha is not a first playable. A whitebox is absolutely not a first playable. The art, animation, mechanics and sound must all be working together, they must be great, and together they must represent something which will outsell competition 2 years from now. But Why?

  So why bother making a first playable if it’s so much work? Because according to Cerny, “It’s the only way to give yourself a proper chance of making a good game.” Cerny Method isn’t just about keeping chaos out of production; it’s about keeping conservatism and arbitrary schedules out of pre-production. Games require innovation and taking chances, and you won’t want to do that once the production train is rolling. By first deliberately setting aside time for chaotic, rapid iteration, you empower yourself and your team to come up with interesting new ideas. You are free to change things on a whim and try ideas that have never been done before. Nothing is too dangerous, too weird, too unfaithful to the genre, because you’re not counting on taking any of the things you make into production in their current form. In fact you should proceed as though you’re going to throw it all out. Then, as the game’s concept starts to congeal, you take your best ideas and give them the most polished presentation you possibly can. Once you have endured the flames of pre-production, you know so many things about your project that most developers don’t get to learn until years into production. You know what works, what doesn’t, what’s fun, and what makes players hate you. You know how the art, animation, and sound mesh with and support the mechanics. You know all the things your avatar can do, and therefore what the levels, obstacles and/or enemies are like. And perhaps best of all, because you’ve been iterating on so many levels, you know how long it takes to make them which means you can actually make an accurate schedule and budget. You are now free to build your game, armed with the knowledge of exactly how to do it, how long it will take, and what it will cost. But there’s one giant caveat emptor: Caesarthumbsdown.gif

  The linchpin of Cerny Method is this: not every game will survive pre-production. If you have made 5 level prototypes, and you don’t have that publishable first playable, that lightning in a bottle that will outsell the competition 2 years from now, you have kill the project. “There’s no point in following this process if you’re not going to old the output of pre-production to a very high standard.” It sounds harsh, but remember the entire point of Cerny Method is to ensure that you don’t find yourself in the situations I described earlier, where you’re in full-blown production mode on a doomed game that will never work. Skipping pre-production is like launching a rocket and then trying to decide where it should land and figure out its trajectory and how much fuel it needs, all while it’s already in the air. When I look back on my failed projects, there’s absolutely no question that I’d much rather have cancelled after 6 months of pre-production than 3 years of production. Easy Street

  But if you do reach the end of pre-production, you have that publishable quality first playable, your game is incredibly great, then congratulations! From here the work is easy; all you have to do is make the game. As Cerny puts it, The publishable first playable proves that you understand precisely what your game is and what it is not, and that you know enough to produce the game. It delineates pre-production, where you know nothing, from production, where you know everything. I’m being flippant of course, there’s nothing easy about making a game. But imagine being able to make levels knowing with 100% certainty that your character’s jump distance isn’t going to change. Imagine being able to look at the overarching design of your game as a single-page spreadsheet. Imagine not needing to constantly re-write documentation, throw out production-quality assets, and have design arguments while a release deadline is looming. I don’t think everybody needs to use Cerny Method. It’s outdated for a reason, plenty of teams are perfectly capable of using Agile or another iterative method. But I’m spearheading a game of my own for the very first time, and I know I’m not the only developer doing so A studio is far more likely to be able to dust itself off after a failure than an indie mortgaging their home. If you’re thinking of putting your own money into a game, for god’s sake do pre-production. Don’t wait until 90% of the production budget’s gone to test a concept Limitations

