HNHacker News
TopNewBestAskShowJobs

terra_t

990 karma · joined February 9, 2010

submissionscomments
terra_t··on If you build it, they won't come, unless...
Maybe I'm old school, but I think it comes down to work.

When I was the publicity directory of a college radio station at a small school, I'd put up 300-400 letter-sized posters to promote a dance with maybe 7-8 distinct designs. This would change the appearance of the campus and pack the house.

A lot of people, doing the same job, would put up 10 or 20 posters of the same design and waste more energy vociferously defending their right to be lazy and that "everybody will see them" than it would take to just make all the designs and walk around putting the signs up.

The main reason people fail at SEO and SMO is that they radically underestimate how much work is required to succeed. That's good news for people who do work hard, because out of 20 potential competitors, 18 of them are easy to beat.

terra_t··on Thingd is building a structured database of every object in the world
If RDF's got a problem it's that it's too simple, at least in terms of the functionality it's offered.

It's complicated as hell to do simple things in RDF, but that's a different issue, not the one I struggle with every day.

terra_t··on Thingd is building a structured database of every object in the world
How's this better than Freebase or dbpedia?
terra_t··on Reddit hacking for votes and profit
Good insight. My thinking today has moved beyond SMO more in the directions of public relations.
terra_t··on Reddit hacking for votes and profit
It's hard to say.

The toughest problem people have starting community sites is getting the community. Dealing with problem behaviors is something you worry about after you've got the community. Just as casinos hire shills, many community site operators use Mechanical Turk or statistical models to get the ball rolling...

Who knows, we might soon have replicants that are more satisfying to interact with than ordinary humans.

terra_t··on Reddit hacking for votes and profit
You can definitely buy them from people who host their own machines in data centers.
terra_t··on Reddit hacking for votes and profit
The people who do things like this, those who do any social media operations (SMO), are smart enough to use it to promote 'acceptable' content that won't be flagged as spam.

If it comes down to a war with the operators of the site, you will lose. If it gets personal, they'll burn you.

Successful SMO promotes content which, plausibly, you'd imagine succeeded organically. Sometimes the content is excellent and sometimes it's a little substandard, but it doesn't stick out as spam. It decorates itself in a pretty cloak to look legitimate.

Unethical tactics in SMO are like steroid use in sports: Steroid users aren't lazy people taking a shortcut, they're people who train hard, can play a good game honestly, but are looking for an edge that will keep them in the market.

It's a tough business because social traffic isn't good traffic for people who play the numbers game. You've got to relentlessly produce fresh content and relentlessly promote it. If you stop, you could be left with little permanent traffic. You'll get hooked on social traffic, however, by the ego rush you get the first time you get a burst of it and it crashes your server.

terra_t··on Debunking the duct tape programmer
"Analysis Paralysis" often has a root cause: incompetence.

A pal of mine was a consultant for a logistics company serving a large country that's north of the U.S. This company was developing a package tracking system in Java, back around 2000 or so, when "Java" was often a synonym for "Project Failure."

Now, when the whole thing started, I told him that I could knock off a prototype in a week in PHP. Nope, these guys decided to spend five months developing a framework for their application, and then they were going to spend the next month developing the actual app.

Well, six months later, I hear that the predictable outcome had happened: management was shocked to discover that the application took 30 seconds to serve up a page view.

My pal sent me a copy of the app and it was an absolute rat's nest: the kind of project where you use 1500 lines of code to say what you could say in 15. The sheer bulk of the code made it difficult to maintain: and it had poor performance written into it's DNA. The real problem was that this wan an IBM shop that was developing its first webapp, developing its first Java app and... nobody knew what they were doing.

-----

Don't get me wrong. There's nothing wrong with front-loading your design and analysis work... IF YOU KNOW WHAT YOU ARE DOING! In the hands of people who have a holistic view of the application and toolset it's based on, front-loading is a beautiful tool that can make implementation and testing a breeze.

