ntoshev

72 ms·

661.
by ntoshev 18y ago
Thanks for the feedback. The performance of the popups is several times better now. Clicking the word clouds might highlight all its occurrences on the page, I think this helps understanding. Maybe we'll implement it in this way.
662.
by ntoshev 18y ago
The tests talk solely about read performance and not transaction throughput. I don't believe write performance can be increased so dramatically, especially not just using a cache-friendly data structure.
663.
by ntoshev 18y ago
You are underestimating what is necessary to do implement tractable algorithms on big dataset. The classic index/search problem would be extrmely simple if it weren't for the big data involved: it is basically hash table lookups plus set in
664.
Ask HN: please review Wordics, a word cloud widget(wordics.lemmatica.com)
6 points by ntoshev 18y ago|3 comments
665.
by ntoshev 18y ago
Hackers are not interested in a career, they are interested in building cool stuff.
666.
by ntoshev 18y ago
Any sufficiently advanced computation is indistinguishable from intelligence. --Arthur Clark, paraphrased
667.
by ntoshev 18y ago
I understand the concept of intentional programming in this way: 1) start with Lisp, complete with macros 2) add multiple skins: some of them are basically a C-syntax representation of the program, some might be visual programming (connecti
668.
by ntoshev 18y ago
What's this, somebody created a Google Custom Search over the torrents tracked by thepiratebay?
669.
by ntoshev 18y ago
Content-based data may not help with Netflix prize, but they do help to quickly drill down to what people want to see. You can pick a "nano category" based on the meaning of the words in it, rating data can be superior but cannot be used in
670.
by ntoshev 18y ago
More practically, the "arbitrary piece of code" can be expressed in a less-than-turing-complete language or language subset, and it would actually be possible to prove useful things about it.
671.
by ntoshev 18y ago
The article really mixes two messages: "behavioral economics works" and "Obama knows what's good for you". The second one is claimed without any evidence, so I chose to ignore it.
672.
by ntoshev 18y ago
Read the article. It claims they have found a better estimate of word importance in a document than tf-idf, which would be very significant. Also it doesn't seem to need text segmentation, which means no need for language-specific tokenizer
673.
by ntoshev 18y ago
Netflix needed a formal measure for their contest, so RMSE is a useful one while "making people happy" is not. A business that relies on recommendation can plug a new algorithm with better RMSE and get improved results immediately, it is an
674.
by ntoshev 18y ago
I would also add the "Theorizing from data" talk from Norvig: http://www.youtube.com/watch?v=nU8DcBF-qo4
675.
by ntoshev 18y ago
Practical natural language apps might require much less, though: see Norvig's spell checker for example. You can probably fit the google index from 1998 on a single modern machine.
676.
by ntoshev 18y ago
Sony Walkman also didn't solve any problem people knew they have at the time. "Solving a problem" is a good approximation of something people want, but it's not a perfect substitute.
677.
by ntoshev 18y ago
Yes, I expect bandwidth to be the bottleneck. But it can be pretty cheap: serveraxis for example has unlimited bandwidth at 10mbps starting from $40 a month.
678.
by ntoshev 18y ago
Thanks, I don't really intend to run it as a business. Without being a forum spammer, I personally used Mailinator a lot and saw the chance to make it easier. I warned about the danger of stealing the account in the FAQ, on the front page;
679.
by ntoshev 18y ago
Thanks for the suggestions, I like them and will probably implement some variation. I hope to be able to run it off a minimal VPS account as I am doing now. The architecture is very slim (no disk access, no interprocess communication) but u
680.
by ntoshev 18y ago
It has some intelligence intended not to DOS legitimate sites: if a site sends a support link along with the confirmation link, for example, the support link won't be opened over and over again. Like Mailinator, the site has also other form
681.
Ask HN: Please review mailop.com, an instant disposable email you don't even need to check(mailop.com)
45 points by ntoshev 18y ago|28 comments
682.
by ntoshev 18y ago
Please add GAE's webapp.
683.
by ntoshev 18y ago
Then you should have said "Internet Explorer opening PDFs is a pain point and scribd solves that". Fair enough, although I can argue Chrome is both much better solution to this particular problem :) Personally I think scribd solves some pro
684.
by ntoshev 18y ago
I do. Google Chrome and Safari both open PDFs very smoothly on Windows. Even in Firefox I am not sure Scribd is an improvement over opening plain PDF (it shows a progress indicator, but it might as well take more time to open in total).
685.
by ntoshev 18y ago
Actual talk recording: http://videolectures.net/wsdm09_dean_cblirs/ Summary: http://glinden.blogspot.com/2009/02/jeff-dean-keynote-at-wsd...
686.
Jeff Dean: Challenges in Building Large-Scale Information Retrieval Systems [PDF slides](carbon.videolectures.net)
2 points by ntoshev 18y ago|1 comments
687.
by ntoshev 18y ago
This article itself is a good example why obsessing with SEO is bad. The article is written for search engines to index, not for humans to understand. It obviously compiles information from various sources, but little effort is made to make
688.
by ntoshev 18y ago
It is not really clear how is it structured, but it seems that Sequoia will own equity in the new startups funded by the YC-managed fund. In this way YC or that new fund won't need an exit strategy.
689.
by ntoshev 18y ago
search.twitter.com is actually provided by a company called Summize that Twitter bought: so it worked for them.
690.
by ntoshev 18y ago
Is the "parallelism" section theoretical, or you can get the Haskell compiler to actually generate multi-core code automatically?

More ›