Notice for HamClock Users

I found a message from N0LSR in my junk mail folder yesterday, and it was important. It’s about the failing Clear Sky Institute server. If you are an Inovato or N0LSR HamClock customer, “all users must switch to the new OHB (Open HamClock Backend) to ensure continued updates and reliable operation.” Find more information on the N0LSR home page.

There’s a quick way to tell if you are on the correct server: “If you are currently on any version below 4.26 and do not get the upgrade offered to 4.27 then you are on the wrong backend server.”

Once upgraded, the version will say 4.27 and you’ll see an envelope just below the version number. Click on the envelope icon for a status message.

1. QRV: Are You Ready?

It may seem as if there is little radio content in this issue, but that’s because the bulk of my radio work landed on EtherHam.com this week. Specifically, I took several days to develop a detailed hands-on look at my new Talk2Me push-to-talk over cellular radio. The pointer for that is below in the New on EtherHam section. If you’re interested in PoC radios, you’ll find this worth your time.

After looking at how these radios and services are marketed, how the ordering and fulfillment process actually happens from the end-user point of view, and how the radio and service function, I’ve concluded: this is not a mature market. Several companies are vying for your dollars. The actual hardware and services offered are in a state of continuous change, so I think it is very difficult for a vendor to keep up with onboarding information, documentation, and even fulfillment steps. This shouldn’t dissuade you from entering this space if you are interested in PoC radio, but do be prepared for a rockier process, from start to finish, than you may be used to experiencing with mainstream amateur radio products. Why would you do this? Because PoC radios can offer an unprecedented level of connectivity and convenience.

2. New on EtherHam

• Push-to-Talk Over Cellular: Hands-On with a Modern PoC Radio — Part 3 in a series. My radio arrived with no instructions. I learned how to use it by pushing every button and chasing every menu. Thirty minutes to enter a WiFi password. An SOS mode hiding behind the flashlight button. A Bluetooth pairing that is not obvious how to undo. And when I finally found the official manual online, it described a radio with no WiFi and no Bluetooth — not the radio I bought. The device itself works, and it puts Kansas City Wide in my pocket, at home on my network and while I’m on the road. Should you buy it? Overall, I’m happy with the device and the service.

• Building a LoRa APRS iGate/Digipeater — Part 1: Why, What, and How — I already run conventional VHF APRS from the lake house, so when I heard about LoRa APRS, I wanted to see what it would take to get a station on the air. (LoRa APRS is a newer parallel network built on chirp spread-spectrum instead of classic AFSK packet.) When I checked the LoRa APRS live map, I found exactly one LoRa APRS station in the whole area, and it hadn’t heard anything over RF in over two months. Here’s how I turned a LilyGo T-Deck Plus into KJ7T-2 — a live 433.775 MHz iGate and digipeater — including recovering from installing the wrong firmware and the hidden BOOT button.

• Weekly Report: July 23, 2026 — Band Conditions, Digital Radio News, Groups.io Digest.

3. New on the Bench

MeshCore Repeater — I bought a MeshCore repeater from an Etsy seller. It is solar-powered with a single 18650 battery inside to keep it running in the dark times. I am halfway between two other MeshCore devices, each about 15 miles away. You can explore the worldwide MeshCore map at https://map.meshcore.io. In the screenshot below, I’ve placed a Random Wire sticker on my location. You can see I have stations to the southwest and to the northeast.

I’ve magnetically attached the repeater to the steel chimney cap.

Even though my device is up above the roofline, I’m on a lake, and that means I occupy a topographic low. I do have a 915MHz mag-mount antenna coming that I hope will extend the reach of this device, but I’m not optimistic. Still, it’s an interesting device and system to explore.

T-Deck Plus — This LilyGo device will probably arrive the day after this Random Wire issue publishes. This is a 915MHz device that I plan to flash with MeshCore firmware so I can talk to my repeater and perhaps get that repeater onto the MeshCore map. If MeshCore doesn’t work out, I can flash it with Meshtastic. I might also try a T-Echo Tracker on 915MHz, too.

T-Echo Tracker — This is a second T-Echo, but this one is configured for 433MHz. It should connect to my LoRa APRS igate/digipeater running as KJ7T-2. I want to use the T-Echo to see how far KJ7T-2 will reach.

APRS testing — I’ve been monitoring APRS on APRSdroid, aprs.fi, and my mobile radio, wanting to see if those were picking up my KJ7T LightAPRS Gateway device…and they are. What surprised me was that my LoRa APRS box, running as KJ7T-2, is also picked up by aprs.fi.

I also see that my internal battery voltage on the LightAPRS Gateway device has come up from the 4.5V range to nearly 4.7V. That’s because I changed the 1A wall charger to a 2A device.

4. Website: EtherHam CSS Cleanup with Claude

This may only be of interest to other webmasters. If you don’t manage a website, read this as: Claude can help you handle some mundane tasks…and feel free to skip it if the idea of cleaning up CSS code makes your eyes cross!

When I was first setting up EtherHam.com, I chose to use WordPress as the content management system. I’ve been working with WordPress for years. While it is a bit bloated for my needs, I know it inside and out. Just like using a favorite radio, there is a certain comfort in that.

I worked my way through several block themes, each requiring some Cascading Style Sheet (CSS) customization. Many fragments got orphaned along the way — leftover CSS doing nothing but cluttering the file.

I’d read that Claude could help manage a WordPress site, but hadn’t trusted it enough to try. This week, I decided to give it a try. First, I backed up the entire site, including making a copy of the CSS file I was about to change. Then I asked Claude for help.

I installed the Claude in Chrome extension when prompted, and after that, Claude had direct access to what I had open in the browser. This process worked well. We cleaned up the CSS settings, and along the way, I identified a few plugins that I had tried but never used, so those got removed.