However, because it delays feedback, prolonged analysis doesn't (necessarily) test the ideas behind a design. Bad ideas can get in, and stay in -- nothing challenges the arrogance of the uninformed.

If, on the other hand, you are humble, and "know what you don't know", a rapid prototype can be a powerful way to get some experience, test your ideas, and, who knows, maybe even ship something.

terra_t··on Reddit hacking for votes and profit
Not necessarily.

Most people find success on social media sites is capricious; not a lot of people read the "new" queue, so your first few organic votes are a matter of luck. Downvotes on reddit make this go double. Even great content can get buried if it doesn't get noticed.

For instance, there's a certain blog on a three letter domain name for which I'll sometimes see two or three posting on the front page at HN. That guy's got a voting ring. You might call it something else, like "I have a twitter account with a lot of followers who use HN", but that's what it is. This sort of practice takes some of the chance and variation out.

An advanced method is to recognize the role of "social proof" in social media. My informant has collected behavioral data sets from certain social media sites and discovered that the C.T.R. on articles increases as the number to the left increases: even if the headline is uninteresting, the high number makes people think that it "has to be good;" in experiments where the number of "ringer" votes was varied, she discovered that the ratio of "organic" to "ringer" votes got better the more "ringers" there were.

terra_t··on Reddit hacking for votes and profit
I knew this purple-haired superheroine who did this to a certain "Zone" which is famous for vapid and incoherent blog postings written in broken English.

She made about 1000 fake accounts and used Tor to make them appear to come from multiple IP addresses. The fake accounts were made with a statistical model that generated plausible usernames, real names, profile pictures and everything -- it could even pair up the gender of the first name and of the picture... most of the time.

Once she wrote a blog post she'd just set the sequencer and her posting would hit the front page in about two hours. She found it was a very consistent way to get 5k page views, sometimes better if the blog posting was good and it got picked up on other social media sites like proggit.

She'd gotten a few of her blog posts to be "all time favorites" on this site, but then another project eclipsed her blogging and she lost the credentials for the fake accounts in a hard drive crash.

Today, this "Zone" is dominated by other people doing the same thing. If you know what to look for, there's a quite unmistakable signature.

She thought about doing the same for reddit, but it was clear at the start that reddit had much better defenses -- it would have been a bigger project, and she probably would have burned some of her proxies before getting it right.

Note that you can get around the IP address problem in several ways. One of them is that you can rent SOCKS5 proxies for $3/month/IP address. Supervillians use bot nets.

terra_t··on What’s Wrong with “HTML5″
Note that Microsoft and Apple want to obliterate flash just as much as anybody does... IE9 shows that even Microsoft realizes that "HTML5" is a far deadlier weapon than Silverlight.
terra_t··on What’s Wrong with “HTML5″
The only person who ever questioned that word for me was a wiccan teenage girl who sees fairies.
terra_t··on What’s Wrong with “HTML5″
People regularly convert verbs into nouns. It's a semantic error, but it makes the invisible process of "making the web better" visible to the muggles.
terra_t··on What’s Wrong with “HTML5″
I think it's harmless, and it crystalizes the idea that things are changing, the same way that the term "Web 2.0" did.

Muggles have a lot of trouble with names. I worked for a big Uni back in the age when Mozilla's browser was caled "Mozilla" and it was impossible to get anybody to take Mozilla seriously. This was a place full of Unix-heads and Microsoft-haters, and we found that in 2002 we still had a 30% market share on campus of the despicable Netscape 4 browser.

When "Firefox" came out, all of a sudden the muggles realized there was an alternative to IE, so Netscape 4 finally died for us.

The project to obliterate flash and replace it with something open needs a name... And "HTML 5" is as good as any, inprecise as it is.

terra_t··on Why customers leave: the breakdown
-150 infographic blogspam.
terra_t··on Ask HN: Subdomains versus Directories?
From my viewpoint it comes down to manageability. I've got scripts that make it take < 2 minutes to set up a new domain or subdomain on my server, and each one of those is in it's own directory so it can be copied, backed up, or moved to another server easily. If, on the other hand, you're using somebody else's hosting system, subdomains could be a big pain.

