HNHacker News
TopNewBestAskShowJobs

romaniv

2,750 karma · joined February 2, 2011

submissionscomments
romaniv··on Show HN: Whiteboardfree – Developer Jobs at Companies That Don't Whiteboard
Whiteboards are okay for ad-hoc lists and diagrams, but they are a poor medium for coding.

Some of the biggest productivity gains I've had over the years were related switching from "writing code down" to "thinking in code" mentality and abandoning linear, paper-like ways of thinking, which hindered me when I was in college.

For anyone who believes that paper and whiteboards are "good enough" for coding, I highly recommend watching some of Victor Bret's or Alan Kay's talks. Pretty much any of them. They consistently touch on the subject of dynamic media as a way to empower people to... well, think better for the lack of better terms.

romaniv··on Proxy, a new JavaScript ES6 feature
Proxies are a basic feature that should be a part of any OO language. Their usefulness ranges from "save yourself a lot of repetitive work" to "impossible to implement without".

Very simple real-life use case: factor out logging and move it outside of normal logic of the object.

romaniv··on AI and Compute
>I don't know if I agree with that, since we are seeing some interesting things

There were plenty of interesting results in AI research before the last two AI winters.

romaniv··on A dozen Google employees quit over military drone project
IIRC, original Recaptcha was about OCR of books that weren't readable. I'm not sure why anyone uses it right now. Like you said, it's obvious where it's going. Not to mention that it's increasingly annoying for users.
romaniv··on OpenAI Research
I would be way more impressed if there was an AI that can play Overwatch. It's one of those games that are fundamentally hard for computers because of meta-gaming, map differences and ability combinations. It would be much harder for AI to converge on some single game-breaking trick, or just bruteforce possible strategies through blind trial and error and memorization.

Also, aim doesn't matter as much, so you wouldn't have to introduce some artificial restrictions. (Reaction time does give you a big advantage, though. You can counter a lot of things simply by reacting to voicelines in time.)

romaniv··on Plastic Bag Found at the Bottom of World's Deepest Ocean Trench
>a large amount of waste generated by these countries is done so in the production of goods meant for Western consumption.

Why does this matter? A factory polluting local lakes and rivers to make goods for export to another state definitely couldn't use this as an excuse. They control the process, so they are responsible for consequences of using that process. So why rules are suddenly different when we talk about countries?

romaniv··on Using the Web for a Day with JavaScript Turned Off
> Is spending the time and resources on creating a non-JS site worth it

Comments like these make me wonder if any of the people saying PE is hard ever tried it at all. The whole point of progressive enhancement is that you don't create a separate non-JS website.

romaniv··on Using the Web for a Day with JavaScript Turned Off
>If your page consist solely on a dynamically generated graph

You can put data for the graph into a table, then use a script to render that table as a graph. If people cared enough, graph libraries could expect heir data in table format.

romaniv··on Using the Web for a Day with JavaScript Turned Off
>It made web app development so much easier

I aim to develop web UIs using (mostly) progressive enhancement. I adopted several practices and developed some libraries supporting this process. As a result:

1. When I need to create something, I know exactly which data structures to use. This is determined by what is available in modern browsers.

2. I can quickly prototype solutions using plain HTML focusing on logic rather than style.

3. Since my data is by definition contained in HTML, I can easily query it using CSS queries. This makes working with nested data a breeze.

4. I factored code I reuse into generic, self-contained behaviors. Things like "when this form is invalid, this control should be inactive". There is usually very little to none page-specific code.

5. Once a behavior is written and tested, "debugging" usually involves simply making sure the page has the right attributes. I can do this by running a CSS query in console or looking at DOM. No breakpoints, no stepping, no watches.