  There are limitations to following Cerny Method to the letter. One of them is that once you transition from pre-production to production, your mindset has to switch from boundless creative openness to staunch conservatism. And you can see this play out in Cerny games. The later levels of Crash 2 don’t have much going on that isn’t present in some form in the early levels. The difficulty increases, and the obstacles are re-combined in new ways but the fundamentals remain more or less the same. Crash offsets this with what Cerny calls a “special mechanic”. For Crash 1, the special mechanic was riding a wild boar. But the special mechanic cannot be an after thought in Cerny Method; it must be iterated and tested during pre-production just like the base mechanics. Crash 3 is the most varied in the series, giving Crash a new ability after each boss. Must of the uncertainty is mitigated by making these powers variations on existing abilities. The only truly new ability is a bazooka. It is also the most cumbersome mechanic in the game, and does not show the same refinement as the other mechanics. The bazooka is an example of what can happen when you aren’t conservative enough during production. I dearly love games like Portal 2, Rayman Legends, Psychonauts, and Antichamber, where each new area seems to offer more innovation than the last. Games like these probably couldn’t be made in Cerny Method because it’s not possible to separate their production and pre-production in that way. Games like these require constant (CONSTANT) iteration and playtesting of near-production-quality levels. Otherwise every new level will feel like an untested prototype. Ubisoft and Valve have the resources to do this, and I do not. Antichamber’s development doesn’t bear comparison because it took so long to make and because the art is so minimal (brilliant, but minimal.) I’m making a game with a tiny team and very limited resources. I don’t want it to take ten years and I don’t want to reach the end of development only to learn that I was doomed a month in. That’s why I’m using Cerny Method.

  2018/04/09
• 
  

  Infoman

  About rust

  Rust is a language designed for writing highly reliable and fast software in a simple way. It can be used from high-level code down to hardware-specific code, and from big irons to tiny devices. Rust is a systems programming language focused on three goals: safety, speed, and concurrency. It maintains these goals without needing a garbage collector, making it a useful language for a number of use cases other languages aren’t good at: embedding in other languages, programs with specific space and time requirements, and writing low-level code, like device drivers and operating systems.

  It improves on current languages targeting this space by having a number of compile-time safety checks that produce no runtime overhead, while eliminating all data races. Rust also aims to achieve ‘zero-cost abstractions’ even though some of these abstractions feel like those of a high-level language. Even then, Rust still allows precise control like a low-level language would.

  Try Rust online in the Rust Playground

  The Rust Playground is an online application that allows you to experiment with Rust by editing and running Rust code from the comfort of your browser; no need to install Rust locally! Several crates (Rust's reusable libraries) are available in the playground to showcase some of Rust's ecosystem.

  Version

  After a long period of iteration and experimentation, Rust 1.0 shipped on May 15th, 2015. This release is the official beginning of the Rust language's commitment to stability, and as such it offers a firm foundation for building applications and libraries. From this point forward, breaking changes are largely out of scope (some minor caveats apply, such as compiler bugs).

  Rust employs three release channels: stable, beta, and nightly. Every six weeks the beta channel is released as a stable version and the current nightly channel becomes the beta candidate.

  What does it do?

  Rust supports a mixture of programming styles: imperative procedural, concurrent actor, object-oriented and functional. It supports generic programming and meta-programming, in both static and dynamic styles. Rust is also capable of creating bindings for C foreign function interfaces.

  The language design pursues the following goals:

  Compile-time error detection and prevention. Clarity and precision of expression. Run-time efficiency. High-level zero-cost abstractions. Safe concurrency. Standard Library

  You can view what the standard library has to offer, and search through it, here. The scope of the standard library will continue to grow in future versions of the language.

  Getting Started

  To get started with the language:

  Start with the book. Read the official frequently asked questions and the questions tagged rust-faq. Open up the standard library documentation. Read Rust by Example. When you are ready for the deepest and darkest arts of Rust, dive into The Rustonomicon. You can find a lot more learning resources (including non-English ones) in the rust-learning repository.

  Getting Help

  Not all questions are suitable for Stack Overflow, and discussions certainly are not. If you wish to ask for a recommendation for an IDE, a library, or wish to discuss about the best way to organize a bit of code, then you may consider alternative venues.

  More open-ended questions and discussions are welcome on:

  the Rust-Beginners IRC channel, the Rust subreddit, the Rust users forum. There is a list of Rust IRC channels on the community page, and there are even language-specific channels if English is not your forte.

  Additionally, the AreWeXYet collection of websites provide helpful summaries and lists of thematic resources:

  AreWeIDEYet AreWeWebYet AreWeGameYet AreWeLearningYet Producing a Minimal, Complete, Verifiable example for Rust code

  Questions on Stack Overflow that seek debugging help ("why isn't this code working?") must include the shortest code necessary to reproduce it in the question itself. Without it, the question is off-topic and is likely to be closed. Here are some common points to follow to reduce your code to create a good MCVE:

  Try to run your code in a different environment. On the Rust Playground In a separate Cargo project Remove unused crates modules types enum variants / members struct members function arguments / return values Combine multiple modules into a single file by rewriting mod foo; as mod foo { / contents of foo.rs / } Replace complete or partial function bodies with the macro unimplemented!(). Remember that some steps might "unlock" other steps; try to apply each step in turn until no steps can be followed! Ensure that your error continues to occur after each change to avoid going down the wrong path.