For one project of mine, I'm keeping my images on a separate domain so I can switch to a different system for hosting images... For instance, host them on a single-threaded web server or move them to S3/CloudFront.

Similarly, if a project involves different technology than the main site or if it may involve different technology in the future, that's a good reason for a subdomain. If you might want to host it differently, that's a good reason for a subdomain.

I don't know what to tell you about S.E.O.; a few years ago there were some people who made subdomains by the hundred because they thought they got S.E.O. advantages from that -- S.E.'s like keywords in the domain. My understanding is that keywords in subdomains used to be unreasonably powerful, people exploited them, and now there's not so great. S.E.'s also appear to have depth limits on the crawls they do a single host, so multiple hosts might also be a way to get more pages indexed. On the other hand, there are powerful ways to get monolithic sites indexed too. Other people think it's better to put all your eggs in one basket so you can get a higher score on one host.

The truth is that people outside the S.E.'s don't know what the real rules are, and the S.E.'s will change them next year.

terra_t··on Making it Suck
I like the principle of "making things more difficult" in exercise. It's a major principle of training in martial arts: you spend a lot of time training to do things in ways that are a little more difficult than when you really do them so that you find it easier when you really need to do it.
terra_t··on Aiming to Learn as We Do, a Machine Teaches Itself
There's truth in this argument, but it doesn't disprove the concept.

As you say, there's one way to solve the grounding problem: if you've got a system that does a job successfully, you've solved the grounding problem in a particular context. For instance, the meaning of the word "red" can be linked to a certain area of RGB or HSV color space. This is sufficient to find photographs that have a lot of "red" in them or to direct a robot to find "red" blocks.

For a system like NELL to really work there needs to be a feedback loop that tests its knowledge base against reality, that is, it has to do something useful. I don't think this is ever quite going to happen in the academy, but I think commercial systems are generally get closer and closer to the capabilities that Cyc was aiming for.

And even though the "usefulness" of knowledge and ontologies is entirely contingent on a particular application, I think there's enough commonality in the human experience that someday there really is going to be a commonsense knowledge database underlying many sorts of applications.

Cyc is perceived as a failure precisely because it's been commercially successful as it exists. Doug Lenat has supported 50 or so developers for 15 years... From a commercial standpoint, that's a wonderful success. Had Lenat been backed in the corner where the only way to succeed would be to make something that changes the world, he might have changed the world -- however, he found a way to pay his way with a system that's good enough for government work.

terra_t··on Thorium Instead Of Uranium: Solution To Our Energy Woes?
Sorry I missed a factor of 1000. Fast breeders on mined uranium have the same problem as plutonium breeders... Except you can breed plutonium faster (but less efficiently) in LWRs!
terra_t··on How to become a freelance web developer
ho hum.

you're always competing with the person who'll try to "copy facebook for $15/hr"

once you've got more than one customer you'll inevitable get in the situation where, if you were honest, you'd have to tell Customer A that his project is late because you've been working on Customer B's project. If you are honest, Customer A will freak out and odds are 50% or so that he'll dump you right there.

I joined the project management institute, and one day looked at the code of ethics I signed onto and I realized I'd be violating it all the time if I was doing typical freelance work for the typical freelance client.

So I won't.

terra_t··on Thorium Instead Of Uranium: Solution To Our Energy Woes?
Realistic breeder fuel cycles also use a mix of actinides that would be unattractive for bomb makers: you're going to start with fuel that's been through an LWR once or twice, so contamination with Pu240 and Pu241/Am241 will be high to begin with and will just get worse the more cycles you spend in a fast spectrum.

