As soon as I read the requirement, I knew that I didn't want to upload just any self-portrait. I knew that I wanted to do something incredibly random, or something incredibly impressive. Then I came up with this idea:

1. Drive out into a scenic, deserted area. Most likely in my new home state of Oregon.
Plan: This task is to be completed in conjunction with a geohashing expedition. I will attempt to pick a geohash which is accessible, yet in a unique and interesting area (compared to where I usually go on expeditions).
Execution: I drove out to someplace east of Bend, Oregon. This is about 200 miles from where I currently live.

---

2. Find a relatively flat, straight stretch of mostly unused road about 100m long.
Plan: If I am east of the Cascades, this becomes a much more likely event, but all in all this state doesn't lend itself all that well to flat, so this may just have to be somewhat of a compromise.
Execution: I drove around local roads, with my GPS unit yelling at me to get back on the highways, until I found a good location with a decent set of mountains in the background.

---

3. Ensure that there is a turnout on the side of the road that would provide a good angle to get the road crossing the field of view of the camera, but mostly down the road.
Plan: This means no trees along the side of the road where there is a turnout. Again, east of the Cascades is the most likely place for this to be viable. I would estimate that the hood of the car needs to be about 10 yards from the edge of the road for this to look right, but it is most likely going to be a trial an error thing, but it should be one I can gauge before the following steps.
Execution: As it happened, I didn't need to be too far off of the road to get it to work correctly, only a few feet.

---

4. Ensure that from the turnout, there is a scenic background behind the road.
Plan: The means few to no trees along the opposite side from the turnout. Once again, east of the Cascades beckons.
Execution: See step 2

---

5. Park and use the hood of my car as a makeshift tripod for the camera.
Plan: I really wish I had a tripod, but I am too cheap to invest. My car hood should be good enough for these purposes though.
Execution: Instead of the hood of my car, I found that a rock on the ground turned out to give a better angle.

---

6. Set the timer on the camera for 15 seconds.
Plan: I am definitely going to want a bit of time to get into position to begin the sprint. If needed I can always have the next section take more pictures and not worry about the timer.
Execution: I didn't do this, reasoning below.

---

7. Set the camera to take a bunch of photos 1 second apart starting once the timer goes off.
Plan: Yay time-lapse photos! My goal here is for 15 pics of me sprinting down the road. At that point I should be somewhere about 100m away, meaning my that I am unlikely to be seen all that well in the picture.
Execution: My camera's default firmware only will take 10 photos at a time. While this would have been enough given experience, I didn't know this at the time. Instead I used an open source firmware which would just take pictures every 1 second until explicitly stopped.

---

8. Position myself as if in sprinters' start blocks at the left of the frame, and sprint down the road while the pictures are being taken.
Plan: The above
Execution: The above, but not at full sprint. It turns out that too many of the pictures are too far down the road to give a good representation of what I wanted. Instead I faked it, and sort of ran in a sprinting fashion, but not at sprinting speeds.

---

9. Cut the background out from all of the photos with me in them.
Plan: This is to obtain all of the action photos of me running. They will then be combined into a single photo in later steps.
Execution: I actually found that it was better to not do this.

---

10. Set a different transparency level for all photos of me. The level of the transparency will be determined by when in time the picture was taken.
Plan: The earlier the photo in time, the more transparent. The last photo will not have any transparency at all.
Execution: This was accomplished by taking each photo, and pasting it into a "blank" as a new layer. Then I set the transparency of that layer according to an algorithm. That algorithm is as follows, for n=0-(number of images minus 1), where n=0 is the last image taken, the transparency was set to 0.8^n * 100 %

---

11. Start with the last photo, and paste in all of the photos of me in reverse chronological order.
Plan: This is to make it seem like the after-images of me are slowly fading into oblivion.
Execution: I actually went the other way around, pasting each partially transparent image onto the last as a new layer, cutting out a hole for the next image of me, and repeating all the way through all of the pics.

---

12. Admire the (hopefully) awesome results.

The final result isn't perfect (one might even call it aperfect), but I feel that it is good enough.