  There's also some information that you should provide about your code:

  It will be assumed that you are using the current stable version of Rust. If you are not using this version of Rust, state your version. This can be found by running rustc --version. If you are using crates, state the versions of every crate that you are using. This can be found in your Cargo.lock file. Books

  Programming Rust Fast, Safe Systems Development Rust Essentials Rust Cookbook Beginning Rust - From Novice to Professional Code Editors & IDEs:

  Atom (Rust language support) Eclipse (plugin) Emacs (plugin) GNU Emacs (Rust-mode) InteliJ IDEA (plugin) Sublime Text (Sublime Text package for Rust) Vim (Rust.vim) Visual Studio Code (Rust extension, Rust extension with Language Server Support) Visual Studio (Visual Studio extension for Rust) Geany IDE (Rust Language support out-of-the-box) Gnome Builder (Rust Language support out-of-the-box) LightTable (plugin) See: Text editor features overview: (Syntax highlighting, Snippets, Code Completion, Linting, Go-to Definition, Syntax Checking).

  history excerpt history

  Code Language (used for syntax highlighting): lang-rust

  8,863 questions tagged rust Ask Question Stats

  created

  7 years, 9 months ago

  viewed

  4290 times

  active

  5 days ago

  editors

  26

  Top Answerers

  Shepmaster 112k9192292

  Vladimir Matveev 61.4k12139186

  huon 49.9k6132164

  Matthieu M. 185k25238459

  DK. 30.4k27088 more » Recent Hot Answers

  What is the stabilization process?

  Why do I get the error “the type of this value must be known in this context” when parsing a string to a number?

  How can I downgrade or install an older version of a tool I installed with cargo install?

  Why does cargo new create a binary instead of a library?

  Converting a Vec to Vec in-place and with minimal overhead

  more »

  Related Tags

  2018/04/09
• 
  

  Infoman

  Amazon Elastic Compute Cloud running Linux/UNIX $2.52

  $0.0152 per On Demand Linux t2.micro Instance Hour166 Hrs$2.52

  EBS $3.54

  $0.05 per GB-Month of snapshot data stored - Asia Pacific (Tokyo)7.511 GB-Mo$0.38

  $0.12 per GB-month of General Purpose SSD (gp2) provisioned storage - Asia Pacific (Tokyo)26.300 GB-Mo$3.16

  Elastic Load Balancing - Application $8.24

  $0.008 per used Application load balancer capacity unit-hour (or partial hour)3.549 LCU-Hrs$0.03

  $0.0243 per Application LoadBalancer-hour (or partial hour)338 Hrs$8.21

  2018/04/08
• 
  

  Infoman

  Berkeley offers its fastest-growing course – data science – online for free

  387 seycombi 13 hrs 89

  http://news.berkeley.edu/2018/03/29/berkeley-offers-its-fastest-growing-course-data-science-online-for-free/

  news.ycombinator.com/item?id=16768021

  By Public Affairs, UC Berkeley | MARCH 29, 2018 Click to share on Twitter (Opens in new window)4K+Click to share on Facebook (Opens in new window)4K+Click to share on Google+ (Opens in new window)Click to share on Reddit (Opens in new window)Click to email this to a friend (Opens in new window)Click to print (Opens in new window) The fastest-growing course in UC Berkeley’s history — Foundations of Data Science — is being offered free online this spring for the first time through the campus’s online education hub, edX.

  Data science is becoming important to more and more people because the world is increasingly data-driven — and not just science and tech but the humanities, business and government.

  David Wagner, Ani Adhikari, John DeNero Data 8X instructors David Wagner, Ani Adhikari and John DeNero

  “You’ll learn to program when studying data science — but not for the primary purpose of building apps or games,” says Berkeley computer science Professor John DeNero. “Instead, we use programming to understand the world around us.”

