Pixel Lights, Robot Mowers, Maslow CNC and Home Automation – HGG687
Jim Collison welcomes Seth Johnson from HomeTech.fm back to Home Gadget Geeks for a wide-ranging conversation about programmable outdoor lighting, home automation, electric lawn care, large-format CNC projects and the growing role of AI in home technology.
Seth starts with an update on HomeTech.fm and his upcoming trip to the CEDIA trade show in Denver before diving into one of his favorite seasonal projects: pixel lighting.
Audio
Video
Going All In on Pixel Lights
Seth has been experimenting with programmable holiday lighting for years and has accumulated several thousand individually addressable LEDs. Rather than having a house that looks exactly the same every night, he wants the ability to change colors, patterns and effects for Christmas, Halloween and other holidays throughout the year.
The discussion covers everything from simple permanent-style outdoor lights to much more involved pixel-light systems.
Seth talks about using HolidayCoro products, including pixel lights, controllers and props designed for elaborate lighting displays. He also explains why floodlights can be useful alongside traditional strings of pixels.
Instead of outlining every surface with LEDs, a few RGB floodlights can wash sections of the house in color. Those same lights can then serve as ordinary landscape lighting when a holiday display is not running.
That flexibility is a big part of the appeal.
For Christmas the house might become red, white and green. Halloween could bring orange and white. Mardi Gras could mean purple, green and gold. The hardware can stay essentially the same while software completely changes the appearance.
WLED, QuinLED and Xlights
For someone just getting started, Seth says products from companies such as Govee provide an accessible entry point. They are relatively easy to install and control from an app.
Once the lighting obsession takes hold, however, there is another level.
Seth talks about starting with QuinLED controllers built around ESP32 hardware and WLED. Those controllers can run addressable LED strings while giving the user much more control over how the lights behave.
For larger and more complex shows, Seth uses Xlights.
Xlights can manage thousands of pixels and create sophisticated sequences and effects. Seth explains that some effects use shaders — the same general concept used for graphics effects in games — and this is an area where he has found large language models particularly useful for generating code.
The Math Behind Pixel Lighting
Installing thousands of LEDs is not simply a software project.
Power supplies, voltage drop, wire lengths, current requirements and power injection all have to be considered. Ohm’s law eventually enters the conversation.
That leads Jim and Seth into an important discussion about where AI should and should not be trusted.
Seth makes the case that an LLM by itself should not be blindly trusted to calculate critical electrical requirements. The better approach is to give the AI access to proper tools — calculators, programs, verification steps and other systems capable of checking the result.
That distinction becomes one of the recurring AI themes of the episode.
AI can be extremely useful when it has a defined task and a reliable toolchain behind it. That does not mean every answer it generates should automatically be treated as correct.
Home Assistant and Lighting Automation
Jim brings Home Assistant into the conversation.
His idea is to let specialized lighting software handle the demanding pixel sequences while Home Assistant acts as the automation layer that decides when those sequences should run.
Seth agrees that this is the more realistic architecture.
Thousands of RGB pixels can represent thousands of continuously changing channels of lighting data. Home Assistant does not necessarily need to generate all of that traffic itself.
Instead, Home Assistant can trigger the appropriate sequence.
That allows the home automation system to provide the intelligence around the lighting without becoming the lighting controller.
Jim relates this to his own Home Assistant automations, including his Storm Guard setup that watches National Weather Service data and responds to weather watches and warnings by changing the state of equipment around the house.
Holiday Lights That Become Security Lights
The lighting discussion also produces a practical idea for people who have no interest in creating a giant Christmas display.
RGB floodlights can double as everyday landscape and security lighting.
Rather than pointing extremely bright lights away from the house, Seth and Jim discuss illuminating the house itself and allowing that light to reflect back into the surrounding yard.
During most of the year those lights can simply provide white landscape lighting. When a holiday arrives, the same hardware can switch colors.
One installation can therefore serve several purposes.
Goodbye Gas Mower
From lighting, the conversation moves into another part of Seth’s increasingly electric outdoor setup: lawn care.
Seth has moved away from his gas mower.
After years of dealing with gasoline, oil, smells and carburetor maintenance, he decided he was done. He found an EGO electric mower on Facebook Marketplace and added it to an EGO ecosystem that already included a weed eater and blower.
The electric mower is not doing all the work, however.
Seth also has Navimow robotic lawn mowers.
Robot Mowers Meet Florida Grass
Robot mower advertisements make lawn care look easy.
Reality is more complicated.
