research

not published work. I'm two years into a degree, not a phd. what lives here is the writing I do while learning something properly: what a technique actually does, what I got wrong, and what it cost me to find out. the post-mortems below are the second half of that, one per bug that actually hurt.

post-mortems

every one of these is a bug that cost me real hours in something I built. written up afterwards: the symptom, how I chased it, what was actually wrong, and what I changed about how I write that kind of code.

• oracle_of_delphiIt Was Never the Model

  four failures in a local voice assistant, at four different layers. two of them presented as the language model being inadequate. none of them were.

• custom_mem_allocThe Block That Was Too Small to Free

  an intrusive free list stores its next pointer inside the freed block. ask it for four bytes and the pointer does not fit, so freeing corrupts the block next door.

• nexus_dbThe Delete That Did Not Delete

  the flush path dropped tombstones it judged redundant, using a definition of redundant that only looked at the memtable. delete a key that already reached disk and the deletion evaporated.

• nano_matchThe Order That Was in Two Places

  a cancelled order went back to the pool while a price level still pointed at it. the next order reused the slot, the book crossed, and nothing crashed.

• nexus_clusterA Cluster That Could Not Keep a Leader

  i copied the election timeout from the raft paper along with its assumption that the network is the slow part. on one machine there is no network, and gc pauses were wider than the timeout.

what's next

• tensors & preprocessing

  the same array work, reshaped to (batch, channels, height, width): resizing, normalization, augmentation.

  in progress
• convolutional networks

  the direct continuation: same sliding window, kernel values learned instead of chosen by hand.

  next
• detection & tracking

  where the motion detector gets replaced by something that holds identity across frames.

  queued