  The course — Data 8X (Foundations of Data Science) — covers everything from testing hypotheses, applying statistical inferences, visualizing distributions and drawing conclusions, all while coding in Python and using real-world data sets. One lesson might take economic data from different countries over the years to track global economic growth. The next might use a data set of cell samples to create a classification algorithm that can diagnose breast cancer. (Learn more from a video on the Berkeley data science website.)

  The online program is based on the Foundations of Data Science course that Berkeley launched on campus in 2015 and now has more than 1,000 students enrolling every semester.

  The intent is “helping students from any major or any background to address questions that interest them, using the powerful tools of data science,” Provost Paul Alivisatos told a congressional subcommittee on research and technology in July 2017. “From linguistics to chemistry, and from history to economics, students from over 60 majors have responded, and this is the fastest-growing program in the history of Berkeley…”

  The Foundations of Data Science edX Professional Certificate program is a sequence of three five-week courses taught by three winners of Berkeley’s top teaching honor, the Distinguished Teaching Award: DeNero, statistics professor Ani Adhikari and computer science professor David Wagner.

  “It’s exciting to be part of an initiative that brings data science to all, regardless of their background,” says Adhikari. “After all, everyone has to be able to reason based on data and to gain insight from data. And data science itself will benefit from diverse perspectives about what’s important. Our goal is to bring data science to the widest possible audience.”

  Anyone in the world can enroll for free; learners who want to earn the certificate will need to pay.

  The first of the three parts starts at 9 a.m. (PDT) on Monday, April 2. Enrollment will remain open after the course begins.

  2018/04/06
• 
  

  Infoman

  View Watchlist Amazon.com Inc. (AMZN) is closing in on music streaming leader Spotify (SPOT) and Apple (AAPL), now in second place, revealing for the first time that it has tens of millions of paid subscribers, a number that is growing at a fast rate.

  Steve Boom, vice president of Amazon Music told Billboard that Amazon Music Unlimited subscriptions have more than doubled during the course of the last six months and that it has tens of millions of customers. Boom said the growth is being driven by an uptick in the number of Amazon Prime customers and the growth of smart speakers like its line of Echo devices powered by its voice-activated digital assistant Alexa. He said that while lots of the news coverage predicts streaming music via smart speakers will become the norm in the future, it's actually happening now. “We wouldn't have grown to this scale if it hadn't been happening already," Boom told Billboard. (See more: Amazon Prime Price Hike Will Bring in $300M More Annually: Cowen.)

  The executive’s comments about Amazon’s streaming music service come just as leading music streaming provider Spotify launched its IPO this week. But Spotify doesn’t have a voice-activated smart speaker line, which is where Boom thinks Amazon can drive more subscribers to its music service. He said the smart speakers have given the Seattle, Washington e-commerce giant an untapped group of music streamers, which is where most of the new paid subscriptions are coming from. The executive noted that it's both older listeners who tend to be slower to embrace new technologies as well as country music listeners, a demographic that hasn’t been quick to move to streaming music.

  Amazon Music claims country songs are streamed 2.5 times more frequently on its platform than on others. Boom told Billboard that the goal is to expand the streaming market, not to go after the same demographics that other streaming services are targeting. "The technology itself is so simple that we don't just rely on those people who I would say are early tech adopters, which has been where a lot of growth in music streaming has been because it's been wrapped up inside of a smartphone. Not everybody wants to listen to music on a smartphone, it turns out," he said.

  Read more: Amazon Music Subscriptions Doubled in Six Months | Investopedia https://www.investopedia.com/news/amazon-music-subscriptions-doubled-six-months/#ixzz5BqiZW4Tr Follow us: Investopedia on Facebook

  2018/04/06
• 
  

  Infoman

  Show HN: Travel spontaneously on a budget

  180 johmu 21 hrs 116

  http://www.wander.am

  news.ycombinator.com/item?id=16742492

  Logo footer Wander is a travel search engine that allows you to find the perfect travel destination that fits your budget and preferences.