Seth deals with Bahia grass in Florida, which he describes as grass that will fight back. It grows quickly during the wet summer months and prefers to remain relatively tall.
His robot may need to mow every day just to stay ahead of it.
Then there are the boundaries.
Seth’s Navimow sometimes becomes increasingly cautious around sidewalks and other edges. As the robot pulls farther away, strips of taller grass develop around those boundaries.
Jim gives this problem a name:
Edge anxiety.
That is where the electric push mower comes back into the picture.
Instead of mowing the entire lawn, Seth can use it to clean up the edges and handle areas where the robot struggles.
Hills, Streets and Robot Mower Reality
Terrain creates another challenge.
Part of Seth’s property slopes toward a creek area, and the robot cannot safely mow every portion of the hill. There are also places where getting too close to the street could leave the mower stranded outside its programmed area.
Seth has found his robot sitting near or in the street after getting itself into trouble.
The result is a much more realistic picture of robot mowing than the traditional advertisement showing someone drinking coffee while a robot effortlessly maintains a perfect lawn.
The robot saves significant work, but the property still needs to be compatible with it, and some human cleanup remains.
Moles, Grubs and Florida Lawn Battles
Seth also spent part of the year fighting moles.
Instead of successfully trapping or poisoning the moles themselves, he eventually addressed the food source.
The moles were feeding on grubs. Seth credits an LLM with helping him figure out how to treat the underlying grass problem once he traced the dead patches back to the grub infestation.
Once the grub problem was treated, the moles moved elsewhere.
That sends Jim and Seth down another lawn-care rabbit hole involving grass varieties, neighboring lawn crews, steep terrain, lawn preparation and the realities of keeping a yard healthy in a difficult climate.
Tree Stump Troubles
Another lawn project came from the removal of a large camphor tree.
As the remaining roots decomposed underground, the ground continued to settle and created a low area in the lawn.
Seth and his friend Henry worked on filling and repairing the area, including determining when to add soil and grass seed.
Jim points out another complication of old tree roots: the decaying organic material can make establishing new grass more difficult, leading to mushrooms, sinking soil and repeated repairs.
It is one more example of why automating lawn care does not eliminate lawn projects.
Building a Large-Format Maslow CNC
Then Seth introduces one of the biggest gadgets of the episode.
A Maslow CNC.
Traditional CNC machines capable of working with a full 4-by-8-foot sheet of plywood can cost thousands of dollars.
Maslow takes a different approach.
Seth bought a kit that works with a standard DeWalt router. Stepper motors, control electronics and four anchored lines allow the machine to move across a large work surface in X, Y and Z dimensions.
Assembly involved what Seth describes as an enormous number of tiny screws, but the result is an open-source large-format CNC system that costs far less than many conventional machines.
He plans to use it primarily with materials such as plywood and plastics.
The Episode Comes Full Circle
One of the best moments comes when Jim asks what Seth plans to make with the CNC.
Christmas and Halloween decorations are possibilities.
Then Jim realizes the CNC could potentially cut the shapes and drill the pixel holes for the same kinds of lighting props they discussed at the beginning of the show.
Suddenly the whole episode connects.
Pixel lighting leads to custom holiday props.
Custom holiday props lead to a large-format CNC.
And the CNC itself can potentially use AI-assisted design tools as part of the workflow.
Seth jokes that his friend Henry can become the CNC operator while Seth stands nearby as the fireman — particularly appropriate since the Maslow documentation recommends having fire suppression available.
AI as a Toolchain
That brings the conversation back to AI.
Seth describes AI systems that can take a description of an object and help generate models for 3D printing and similar fabrication workflows.
The important part, in his view, is not simply asking an LLM to magically solve everything.
The useful model is giving the AI a clear path and the proper tools.
That same approach is showing up in his work on HomeTech.fm.
Seth has been building custom backend tools for the podcast, including a show board tailored specifically to the way the HomeTech team produces episodes and tracks ongoing projects.
Instead of adapting their workflow around generic software, they can increasingly build software around their workflow.
Is There an AI Bubble?
Jim and Seth also discuss whether today’s AI market is a bubble.
Seth believes there are bubble-like elements, but he does not necessarily expect one dramatic collapse. His view is that the market could instead continue to evolve and consolidate over time.
Some companies may disappear.
The underlying technology will not.
One difference between this technology cycle and some previous ones is the availability of downloadable models. A capable local model can continue running on hardware regardless of what eventually happens to the company or cloud service that originally produced it.
That leads into a discussion of increasingly expensive GPUs, hard drives and other computer hardware.
