> Temporary files usually start with a dot or a dollar-sign.. to make sure that Nginx never serves any files starting with either of those characters...
Have these disputes changed YC views regarding single founders?[0]
I think YC has always favored founding teams with a dominant leader[1], but I wonder if these break-up experiences have pushed it a bit further towards single founders with strong teams.
I guess there is no change, since the advantage YC sees in multi founders is probably only for the time interval before these disputes, at which point they are already doomed. But what I may be missing: could amicable breakup ever help a company avoid the dead pool?
Another thing I find missing from the discussion: is this move also meant to push new founders to profitability faster? PG seems to be mentioning it more lately ([2], [3] etc.).
[0] Honest question- I hope it does not spark another single vs. multi founder debate.
I find it interesting that they awarded the prize to Shapley after Gale's death[0]. I always thought that there was some rule that prevented them from awarding the Physics Nobel Prize to Aharonov after Bohm died[1].
While I may agree that the US should do more to encourage immigrant entrepreneurs, I do not think that the author chose a good openning example.
The author points to statistics about tech companies started by immigrants and presents Desai as part of the team that developed new technology, but Desai studied MBA and he was only doing an administrative job at IR Diagnostyx[0].
Moreover, the openning example does not seem to be precise; the author claims that Desai was not given an opportunity to start his business, which does not seem to be aligned with Desai being affiliated with IR Diagnostyx from 2009 to Feb 2012[1]. Furthermore, upon graduation in 2009, the US did give Desai 12 months of OPT[2], for which self-employment does qualify. In fact, if Desai had studied Science/Tech/Engineering/Math, he would have been give
an an extension of 17 months of OPT[3], for a total of ~2.5 years to work on his business.
He's just refuting Khosla's agument and pointing out that if VCs were really repelled, then the valuations could not be high (since these are determined by the demand of the market).
IMHO, it's best to prototype in Octave and then build in python. I find that the Matlab/Octave syntax is too focused on linear algebra, so it's better for small prototypes (and for people coming from non-SW fields). For big projects, I prefer the 0-based arrays, more than one function per file, and all the rest of the python goodies. I estimate that 70% of my time is usually spend preparing the data (e.g. parsing xml, or some other files, etc), for which I find python more suitable.
In fact, I usually work with them side by side, testing ideas in Octave, then implementing these pieces into a large python project.
Edit: this has also been discussed here before, e.g.
A lot of the current post seems to be taken straight from there. Compare, for example:
> in Alzheimer’s patients their short-term memory hardly works at all, but the long-term memory is still active. They know the green and yellow bus sign and remember that waiting there means they will go home.
with Goebel's quote:
> 'Their short-term memory hardly works, but the long-term memory is still active. They know the green and yellow bus sign and remember that waiting there means they will go home.'
> But then if you think about it for two seconds you have to wonder why we want a good signal of these students’ ability. This is not assessment for accreditation so who cares about getting such incentives right? What one surely wants are problem sets that signal to the student whether they had mastered the material or not
First, I think their business model (talent discovery and job placement) depends on good assessment of students' ability. At least, this is Udacity's business model[1].
Second, I think that the OP's assumption that this is not for accreditation is wrong. I think they do want Coursera's statement of accomplishment to be valuable on it's own (even if "Stanford" is not mentioned in it). For instance, if I were an employer, I'd hire anyone with Coursera's statement of accomplishment in a challenging class as the Probabilistic Graphical Models class [2]. Of course, the only problem is, as the OP mentions:
> For online courses, no one has cracked how to verify whether an identified student is the same person as the one doing the assessment.
[2] The class started with 44000 students [3]; by the 4th week there were about ~2000 left [4]; my guess is there are around 1000 left after the crazy 5th assignment... Though, One may argue that these statistics in part are due to rough edges and cryptic instructions in some of the programming assignments and quizzes.