  Find us on Facebook

  Follow us on Twitter

  Check out our Instagram

  Download Wander&Aspire from the Chrome webstore

  terms | privacy

  2018 © WanderApp

  2018/04/04
• 
  

  Infoman

  Spotify opens at $165.90, valuing company at almost $30B

  93 harshgupta 50 mins 52

  https://techcrunch.com/2018/04/03/spotify-opens-at-165-90-valuing-company-at-30-billion/

  news.ycombinator.com/item?id=16746680

  Spotify opened on the New York Stock Exchange at $165.90, giving the company a market value of nearly $30 billion.

  The first trade didn’t happen until 12:45pm Eastern. This is halfway through the trading day, and a record for the latest time for a public debut.

  Spotify isn’t selling its shares on the stock market, meaning the company isn’t raising any money today. Instead, the event known as a “direct listing,” is a collection of transactions from existing shareholders (like employees and investors) selling shares directly to stock market investors.

  Spotify is basically trying to recreate the secondary market activity that happened before it went public. The company says that in 2018, shares traded on the private markets between $90 and $132.50. Since Spotify didn’t do an IPO, it set a “reference price” of $132 per share, which would have given the company a valuation of $23.5 billion.

  Unlike a traditional IPO where employees don’t sell shares for months, known as a “lock-up,” Spotify insiders are already allowed to sell.

  If few people opt to sell, it will drive the share price up, because of limited supply. If a lot of insiders sell, the reverse could happen, if investor demand doesn’t match it. This process may lead to increased volatility in the first few days or weeks of trading.

  In the long-run, Spotify’s performance in the stock market will largely depend on its business performance and outlook.

  Some investors are concerned that Spotify will run the course of competitor Pandora, which has struggled as a public company, partly due to hefty artist fees. Others argue that Spotify could be viewed as a Netflix, which has been successful at its entertainment licensing agreements.

  It’s certainly a big and growing business.

  The company says it is present in 61 countries and its platform includes 159 million monthly active users and 71 million premium subscribers.

  Spotify had 4.09 billion Euros in revenue last year (or close to $5 billion), compared to 2.95 billion Euros (about $3.6 billion) the year before. 2015 saw 1.94 billion Euros in revenue (about $2.38 billion).

  Losses for last year were 1.2 billion Euros ($1.47 billion), which compares to 539 million Euros ($661 million) the year before.

  Spotify previously raised about $2.7 billion in both debt and equity financing. Tencent, Tiger Global, Sony Music and Technology Crossover Ventures (TCV) are amongst its largest shareholders.

  2018/04/04
• 
  

  Infoman

  Ask HN: What are the best MOOCs you've taken?

  119 csdrane 1 hr 36

  news.ycombinator.com/item?id=16745042

  sampo 11 mins Most fun: Pat Pattison, Songwriting, Coursera. Very good lectures, very good material, very well presented. I used to think that the best feature of MOOCs is the automatic grading and feedback from programming homework, but in this course, for the homework songwriting you gave and got feedback from 3-5 random people in the course, and it was not only useful but this feeling of togetherness with strangers was even better than getting instantaneous feedback a from programming homework. There is no reason teaching art wouldn't scale to MOOCs as well.

  Nicest: Andrew Ng, Machine Learning, Coursera. Interesting topic, well-planned material, very well avoids going into the mathy details, but still conveys a feeling of understanding of the topic, so accessible to a wide audience. (Martin Odersky, Functional Programming Principles in Scala, Coursera, was almost equally nice, but had some rough edges in the first run.)

  Most interesting: Probabilistic Graphical Models, Daphne Koller, Coursera. Very interesting topic. I took the first run of the course and it had lots of rough edges. Needs a lot of work to apply the lectures to the homework. I haven't seen such a demanding course since I took quantum mechanics at university.

