wonsungi

I agree Tomayko's method is very helpful, but I came to the conclusion it's better to just commit as often as possible.

- Multiple changes that affect the same line(s) are very difficult to pick apart.

- It is easier to write a meaningful commit message while the change is fresh in your head; otherwise a lot of time is spent reviewing your changes.

I suggested having a "Compose" button instead of a link in 2004. I noticed no one, including myself, was looking for a link when they wanted to write a new email because the other active Gmail actions like "Search" and "Delete" all had buttons.

Looking for a co-founder in Seoul? Send an email to [email protected]

Otherwise, I found this list of over 60 start-ups in Korea [English]: http://codeinkorea.com/46

Direct link [mostly Korean]: https://spreadsheets.google.com/ccc?key=0AiIm7SFFVfcFdEhpRGV...

Also highly relevant to why 165 University would be lucky: downtown Palo Alto is within walking distance from "the quad" on Stanford campus (just down Palm Drive).

I was told FaceBook HQ was at 150 University Ave; just down the street: http://maps.google.com/maps?f=d&source=s_d&saddr=165...

But they may have very well leased space in 165 University Ave; I was also told by the time they moved, they had a complex shuttle system to get between their many office spaces in Palo Alto.

Try http://www.bkmks.com: (Google) search box at the top; organize your favorite links below via drag & drop. Fast and simple.

I just wish it had tabs and easy/multiple delete (I hacked together a javascript bookmarklet to work around the multiple-clicks-to-delete-a-single-link issue)

I often take photos of textual signs for use later. Take it to the next level: OCR text from the built-in cameras. A friend suggested OCR from live video was even possible (on an iPhone!) My vision was being able to enter text by scribbling something on a napkin, and OCR'ing the text in.

Another cool method is Dan Bricklin's Note Taker (http://news.ycombinator.com/item?id=997264) Just need to add OCR.

Now, my experiments with OCR algorithms didn't work too well. Apparently the training data is the most important component.

Handwriting recognition is also possible, but still images lack the very useful stroke info.

Transcript: http://www.february-7.com/features/conan.htm