6. The most important part: I can add one behavior at a time and the result is something that works and makes sense.

7. A lot of UI "logic" I used to have in scripts naturally migrated to CSS.

I like this process way more than fiddling with tons of page-specific "glue" JavaScript. I especially like it when I'm in a crunch, because it pre-defines a lot of the things I would have to "design" on the fly in a traditional development workflow. Also, if I run out of time I have a working (if ugly) app. If I introduce a bug somewhere in UI or run into a compatibility issues, it usually doesn't result in the entire user workflow stopping dead.

romaniv··on Three-day no-meeting schedule for engineers
This is actually a very acute and non-obvious observation. People might get tired during meetings, but this will mostly affect following meetings, rather than engineering. And it's easier to focus on an activity (and optimize it) if it lasts a day rather than an hour. This might end up improving the overall "meeting culture" in the company, teaching people to schedule breaks, be on time and so on.
romaniv··on IRS’ 60-Year-Old IT System Failed on Tax Day Due to New Hardware
>The question for the industry remains: how do we deal with systems that should be up for 10 years? 25 yrs? 50 yrs? 100 yrs? We didn't even have computers as they exist today 50 years ago.

Is this a question the industry actually cares about? If anything, I'd say it got much worse at creating long-living software. It sucks to run a mainframe-based system from 50 years ago, but running a 10-year-old Web Forms application can be borderline impossible. I'm scared to think of what will happen with all the cloud-based stuff in the next 10 years.

>How do we separate business logic in a self-contained way that can be independent of hardware and (from a certain level) software?

I often say that there is no such thing as "business" logic. There is just logic. The least maintainable systems I've worked with were the ones with the most sophisticated attempts at abstraction.

In my opinion, the key to software longevity is making systems that allow for local reasoning, i.e. analyzing parts without knowing the whole. This can be achieved in a lot of different ways, but it needs to be a deliberate goal in system design.

romaniv··on An open source implementation of DeepVoice 3: 2000-Speaker Neural Text-to-Speech
>Google seems to have the best know-how.

Funny how they publish their "research papers", yet no one else is able to implement their engine with even remotely comparable results.

romaniv··on Neural Network Quine
>Biological life began with the first self-replicator

I never understood why people assume that those biological self-replicators were in any way analogous to quines. Quines generate information from within themselves. Proto-life would have to assemble copies of itself from molecules available in the environment. It seems like completely different principles of work.

romaniv··on CSS: The bad bits and how to avoid them
This argument doesn't stop a myriad of widely used JavaScript frameworks that rely on custom attributes. Besides, if you really care about validity, you can use

  <div data-my-block="my-element my-modifier">
It's still shorter, more readable and avoids repetitions. The difference becomes even more pronounced as the complexity of your styling goes up.
romaniv··on CSS: The bad bits and how to avoid them
Never understood why people prefer verbose and repetitive BEM classes over custom properties and appropriate CSS selectors.

  <div class="my-block my-block__my-element my-block__my-element--my-modifier">
becomes

  <div my-block="my-element my-modifier">
This solves for specificity, and uses built-in CSS features instead of a "manual" workaround.
romaniv··on Garry Kasparov: Open-Ended AI [video]
>Kasparov apparently continues to believe that his expertise in chess transfers to other disciplines.

There is a lot of irony in this statement, since he speaks about expertise of AI systems in Chess and other games transferring to more open domains.

romaniv··on Using AI to match human performance in translating news from Chinese to English
> The proof of a theorem can be memorized without any understanding just as easily as algebraic rules.

Sure, but in practice students who rely purely on memorization can't answer questions that go beyond what's directly covered in textbooks.

romaniv··on Using AI to match human performance in translating news from Chinese to English
Microsoft Research also seems to be much more willing to create usable demos of their work.
romaniv··on Using AI to match human performance in translating news from Chinese to English
That's not his argument at all. His argument is that just because you can do some task doesn't mean you "understand" it.

You don't need some fancy philosophy and complex thought experiments to see what he means. Just look at how people learn math. You can do calculations by memorizing algebraic rules, but that's not the same as understanding why those rules exist and what they mean. Even though you will calculate answers correctly in both cases, we all know there is a qualitative difference between them.

Back to Searle. His argument is that everything computers do is analogous to rote memorization and that transition to understanding requires something computers don't have.