  Best organized: Jennifer Widom, Databases, Stanford. This is not the flashiest of a topic, but oh boy was it well organized. Runs like a clockwork. Everything in the lectures is relevant, everything from the lectures can be applied and is tested in the homework, there is lots of homework (but still not enough to make you remember SQL for the rest of your life if you don't keep using it), weekly homework has a nice progression from simpler things to medium difficult things, and the web environment is well designed, and gives wonderful feedback and guides you to get your queries correct.

  dmytrish 5 mins

  Probabilistic Graphical Models, Daphne Koller

  – the first course in the specialization has a very good and engaging start, but the gap between lectures and problems widens quickly after that (maybe that's why the author boasts about a "challenging" course "not for everyone"). I'm hesitant to take the next course of the specialization.

  sampo 0 mins I wrote about the first version when it was all in one course. It seems they have split it into 3 courses now. Probably it's less dense now.

  lr4444lr 3 mins I concur completely with this assessment of Ng's and Widom's courses.

  akbarnama 0 mins Most fun and learnt a lot in Introduction to Mathematical Thinking taught by Keith Devlin. I did this course from Coursera in 2012. The most fun part was the forum where students collaborated to discuss and gain better understanding of the problems.

  rectang 40 mins Khan Academy math, because of the exercises.

  They are consistent, not very buggy, gamified, and consumable in small or large amounts. Sal Khan is a good communicator and the videos are decent, but it's the exercises that make Khan Academy exceptional.

  blcArmadillo 34 mins Andrew Ng's Machine Learning Course on Coursera: https://www.coursera.org/learn/machine-learning

  vector_rotcev 39 mins Learning how to learn, Barbara Oakley, Coursera.

  By far and away the best learning course I've taken in my life as well, I wish it had been available before I had completed my formal education.

  OldSchoolJohnny 10 mins Everyone who wants to learn anything in life should take this course first, I can't endorse it enough.

  neovive 24 mins Over the past few years, I've watched a few courses on Udacity, Coursera and EdX. I prefer taking ad-hoc courses to fill knowledge gaps (statistics, AI, programming, math, etc.), so I can't give a full review of the complete Nanodegrees, Certificates, XSeries, etc. I usually watch the lessons as needed without completing the entire course; mixing and matching MOOC courses with video learning sites (e.g. Datacamp, Youtube channels, Khan Academy, Egghead, etc.)

  If I had to pick a MOOC platform, I prefer Udacity's more hands-on approach, but enjoy courses on EdX and Coursera. The quality of all three MOOC platforms is excellent. It's an amazing time for autodidacts!

  If you're starting from scratch, without any background knowledge, the certificate programs with access to mentors are a great place to start. The curriculum is designed by industry professionals and/or experienced professors. This saves you time, keeps you focused and offers a place to get help when needed.

  jitix 18 mins Do you have any recommendations on statistics for beginners?

  neovive 3 mins Udacity has two free introductory statistics courses that they recommend prior to starting their AI nanodegree program:

  Intro to Descriptive Statistics [https://www.udacity.com/course/intro-to-descriptive-statisti...]

  Intro to Inferential Statistics [https://www.udacity.com/course/intro-to-inferential-statisti...]

  Khan Academy also has very in-depth coverage of statistics, starting from the basics. https://www.khanacademy.org/math/statistics-probability

  bitL 2 mins Udacity's Self-driving Car Nanodegree by a wide margin.

  From the rest, MIT's Underactuated Robotics was pretty rad, Udacity's Deep Learning Foundations Nanodegree was very useful, Ng's Machine Learning was made super easy. The School of AI's DApps/Blockchain course so far looks pretty good as well.

  zachwill 27 mins Not your usual answer for HN, but the best online courses I've ever taken are Chris Orwig's photography stuff on Lynda.com. Most local libraries have a free subscription with Lynda, and the way he teaches photography/Photoshop/etc was so useful to learn during college. It's not math or machine learning, but the guy is an absolute master at his craft -- and offers some of the clearest explanations on his line of thinking when working on projects.

  billdybas 2 mins MoMA's "Fashion as Design" course was pretty interesting: https://www.coursera.org/learn/fashion-design/

  loganekz 32 mins Functional Programming Principles in Scala [1] taught by Martin Odersky, professor at EPFL and creator of the Scala language.

  [1] - https://www.coursera.org/learn/progfun1

  MattyMc 5 mins The Hardware/Software Interface from the University of Washington (previously offered on Coursera). As a non-CS major, it gave clarity to a lot of the magic that happens when you write code. Fabulous course. https://courses.cs.washington.edu/courses/cse351/

  anarchimedes 21 mins I like the dialogue between Hastie and Tibshirani in their statistical learning course from Stanford [1]. I found the accompanying ISL book and c-cran depositories helpful for when I wanted to go deeper beyond the lecture.