Now, hypothetically, a country like Iran could develop a fast reactor that's optimized for producing plutonium for military purposes, but practically this would be much more difficult than developing a heavy water reactor like the U.S. used at Savannah River.

The isotope which is most problematic in commercial fuel cycles would actually be Np237. It's easy to separate Np chemically from other elements, and Np237 is longer lived than other isotopes of Np, so it can be prepared in a very pure form. Np237 has a low spontaneous fission cross section and a critical mass close to pure U235 -- it would be an attractive material for primitive gun-type bombs.

Note that Np237 is produced from U235 by the chain of absorbing two neutrons, making U237, and then beta decay to Np237. A plutonium-fueled or thorium-fueled reactor isn't going to make as much of it as our current reactors do.

terra_t··on Thorium Instead Of Uranium: Solution To Our Energy Woes?
And that gets us to the central problem of our energy policy.

Utilities are moving towards a short-term thinking model where they install natural gas turbine with a low capital cost and use electricity deregulation to offload the risk of volatile natural gas prices to the consumer. "Renewables" are also driven by short-term motives: making a quick buck by exploiting government subsidies.

Nuclear energy can give us a sustainable energy source the thousand-year timescale (energy-wise, breeders can certainly consume seawater U238 profitably,) but we need to be thinking on the timescale of decades, not the next quarter, to get there.

terra_t··on Thorium Instead Of Uranium: Solution To Our Energy Woes?
The trouble with fast breeders is that you need a huge fissile inventory to get one started.

A study of the PRISM design, a modern fast breeder that uses integral reprocessing to reduce the inventory of plutonium that's sitting around, not fissioning, indicates that if we reprocessed all spent fuel in the U.S. today, we'd have enough Pu to power about 33 MWe worth of breeders, about 1/3 of the current fleet.

It comes down to energy vs power. "Fast" breeders can tap a huge energy resource, but they can only produce a low level of power because they tap that energy slowly. That's the whole reason they use highly reactive sodium as a coolant; to reduce fissile inventory per unit of power, you need a coolant with incredibly high heat conductivity. Lead is a much easier material to handle, but a lead-cooled fast reactor has 1/3 the power density of a sodium reactor, which means 3x the fissile inventory.

Fast breeders, therefore, can sustain energy production over a long period of time, but they can't drive a rapid expansion of nuclear energy, like the 5-fold increase we'd need to replace fossil fuels and stop global warming -- it becomes hard to justify the economics of a plutonium economy in the same way... You have to be thinking more than "seven generations" ahead to see the economic boon.

Thermal reactors effectively trade a moderator material (water, graphite, etc.) for (relatively scarce) fissile material. Thermal reactors (even today's LWR) extract 10-20x as much power from fuel than do fast reactors with less aggressive design (although the LWR extracts only 2% as much energy, in the long term, as a breeder could.)

A thermal breeder, based on thorium, could provide the best of worlds. With small inventory, it's possible to meet high power requirements, but by using abundant thorium instead of rare U-235, be able to sustain that power for thousands of years.

terra_t··on Ask HN: Underperforming, feeling stuck
You can be depressed without realizing it. Part of you can think that you like your job, while another part of you hates it. The best way to detect burnout isn't to ask "do you feel burned out?" but to ask "can you get the work done?"

I was stuck in the code mines a few years ago; I'd always find some way to convince myself I was happy about what I was doing, but then I read this blog about SEO and started dreaming about making my own web sites that make money...

At that point my mental health really cracked and I couldn't focus on 9-to-5 work at all anymore. It was a sign that I had to change what I did, and gradually, that's what happened.

terra_t··on Ask HN: Is anyone still excited by Diaspora?
Couldn't care less.

Systems like Facebook depend on a dual sided market. There's nothing to read if nobody is writing, and why write if nobody is reading?

I had a friend years ago who ran a multimedia-oriented site that only took open formats like Ogg Vorbis. Although hypothetically, Ogg Vorbis is a free format which might be accessable on more operating systems and devices, practically, publishing in Ogg Vorbis is like speaking in Pig Latin. MP3 files just play for most people on the devices and OSes that they actually use w/o any trouble.

