Full-stack wireless tech for IoT & UAVs. 100% self-developed TPUNB® LPWAN | RF • LoRa • Wi-Fi • BLE modules | DTUs • gateways • ground stations
HongKong
- Tweets 176
- Following 178
- Followers 46
- Likes 85
2030: Your kitchen AI demands a 30% tip.
Your wallet AI negotiates it down to 15%.
Your marriage AI recommends paying 50%.
Welcome to the agent economy.
Can't wait to select "No tip" for my wife's cooking.
ALT An Apple Home View device in a kitchen displaying an interface labeled "Would you like to leave a tip?"
1
20
The real shift isn't just bringing every smart device into one ecosystem.
It's moving from connected devices to coordinated intelligence.
Interoperability lays the foundation. AI agents could turn understanding into real-world action.
That's the direction we're exploring with TP Grow.
苹果终于对智能家居动真格了，但它没自己造锁造摄像头，而是拉着 LG 和施耐德直接在现场立了块新招牌 Apple Home Collection。
以前折腾 HomeKit 的老哥应该都懂那种折磨，买三样设备要装四个 App、注册三个账号，断个网全家设备直接瘫痪。
这次苹果把规矩全改了：
不用再下载第三方软件，不需要注册杂七杂八的厂商后台，拿 iPhone 碰一下直接配对完成。
LG 负责制造门锁、双摄门铃与监控摄像头，施耐德负责全屋温控器。
门锁利用 iPhone 和手表里的 UWB 芯片，走近门口自动开锁，
温控器吃透了 Thread 和 Matter 标准，家里宽带断网，局域网内的设备照样互联互通。
表面上看是挑了两个巨头搞联名，背后真正的杀招是苹果把最重的制造包袱全甩了出去，自己死死把住配对入口、原生界面和端到端加密。
合作厂商碰不到你的家庭隐私数据，用户也不必忍受一堆臃肿的后台。
这一整墙的门铃和温控器只是前哨站，配合现场同步发布的 HomeView 智能中枢，苹果正在用最轻的姿态，把系统级 AI 真正送进你的客厅。
1
4
Most IoT devices: "Something's wrong."
This little guy: "Say less."
From sensing to acting. That's where things get exciting. 👀
This little machine spots a fire and puts it out by itself. No human. No app.
Watch it work. A flame sensor catches the fire → the light flips from green to red → a pump kicks on → it shoots a jet of water straight at the flame.
It even aims. A servo swivels the nozzle toward the fire, so the water lands where it's actually burning instead of spraying everywhere.
And he built the whole thing at home. The whole setup fits on a dinner plate: a board, a sensor, a pump, a water tank and a battery.
Here's the part that gets me: detect, aim, extinguish on its own is how premium fire suppression systems work.
The expensive stuff built into real buildings. He pulled off the same logic on his desk at home.
The big version costs a fortune and protects a whole building. This one runs on parts cheaper than lunch.
4
Four pieces of tape. One smart home dream.
Meanwhile, we've got a little 4-inch idea brewing. 👀
1
30
They gave AI agents a body.
What if we gave them a whole world of devices to interact with?
We're exploring that side of physical AI too. 👀
Earlier this year, we were just curious:
"Can we bring AI agents like Claude Code, Codex, and Hermes into the physical world?"
Give them eyes, ears, a voice, and a body they can control to understand the world and act in it.
That’s how Lamp started: autonomous.ai/lamp
1
25
Giving AI agents a voice is cool.
Giving them a way to interact with the physical world? That's where things get interesting.
We're working on something. 👀
Introducing Interface, the primary hardware for the agent era.
Interface lets you control all your agents from your hand.
$99 for early adopters. Ships in January.
1
7
The future of smart homes isn't about adding more tech. It's about making the tech disappear into everyday life.
That's the part we're excited about.
NEW: Apple - in an unusual partnership with LG - is readying a Doorbell, Deadbolt Lock, Thermostat, Indoor Camera, Outdoor Camera, Floodlight Camera and more to rival Amazon and Google as part of its major smart home reboot. bloomberg.com/news/articles/…
5
6 inches for the smart home? 👀
Meanwhile, we've been playing around with 4 inches on our side.
8
Same problem. Our answer.
No Wi-Fi. No cellular. Still connected.
We’ve been field-testing our off-grid communicator in real outdoor conditions, with line-of-sight communication reaching up to 10 km in open areas.
When the network disappears, the conversation doesn’t have to.
#OffGrid #Wireless #OutdoorTech #IoT
Your apocalypse phone doesn’t need a cell tower.
meet the Doomsday Phone, a DIY communicator called Chatter, powered by an ESP32 and LoRa radio
it sends text messages directly to another compatible device within range
no SIM card
no Wi-Fi
no mobile operator
a built-in screen and physical buttons mean you don’t need to pair it with your smartphone
the video claims up to 10 km of range, depending on terrain, obstacles and antennas
its strength is simple: two powered devices can communicate without relying on a working mobile network
for a camping trip, a remote location or a network outage, that could mean getting an “i’m safe” to someone who needs to hear it
1
26
Every time I think we’ve found the limit of the ESP32, someone does this.
512 WS2812 LEDs. 240fps. Controlled from a Nintendo Switch.
There’s always a little more performance hiding in these tiny boards.
#ESP32 #Embedded #LED #Maker
78
Camera sees it. LoRa reports it. Edge gateway decides what happens next.
Why send the whole video when the useful part is only a few bytes?
What if your AI camera didn't need Wi-Fi or cellular?
Imagine a camera running lightweight vision at the edge, detecting people, vehicles or objects locally, then sending the important results across a LoRa mesh.
Instead of pushing the entire video feed, the device could send AI-generated metadata, alerts, timestamps and selected low-resolution frames between nodes.
That means much lower bandwidth while keeping the system useful in remote areas, farms, disaster zones or large sensor networks.
The interesting engineering challenge is making the camera, edge AI, ESP32-class hardware and LoRa mesh work together within tight power and bandwidth limits.
Who wants to build a battery-powered AI camera that can communicate without Wi-Fi?
2
36
RF detection isn’t the only game in town. Acoustic + edge processing + wireless backhaul makes this a pretty interesting sensor node.
What if an ESP32-S3 could hear a drone before you could see it?
Agam Rossen's VolAnti is an open-source acoustic drone detector built around an ESP32-S3, four MEMS microphones, e-paper display and LoRa radio.
It analyzes the harmonic signature of spinning propellers instead of looking for radio signals, making it interesting for detecting fibre-optic FPV drones that can leave the radio spectrum silent.
The concept could be useful for perimeter monitoring, wildlife & environmental monitoring, industrial site awareness, event security, search and rescue operations and early-warning sensor networks, especially where visual detection or conventional radio-based detection is difficult.
The creator reports 104 m measured detection, 0.23 s alert latency in its fastest detector, and roughly 18–22 hours of battery life.
What I really like is how much signal processing, embedded hardware and wireless communication has been packed into a small device.
Would you build an acoustic sensor like this?
2
33