diff --git a/RETROSPECTIVE.md b/RETROSPECTIVE.md
new file mode 100644
index 0000000..95aa3c3
--- /dev/null
+++ b/RETROSPECTIVE.md
@@ -0,0 +1,48 @@
+# a short retrospective on the project
+
+## why?
+
+I intended to learn about how to use FPGAs, and figured that
+graphics would be an easy way to get feedback on my progress
+(I also just like computer graphics).
+
+## challenge
+
+I used a [pico-ice](https://pico-ice.tinyvision.ai) development board
+for this project. Since the iCE40UP5k (the FPGA) is not
+attached to the USB data pins, I needed to communicate
+between the microcontroller (a raspberry pi RP2040 chip)
+and the FPGA. For the amount of data to be transferred,
+even at 64MHz, SPI would be too slow. So, I had to design
+a custom protocol for communication between the two chips.
+Additionally, the were not enough pins for two one-way
+data buses. Thus, a bi-directional bus was needed.
+
+## how?
+
+The RP2040 microcontroller has a cool feature called PIO, short for
+'programmable input-output'. Essentially, it is a tiny state machine
+embedded in the microcontroller to manage the state of GPIO pins
+in order to avoid bit-banging, saving CPU time. I had to learn
+how to use this feature (and many others) of the RP2040.
+Raspberry Pi has good documentation [available online](https://datasheets.raspberrypi.com/pico/raspberry-pi-pico-c-sdk.pdf).
+With this, I was able to create a bi-directional bus without
+the use of bit-banging in order to transfer frame data from the FPGA to the RP2040.
+
+## results
+
+Overall, the project worked better than expected. The video displayed fine over
+USB Video Class, but the bi-directional data bus had synchronization issues, so I
+had to lower the clock speed of the FPGA so that the RP2040 could over-sample.
+Furthermore, the implementation of the mandelbrot fractal renderer could use
+some work: it doesn't use pipelining at all.
+
+Here is a picture from the 'camera': ![mandelbrot fractal](./mandelbrot.png)
+
+## evaluation
+
+Looking back on this project, I realize I should have invested
+more effort into designing a good interface between the two chips
+on the pico-ice board. One of the major pain points of this project
+was the improper synchronization over this bus. Overall, though, I
+learned a lot about SystemVerilog programming and microcontrollers.
diff --git a/mandelbrot.png b/mandelbrot.png
new file mode 100644
index 0000000..a22389d
Binary files /dev/null and b/mandelbrot.png differ