Even though there is an articulate group of people who are concerned with privacy and other philosophical problems with Facebook, the fact is that most people just don't care -- if you ~really~ cared about those things, you probably wouldn't be oversharing your life on a social network anyway. In fact, from a 2010 perspective, Facebook's early policy of only publishing to your 'close friends' looks medieval: the industrial spy, potential employer, litigator, or other person who wants to use information to harm you is going to extract it one way or another, from such a network. If you really want to keep something a secret, keep your mouth shut.

terra_t··on What developers think when you say "Rock Star"
+10

I worked at one of these places, as a contractor. They never offer health insurance. They buy a lot of pizza and junk food, and give out lots of cheap praise, but will never send you to a conference or otherwise contribute to your well-being or professional development.

These guys offered me a permanent position, and I turned it down for a real job.

terra_t··on Could jgc.org be monetized?
Sure, computer science students get paid a lot, relatively, when they get out of college.

The trouble is that, unlike white collar professions, pay for programmers doesn't go up as they mature -- instead, it goes sideways or down.

terra_t··on (How to Write a (Lisp) Interpreter (in Python))
Corrected Title:

(HowTo (In (Write LispInterpreter) Python))

[if the title actually was LISP syntax, everybody would be programming in LISP today]

terra_t··on Michael Stonebraker: Why Enterprises Are Uninterested in NoSQL?
you can write

select sum(x=1) from y;

in standard SQL, Oracle, and Mysql. True == 1 in standard SQL, False == 0. Both pgsql and Microsoft SQL server define a nonstandard boolean type that requires you to add a cast or an if statement, bulking up the query.

I did a shootout of mongodb and the three RDBMS systems (!Oracle) I mention for building a system to represent data from Freebase. It was possible to make a VARCHAR(4096) in mysql and only index the first 64 characters which meant I could map freebase types to mysql tables without running into index limitations -- I wanted the better GIS capabilities in pgsql, but I didn't want to double the size of my tables and queries to be able to handle strings losslessly.

terra_t··on Could jgc.org be monetized?
The first thing to do is to divide your revenue numbers against your page views: you get an eCPM of 22 cents per 1000 page views from adsense and 24 cents from amazon, for a total of 46 cents.

That's pretty good for a programming blog, in fact, the best I've seen. Programming blogs don't monetize well, for quite a few reasons:

* Many programmers refuse to buy software and other programming tools * Outside of a few special enclaves, programming is a relatively low status and low paid occupation. (Yes, you'll find people in the valley who get paid $120k a year, but every one-horse town has a job shop that wishes it could pay programmers $30k a year and may or may not offer health insurance... Sure they get turnover, but they think they'll make it up in volumes) * Programming blogs attract a lot of people from India and other developing countries who are very poorly paid, cheap and don't have credit cards anyway * There's just too many programming blogs... The cost ofadvertising is set by supply and demand if there's a lot of inventory and not enough advertising, the price is going to be low

What to do? You may (or may not) be able to double your revenue by making your ads more intrusive; if you make your blog look like hell, you'll lose the social traffic, however.

The best things to do are:

* Not blog. Permanent content that attracts search traffic pays better than blogs that are driven by social traffic. Blogs aren't all bad, because (1) you write a lot of content and (2) the social traffic results in a lot of links which helps you in search, but a search-first strategy is better still. * Pick another topic. One reason so many people make so little money on ads is that 80% of content creators create content about 20% of the topics. If you put on a blindfold and threw darts at a thesaurus, you'd find something that monetizes better than the average fool who "follows his passion"

Oh yeah, if you need to "make money fast" don't forget the value of your domain name. You could get $5000-$15000 in one lump sum for a three-letter name. That's 8 years to two decades worth of your ad revenue...

← PreviousPage 4 of 10Next →