  [1]https://lagunita.stanford.edu/courses/HumanitiesandScience/S...

  idiocratic 39 mins Robert Sedgewick's Algorithms has been one of the best for me, not only as a general refresher on algorithms, but also as a way of better understanding complexity notations.

  mrnaught 12 mins Walk through of each algorithm with visualization is amazing in that course.

  grdvnl 20 mins I took the Programming Languages course on Coursera which is so far the best course for me. It changed the way I learn any new programming language.

  https://www.coursera.org/learn/programming-languages

  I see that they have split the course into 2 parts.

  adenadel 14 mins I'm going to go off the beaten path here. I really enjoyed the edX course Molecular Biology (MITx - 7.28.1x). They teach DNA replication and repair

  gringoDan 26 mins 3b1b's series on Linear Algebra is essential for an intuitive understanding of the topic: https://www.youtube.com/playlist?list=PLZHQObOWTQDPD3MizzM2x...

  (Really anything by 3b1b)

  mcjiggerlog 3 mins Can anybody recommend any courses for learning WebGL?

  ScoutOrgo 19 mins Fast.ai's (fast.ai) deep learning and machine learning courses. No ads, good notes/forum, and very approachable material for anyone that knows basic coding.

  samuraijack 13 mins Compilers by Alex Aiken. https://lagunita.stanford.edu/courses/Engineering/Compilers/...

  rjammala 15 mins Algorithms by Tim Roughgarden on Coursera

  mcintyre1994 38 mins fast.ai is outstanding, if you're at all interested in deep learning.

  user2994cb 23 mins Dan Boneh's Cryptography Part I on Coursera. Will we ever get Part II? Enrolling for Sept 2018 according to Coursera.

  chongli 24 mins Programming is for everybody on Coursera. Taught with Python, extremely approachable for non programmers. Teaches you fun stuff including how to use sqlite and how to scrape websites, use JSON APIs, and more!

  contingencies 27 mins Venture Deals. https://www.kauffmanfellows.org/online-course-venture-deals/

  Fnoord 32 mins Learning How To Learn [1] by Dr. Barbara Oakley, Dr. Terrence Sejnowski available on Coursera & elsewhere.

  [1] https://www.coursera.org/learn/learning-how-to-learn

  fermigier 23 mins "Learning How to Learn" (already cited) and "Critical Perspectives on Management" (by Rolf Strom-Olsen, available on Coursera, starts in 2 weeks!).

  ajudson 32 mins Nand2Tetris (part 1)

  dmytrish 11 mins The second part is awesome too.

  The first part shows how to design an unoptimized and simplistic, but complete and working 16-bit CPU and RAM from logic gates.

  The second part builds a whole software stack on top of it using a virtual stack-based VM:

  - CPU assembler;
  - a (AOT) compiler from the VM opcodes 
    into the CPU assembly;
  - a compiler from the high-level language 
    called Jack (an educational mix of Java/C 
    with many complex parts removed) into the 
    VM opcodes;
  - a standard library for the Jack language 
    (Screen/Keyboard/Output/Math/String/Array/
    Sys/Memory classes), including drawing lines/
    circles and bitmapping glyphs into video memory 
    for text rendering;
  - your own project (usually a simple game and 
    sometimes marvels like [0]) written in Jack 
    on top of all of that;
  

  The courses are definitely very challenging and some previous exposure to the topics is desired. [0] https://github.com/QuesterZen/hackenstein3D

  vowelless 22 mins Learning from Data, the Caltech course.

  2018/04/03
• 
  

  Infoman

  

  2018/04/02
• 
  

  Infoman

  Henri Matisse

  For me, the subject of a picture and its background have the same value ... there is no principal feature, only the pattern is important

  

  2018/04/02
• 
  

  Infoman

  Experiment, Measure, Repeat

  47 ecaml 1 day 6

  https://blog.buildo.io/experiment-measure-repeat-5e2c389e63ff

  news.ycombinator.com/item?id=16715718

  source: https://elearningindustry.com/wp-content/uploads/2015/11/10-online-research-tools-every-online-learner-know-800x600.jpg A few months ago, I read an article from Michael Simmons where he essentially said that

  “Deliberate experimentation is more important than deliberate practice in a rapidly changing world.” Even though the article focused more on experimentation for personal improvement, the first thing that came to my mind was something like: “this should be one of the core values of every company” (here’s a Forbes article that tells the same).