Both Jim and Seth have accumulated systems that suddenly look more valuable as hardware prices rise.
Local AI and Home Servers
Jim talks about his RTX 3060 system and the ability to run smaller AI models locally.
Seth describes a Dell server with two processors, 192GB of RAM and multiple Tesla GPUs.
Those systems may not always compete with the newest cloud infrastructure, but they can provide useful local capabilities without depending entirely on a hosted AI provider.
Storage enters the conversation as well.
Seth notes that refurbished 12TB hard drives that previously sold for around $100 have become much more expensive, while Jim discusses maintaining a large amount of storage and an Unraid server.
Machine Learning for Podcast Audio Cleanup
The final technical turn of the episode moves from home automation to podcast production.
Seth previously relied heavily on Auphonic for audio processing and continues to consider it an excellent tool.
But he has also trained his own machine-learning-based processing workflow around the voices and audio characteristics of the HomeTech.fm hosts.
The system can remove echoes and background noise in ways that would be difficult to accomplish with traditional audio processing alone.
At trade shows the cleanup can be so aggressive that Seth sometimes has to mix some ambient trade-show sound back into the recording so listeners understand that the hosts are actually on a show floor.
Synthetic Voices Still Have a Problem
Jim then describes experimenting with ElevenLabs after a recent podcast recording failed.
He considered generating part of a show with a synthetic version of his voice.
The result did not convince him.
The problem becomes especially noticeable when the listener already knows the real speaker.
A synthetic voice may sound impressive when it belongs to an imaginary narrator. Recreating the exact cadence, inflection and familiar patterns of someone listeners have heard for years is much harder.
Jim points to the opening line he has delivered hundreds of times introducing the show. Even with a large collection of recordings of his voice, reproducing that familiar cadence convincingly proved difficult.
Sometimes the simpler technology still wins. Jim realizes afterward that he could have simply reused a clean recording of the real introduction — a fix Seth agreed would have worked well.
Episode Chapters
[00:25] Holiday Lights and Home Automation
[01:35] Trade Show Update
[06:58] Getting Started with Pixel Lights
[17:24] Halloween and Seasonal Colors
[22:43] AI, Math, and Light Sequences
[31:47] Home Assistant for Lighting
[38:03] Security Lighting Ideas
[39:53] Electric Mowers
[42:30] Fighting Florida Grass
[43:44] Moles and Lawn Battles
[50:26] Edge Anxiety
[55:32] Tree Stump Troubles
[57:53] Large-Format CNC
[1:04:06] AI and Home Projects
[1:08:21] The AI Bubble
[1:10:47] Storing Hardware
[1:12:26] AI Voice Cleanup
[1:14:19] Synthetic Voice Tests
[1:17:02] Show Wrap-Up
Key Takeaways
- Pixel lighting scales quickly. Easy app-controlled lights are a good starting point, but larger displays require controllers, power planning, voltage calculations and dedicated sequencing software.
- Home Assistant works well as the automation layer. Let specialized lighting software generate complex pixel sequences while Home Assistant decides when they should run.
- Robot mowers reduce lawn work but do not eliminate it. Edges, slopes, fast-growing grass and difficult terrain can still require a conventional electric mower.
- Large-format CNC is becoming accessible to DIYers. Seth’s Maslow CNC can work with full plywood sheets without requiring a traditional multi-thousand-dollar CNC table.
- AI is most reliable when paired with tools. Programming assistance can be extremely useful, but calculations and critical technical work should be verified with appropriate tools.
- Synthetic audio is better at cleanup than impersonation. Machine learning can remove noise and echo extremely well, while reproducing the recognizable cadence of a familiar speaker remains difficult.
Products / Platforms / Technologies Mentioned
HolidayCoro, Pixel/addressable LED lighting, RGB floodlights, Govee outdoor lighting, QuinLED, ESP32, WLED, Xlights, DMX lighting, Home Assistant, Navimow robotic mowers, EGO electric mower, EGO weed eater, EGO blower, Maslow CNC, DeWalt router, Unraid, RTX 3060, Dell server / Tesla GPUs, Auphonic, ElevenLabs, local AI models, CEDIA
Links and Resources
- HomeTech.fm: https://hometech.fm/
- HolidayCoro: https://start.holidaycoro.com/
- Maslow 4.1 CNC Kit: https://shop.maslowcnc.com/products/maslow-4-1-kit
Listen and Subscribe
Find Home Gadget Geeks episodes, show notes and subscription options at:
