tl;dr: A developer wrote a SIMD-accelerated FAST feature detector for the ESP32-S3's 128-bit vector instructions, roughly doubling throughput from 5.1MP/s to 11.2MP/s and processing QVGA frames in ~6ms. Key obstacles included the ISA's lack of misaligned reads, signed-only 8-bit comparisons, and saturating arithmetic — worked around with register slice extraction and an XOR-with-0x80 trick to convert unsigned pixels to signed range. The result enables real-time 30fps VGA computer vision on a $2 chip.