  Here at buildo, we take experimentation and continuous improvement seriously: it’s our core value. The goal of this post is to share with you how we do experiments internally, also adding some advice that you could follow, hoping to be useful for some of you that do experimentation or that want to start (you should!).

  In October of the last year, we started noticing that some of our Slack channels were noisy and difficult to follow, especially in high work-load weeks. People were speaking about different topics in the same channel (e.g., two issues of the same software project), with unrelated messages alternating between one topic and another.

  We decided to solve such pain point like we usually do: with an internal experiment. Luca Cioria, who proposed the experiment, became its owner and did the following:

  defined a goal: “Make Slack easier and faster to read, considering the growing amount of messages.” defined how to try to reach the goal: “Use threads as much as possible, and create a culture of respecting channel’s quiet.” defined a metric to be measured to evaluate the success of the experiment: “The percentage of messages in threads.” defined success conditions for the experiment: “People like threads and accept to use them” and “there is a significant increase in the metric (set to 70% mostly as a psychological goal)” defined a date for the final evaluation (i.e., success or failure?): “2018, 10th week.” Luca then presented all this to the rest of the team, which agreed to start; he then created a plot of the success metric and added it to our internal dashboard, to create awareness about the progress.

  This is the result so far:

  Percentage of messages in threads (in blue) VS Slack total message count (in green) The experiment was considered successful, and the outcome was a new dedicated paragraph in our internal Slack guidelines.

  Our internal dashboard, near the beloved coffee machines I think that having experimentation as one of the core processes in your company is extremely important, and I believe that creating such culture of experimentation is not as difficult as it looks.

  There is no secret sauce to do it, but I will now share some essential guidelines that we follow and that you could try to use too:

  Set up recurring meetings to discuss pain points and ideas, and to come up with possible experiments to address them. Once you have a set of possible experiments, prioritize them and work only on few of them simultaneously. We prioritize based on value and effort of every experiment (e.g., we give priority to high-value-low-effort ones) Documentation is essential: every experiment shall have an owner, a goal, some success metrics to be collected, and a deadline; all this shall be clearly stated in a so-called “canvas” (we use DropBox Paper for that, and for almost all our internal processes. Francesco Negri spoke about it in another post) Use a data-driven approach: collect data and plot the experiment metrics; to create awareness, make them easily visible to everyone. For example, we just put a dashboard near the coffee machine, and it works. Quantitative metrics-based goals are good, but not reaching them should not automatically result in failure: there should be a final meeting for every experiment, to discuss its outcomes Embrace failure and avoid maintaining useless things: failure is a vital part of the process. If an experiment is not considered successful, ask yourself why: this could result in inspiration for future improvements. Most importantly, archive or delete everything related to it, to avoid having to maintain unneeded tools or processes Although this works for us, we encourage you to take inspiration from what we wrote to build your own process of internal experimentation. The process itself shall be continuously improved, like basically everything in your company. At buildo, nothing can “escape” continuous improvement: the dashboard near the coffee machine started as an experiment, but also this blog, the process I described, and everything else we do and use every day.

  At first, it seems a lot of work, but it’s easier than it looks: call the first meeting with your team, find a way to collect and display metrics, and there you are, ready to experiment. Happy improvement :)

  To build our experimentation process, we took inspiration from the well-known PopcornFlow. For example, here’s the PopcornFlow experiment canvas template Special thanks to Jacopo Romei and to Avanscoperta, who help us continuously improving our continuous improvement processes over time We plan to write something more about our internal processes in the next posts: if you liked this, stay tuned!

  —

  If you want to work in a place where we take experimentation and improvement seriously, and where we care about the quality of our development workflow, take a look at https://buildo.io/careers

  2018/04/01
• 
  

  Infoman

  Generating $32,000 in 30 days selling a crypto-filled USB stick

  134 patwalls 15 hrs 95

  https://www.starterstory.com/cblocks

  news.ycombinator.com/item?id=16708526

  2018/03/30