I agree. The PGM class is much more challenging than the machine learning and algorithm classes, even though (or perhaps, because) Tim and Andrew are doing such a terrific job with the videos of ML and Alg, respectively. One also notices a steep decline in the number of PGM quiz attempts in the published statistics. However, I think that some of the difficulty of the PGM class should be attributed to the rough edges in the programming assignments. We rely heavily on the forums to collaboratively decipher them (as well as some ambiguous questions on the problem sets). But then, this also shows that Coursera's platform enables positive student interaction...
I probably used every curse I know while working on the 5th PGM assignment, but I am the first to admit that I did learn a lot, and I'm sure they will fix the rough edges for the second run of the class. Moreover, the entrepreneur in me is very inspired to see fresh-from-the-oven code. Sure, I've already seen apps that were lacking some features because they were released fast, but here I have an opportunity to actually see some fresh code (since we are basically asked to "fill in the blanks" in Daphne's implementation). I admire Daphne's entrepreneurial courage to publish something even if is is not 100%.
Also note that the PGM class has many more pointers to recent research in the field [1], which I think the OP would find interesting.
Finally, I think the OP is missing a bit of the bigger picture - comparison of undergraduate studies in Israel and the US. I could probably write a long post about this one day (I've taught science undergrads in both college systems), but at least I should point out a couple of things. In Israel, there is much more focus on the major; about 95% of the classes are in the major field of studies. (One chooses her/his major before applying to college). There are (almost) no GE classes [2]. There is probably no one attending CS classes who does not major in CS (or double major). There are advantages and disadvantages of each system. But as a result, you can put more challenging content into CS classes in Israel.
Therefore, it would probably be more fair to compare a CS class from the Technion to upper-division or first-year grad class. That is, instead of comparing it to Stanford's CS161 (or any other 100-199 classes), it would be more fair for the OP to compare it to the level of CS228/CS229A (or almost any other 200-299 classes).
[1] Not all of these are plugs to Daphne's research- I even remember some pointers to Thruns' papers. (he's sort of a competitor.)
[2] Perhaps this is to allow 3 years college, and save some time we spend serving in the army.
There could be other time-based methods to encourage students, rather than hard deadlines. For instance, suppose the class had a bag with unlimited (or large) number of question, and required each student to score 100 points to pass. Students would have unlimited number of attempts at quizzes/assignments, but their score decays with time (think HN submission ranking decay). That is, if you worked hard and scored 70, but then you were busy/away for a month, your score drops to, say, 50. So you are encouraged to finish the questions in a timely manner, but on the other hand, you can also make up for that missed month by putting more work when you are back.
With this method or a something similar, you do not "lose everything" when you miss a deadline. IMHO, "losing everything" for missing a deadline is part of the reason for the (exponential?) decay in the number of students participating in Coursera's PGM class[1,2].
[1] Based on PGM's "quiz's highest score" statistics, the total numbers of participating students are 6950, 3500, 2650 for week1, week2 and week3 respectively (other weeks' stats are not complete as of this writing).
[2] Another interesting stats is the ratio of perfect scores to all other scores: 25.2%, 34.3%, 49%. This could suggest that the deadline system filters out most students but the very top ones (assuming uniform difficulty of quizzes, and uniform quality of lecture videos/slides, etc.). This may not be what you want from an educational perspective. And, yet it may actually be what you want to support a business plan similar to Udacity- discovering talent. (see my comment below)
Regarding Udacity's business plan:
In June Thrun "took the next step: cofounding KnowLabs... He pulled in David Stavens... as CEO… Thrun decides that KnowLabs will build something called Udacity… Stavens is thinking about potential business models. Though Thrun cringes at the notion of charging students, people might eventually pay for add-ons—say, TA services, study aids, or offline materials. He also considers other revenue streams. Near the end of the term, he emails his top 1,000 students, the ones with perfect or near-perfect scores on homework and tests. The subject: Job Placement Program. Thrun solicits résumés and promises to get the best ones into the right hands at tech companies, including Google. A recruiter who places a hire typically earns 10 to 30 percent of an engineer’s first-year salary, which might be $100,000. Stavens figures he could charge much less. After all, KnowLabs discovers talent in the course of doing business."[1]