2018 review — quiet year, Zigbee goes local, the doorbell got scary
End of 2018. The quietest year here since 2012 — and I called it.
Scoring the 2018 forecast
| Prediction | Confidence | Outcome | Verdict |
|---|---|---|---|
| Zigbee comes off the SmartThings hub onto a dedicated coordinator | 60% | ConBee + deCONZ in October | ✓ |
| Complex automations move off YAML to a flow tool | 65% | Node-RED, February | ✓ |
| Smart-home camera category consolidated by a big acquirer | 55% | Amazon bought Ring in February | ✓ |
| MQTT becomes the house's internal message bus | 50% | Mosquitto broker, alongside Node-RED | ✓ |
| Posts slow to ≤ 3 for the year | 80% | Three posts. Hello. | ✓ |
5/5 ≈ 100% — but the honest read is that several of these were within my control, so it's less "good forecasting" than "I did the things I planned." The Ring acquisition was the only outside call.
What changed in 2018
Three moves, each pushing the same direction — local-first:
- Node-RED (Feb) — the automations got a brain that can hold state, and a visual debugger I trust.
- MQTT broker (Feb) — the house got a nervous system. Everything publishes; anything can subscribe.
- ConBee + deCONZ (Oct) — the Zigbee mesh came off the SmartThings cloud-bridge onto a coordinator I own.
And one move I didn't make but thought hard about: Ring. Amazon's acquisition crystallized why I keep cameras local. Didn't buy one. Won't.
What worked
- The MQTT bus. Watching
home/#scroll by is still the clearest window into what the house is actually doing. Adding a device is now "point it at the broker," not "rewire the integrations." - Aqara sensors on deCONZ. They drop constantly on SmartThings and behave perfectly on a local coordinator. Cheap, reliable, mine.
What didn't
- Z-Wave is still on the SmartThings hub. The migration I keep deferring. The hub now exists almost entirely to be a Z-Wave radio, which is an expensive antenna.
- Documentation debt. Three posts means a year of changes that live only in git commit messages and my memory. The day job won that trade, but the journal suffers.
The day-job thread
For the record on why it was quiet: I spent 2018 building a connected medical device — BLE, phone-as-gateway, a compliance surface that makes "move fast" illegal. It's the most relevant work I've ever done to this hobby, and the least time I've ever had for the hobby because of it. The patterns cross-pollinate constantly: the MQTT event bus at home is the same shape as the telemetry pipeline at work, just without the auditors.
Forecast for 2019
| # | Prediction | Confidence |
|---|---|---|
| 1 | Z-Wave finally comes off the SmartThings hub | 55% |
| 2 | A major cloud platform kills a beloved integration, proving the local-first thesis | 65% |
| 3 | MQTT becomes the primary integration path, not a side bus | 60% |
| 4 | Posts stay ≤ 3 (day job runs through 2019) | 75% |
| 5 | The house survives at least one more cloud outage untouched | 70% |
What's next
Z-Wave off the hub — for real this time. And I have a suspicion that 2019 is the year some big platform unceremoniously kills an integration thousands of people depend on, and everyone re-learns the lesson I've been writing down since 2017. The house is ready for it. Most aren't.
Six years in. Quietest year yet, most architectural progress yet. Local-first is no longer the goal — it's most of the way to being the default.