I remember my start thoroughly. I had been over at a friends house and I saw that he had a Rubik's cube (Wal-Mart brand). He could solve one face of the cube pretty much at will and I became angry that I couldn't even solve one face of the cube. So, I borrowed it and I looked up an online tutorial on nerdparadise.com (it was a horrific tutorial in retrospect, with no proper notation and strange orientation although it did teach a layer-by-layer method). After becoming disgruntled with this tutorial, I went to the website of Lars Petrus. After failing there (LOL I SUCKED) I ran across Tyson Mao's layer by layer video tutorial and mastered each step of all the videos. After this, I completed my first solve at my dinner table in front of my parents. It was about 5 minutes long. With mere practice I grew to the point where I could get about 2:30. At this point, I decided it was time to learn more steps. I replaced the cumbersome "1-middle layer edge at a time" with the "keyhole" method of edge insertion (which, actually, every beginner should be taught...) and then only used the edge insertion algorithm for the final middle layer edge. After this, I learned the 7 OLL algorithms for cases where all edges are solved. The "keyhole" step got me to about 1:10 (a very frustrating period, when one is trying to get their first sub-minute solves!), the 2-look OLL algorithms got me down to the 45-50 range. I won a pretty fun little competition at our middle school with this time (by far). By this time, I had already put silicon lubricant in the cube and had replaced the stickers with CubeSmith stickers. From there I practiced an intuitive F2L. From then on, my speedcubing time has been almost directly proportional to how well I execute the cross+F2L step. I am now down to the 18.xx. I plan to learn full OLL this summer (since I am already pretty close).