Dispatches from the work in progress
How thoughtful choices in color, material, finish, and light shaped Everbowl's identity long before its technology came to life. Introduction When people pick up a product for the first time, they don't start by evaluating its spec sheet. They make an almost instant judgment based on
Rebooting a device sounds harmless until you remember what it's for. When a pet-monitoring station restarts, it isn't watching anything — no weight readings, no captures, no heartbeat. A safe operating-system update costs a reboot, and a reboot costs a small hole in a pet's
Most monitoring conversations are about thresholds and routing - how many failures before we restart something, how long before we page someone, who gets woken up at 3am. That's real work. But it quietly assumes something worth verifying: that the signal you're measuring actually means what
Building products that protect themselves without turning every fault into a service nightmare A hardware protection circuit is usually designed with one question in mind: "How do we prevent this fault from damaging the board?" That question is important, but in real products it is incomplete. A product
Somewhere in most mature codebases there are three ways to schedule a recurring job. Nobody chose that. There is no design document proposing three schedulers, no meeting where a team weighed the options and landed on "several." Look closely and each one had a good reason at the
Our IMU bark classifier now has 11 training generations behind it. This post covers what changed between v8 (our last reported state) and v11 (live as of August 5), introduces a real ambient audio clip where the audio classifier was completely fooled but the IMU was not, and explains the
How our "Stage 2.5" design freeze helped eliminate scope creep while keeping Everbowl moving toward production. Introduction Every hardware team eventually has to answer a hard question: are we done designing? The temptation is always one more improvement. A bracket that could be a little stronger. A
The device is on the table. Its LED is breathing. Your phone's system Bluetooth settings can see it. Your app scans, and finds nothing. This is the single most common way a Bluetooth Low Energy integration fails, and it is almost never a bug in your code. It&
Every radio on a pet tracker has a horizon. Bluetooth reaches across a room, maybe a house on a good day. The moment a dog slips through a gate and trots down the street, that link is gone — and that is precisely the moment a pet parent most needs to
A camera is the obvious way to know whether a pet is in the room — and often the wrong one. It fails in the dark, it's blocked by a blanket or the side of a bed, and it puts a lens in the most private corners of someone&
Every system I have run at scale has taught me the same lesson in a different costume: failures don't hide in the components you own — they hide in the seams between them. This is the story of one such seam. Our notification pipeline had a monitored backend on
Every RAG tutorial follows almost the same flow: convert the user's query into an embedding, retrieve the Top-K most similar documents, and pass them to the LLM. The part that's usually skipped is a simple question: Why K? Why retrieve 3 documents instead of 1? Why
In the first post in this series we described a rule-based sleep detector for our collar wearable: bucket the IMU stream into one-minute epochs, score each epoch by how much the acceleration is changing, and call a long enough run of quiet minutes a sleep bout. We ended that post
Keyspaces looked like the safe choice. Cassandra was already running in the stack, so letting Temporal reuse it meant no new engine and everything staying in-account. Then we hit the wall: Temporal's Cassandra schema depends on conditional batch writes, the lightweight-transaction mechanism behind shard-ownership fencing, and Keyspaces doesn&
A pipeline written as a script is a promise that nothing will go wrong. Modeling it as a directed acyclic graph is what you do after something does. Every multi-step pipeline starts as a function. Fetch the data, clean it, score it, summarize it, store the result, notify someone. Six