Whether you buy his argument, two things are clear. First, there is a difference between just producing results and understanding the process. We all experienced this difference. It's all theoretical as long as you stick to simple tests (like multiple-choice exams), but becomes relevant when you suddenly expand the context (like requiring the student to prove some theorem instead of doing a calculation). Second, we also know that for humans this difference isn't just quantitative. Memorizing more algebraic rules and training in their application will not automatically result in students gaining understanding of mathematical principles.

romaniv··on Surprising Things About School
I would be ecstatic if I had an option to replace some of my university classes with MOOCs. The leap in quality of lecturing would be gigantic. MOOCs made me realize that most of the subjects in college I found "complicated" really weren't. I just had horrendous professors and bad books.

I'm currently in the process of "fixing" my college education by watching lectures on fundamental subjects like biology (which I never had in college for some reason), chemistry (which was taught really badly) and linear algebra (taught awfully). It's astounding how many new things I'm picking up.

Plus, there are classes like AI, Cryptography, Model Thinking and so on. Nothing of this sort was available to me in college at all.

romaniv··on Surprising Things About School
>Nearly everyone recognizes that MOOCs are by and large a failure with ~2% completion rates, but they make us feel good because now it’s the students’ fault not they’re not learning, not the school’s

Before: simply getting familiar with a subject required me to jump through bureaucratic admission hoops, go through arbitrary prerequisites, pay tons of money, bend my life around some class schedule and then get mediocre lecturing from some random guy who happened to teach the subject at the nearby university at that time.

After: I can watch lectures from world-class professors and universities at any time, starting with any difficulty level, on a wide variety of subjects. I can also get free help from other people if I need to.

How is this a failure? MOOCs are one of the greatest thing that ever happened to the Internet and education.

romaniv··on Unexpected challenges of making money on the internet
Last time I read about it, using 3D Secure changed your liability agreement with the credit card provider. Not to mention that its implementation is a confusing mess that bounces you between badly designed websites with questionable security.
romaniv··on Stimulus: A modest JavaScript framework for the HTML you already have
Allow me plug something. If you just want to "AJAXify" a few links and forms you could try the splice-into behavior from this repo:

http://quickenloans.github.io/Behaviors.js/

It has no server-side dependencies and should be easy enough to use. But it doesn't use push history, which limits its usefulness for more extensive use cases.

romaniv··on Fizz Buzz in Tensorflow (2016)
>I'm also wary of people who see themselves above mundane requests.

Let me give you an analogy from another domain.

"My company requires excellent communication and writing skills, so would you mind talking a 5h grade spelling and grammar test?"

romaniv··on Lessons from Optics, the Other Deep Learning
Then maybe we should put more resources to study other approaches and higher-level concepts, like statistical learning research done by Vladimir Vapnik.
romaniv··on Lessons from Optics, the Other Deep Learning
>I think it would be healthier for the community to make it a habit to explicitly document why a certain topology and layer sizes were selected.

Also, which other topologies were tried and failed to produce good results. It's amazing that this information is missing from most modern ML papers.

romaniv··on Fizz Buzz in Tensorflow (2016)
It is condescending. Asking people to do FizzBuzz indicates that you expect a large portion of your candidates to be completely clueless. That's not a good message to send.

Think about the following. FizzBuzz filters out some extreme cases of incompetence, but tells you nothing of value about the people who passed. You can use the same chunk of interview time to ask a more challenging question that will still filter out the extremely incompetent, but also give you some level of confidence in the people who passed. So why would you prefer to ask a question that gives you less useful information?

It's not unlike a sorted array search algorithm. To maximize your efficiency you start in the middle, not at some arbitrarily chosen low point.

romaniv··on Fizz Buzz in Tensorflow (2016)
This legitimately looks like some legacy apps I had to deal with in real life. The language and the framework of choice were different, but the principle was the same.
romaniv··on Designing Windows 95’s User Interface
> At the time, it was a huge leap forward in desktop computing.

After seeing some old demos of Xerox Star, I'd question that.

https://www.youtube.com/watch?v=Cn4vC80Pv6Q

romaniv··on Designing Windows 95’s User Interface
Desktop paradigm isn't a Windows 95 thing. It's Xerox Alto thing.
← PreviousPage 6 of 30Next →