Since I was on a roll, I asked Claude for help making formatting consistent between the home page, sub pages, and posts. In just a few minutes, we aligned formatting in these different views.

I posted the exact CSS file Claude helped me clean up on EtherHam, if you’re interested. It is nicely organized and well commented.

If you’re curious, you can download a copy as a text file.

Bottom line: EtherHam.com looks fine and the styling is cleaned up and consistent throughout the site.

5. Vehicle Super Starter: Why I Didn’t Buy It

This isn’t amateur radio, but it could intersect with licensed operators if they run a mobile radio in their rig.

What is this about? I thought about buying a super capacitor for my truck. The sales messaging I’ve been seeing seemed compelling and I find the concept quite clever, so I gave it more than a cursory look.

Ancel sells a device called the Super Starter — a permanently installed super capacitor that wires into your battery terminals and sits under the hood, taking over most of the cranking load every time you start the engine. The pitch is straightforward: your battery stops absorbing the punishing current spike of a cold start, wear and tear drops, and the battery lasts longer. No cables, no remembering to charge a separate pack, no getting out of the truck. Connect it once, forget it exists, and let it quietly do its job for years.

I’ve had dying and weak vehicle batteries, so I really like the concept. I like anything that takes a known failure mode — batteries dying at the worst possible moment — and tries to engineer it away rather than just selling you a booster with a flashlight and jumper cables.

But a few things stopped me before I clicked buy. First, the fit-to-cost problem: the base model is rated for engines up to 3.0 liters, and my Ridgeline runs a 3.5L V6. The unit that actually matches my truck runs closer to $220 than the $150 headline price on the smaller model — a detail that’s easy to miss if you’re moving fast.

Second, the marketing language has a familiar look to it. If you’ve watched the push-to-talk-over-cellular space the way I have, you’ve seen this pattern before: a genuinely interesting piece of engineering wrapped in urgency — limited-time pricing, breathless “why didn’t this exist sooner” copy, glowing coverage that reads like it was written by someone with an affiliate link. None of that makes the underlying tech bad. But it’s usually a sign the vendor wants you to make the purchase decision before you’ve had time to ask the boring questions, like whether the battery-life claims hold up after a real winter under the hood. This product launched this year. There’s no multi-year data available yet, independent or otherwise.

However, there is a review published by Car and Driver that touches on the utility of the device for cold winter vehicle starts. It sounds solid, and it is a real-world, hands-on kind of review. If you’re interested in this tech, this is worth a read: This Clever Supercapacitor Accessory Is a Smart Fix for Cold Car Starts.

Third, and this is the part that’s specific to the Random Wire audience: if you’ve got a mobile rig hardwired into your vehicle’s electrical system, this device is doing something to your crank and start-up voltage profile that nobody’s characterized yet. It’s not a fuse-blowing concern in the classic sense — the Super Starter probably doesn’t add current through your radio’s own power leads, so an overcurrent trip on that circuit isn’t the likely failure mode. The more concerning question is voltage transients on the 12V bus itself. During the crank event, the capacitor is absorbing most of the starter’s draw and then handing off to the battery and alternator; right after start, it’s pulling current back from the alternator to refill itself. Either transition could produce a brief dip or overshoot on the bus that a normally-fed rig wouldn’t otherwise see — and mobile radios are already notoriously sensitive to exactly that kind of thing, which is why so many of us add extra transient suppression to a hardwired install in the first place. I haven’t found research measuring the 12V bus with and without one of these installed, from Ancel or anyone independent. Until someone runs that test, I wouldn’t want it anywhere near several hundred dollars of mobile radio gear.

Fourth, and the part that actually killed it for me: the thing only helps the vehicle it’s bolted to. I have kids with vehicles in various states of mechanical health, and I’m the guy who often gets the call when one of them won’t start. A device clamped to my own battery terminals is no use to them. This means I’d still need to carry a portable lithium booster in the truck regardless of whether I installed the Super Starter. And if I’m already carrying a booster that covers every situation the Super Starter covers plus the ones it doesn’t, the case for spending another $220 on a single-purpose, single-vehicle install gets thin fast.

So here’s where I landed: it’s a legitimately clever idea, and I like the concept. If you’ve got a vehicle that sits for long stretches, hauls a heavy accessory load, or otherwise chews through batteries, I understand the appeal. But for a daily-driven rig with a battery in decent shape, and a household where “helpful” means being able to jump anyone’s vehicle, not just your own, a $60 portable booster under the seat does more for less money — and doesn’t ask you to trust a new product category on faith. I’ll keep an eye on this space. If the independent, multi-season reviews start showing up and the numbers hold, I may revisit it. For now, my jumper cables and a charged lithium pack are in the truck.

6. Short Stack from the Interwebs

Pulled from the aggregated feeds on EtherHam (etherham.com/feeds/), items from July 20–23.

Antennas

• The Crown Antenna: A Wire Loop That Hugs the Roof and Works the World — Hamradio.my — A roof-hugging wire loop design for HF. Worth a look if you’re antenna-restricted.

• Part 2 of… Antenna solve one problem to only create another — Amateur Radio Weekly — the classic antenna-tinkering trap: fix one issue, discover a new one. Relatable.

Propagation & Science

• HamSCI Ray Tracing Working Group — HAMSCI.org — HamSCI’s group working on ray-tracing tools for ionospheric propagation modeling.

• The ARRL Solar Update — ARRL News — Latest solar/propagation conditions rundown.

Satellites

• UmKA-1 (RS40S) Satellite to Transmit SSTV, Audio Clips — Amateur Radio Daily — New satellite planning SSTV and audio transmissions — worth tracking if you’re into satellite work.

APRS & Digital

• APRSdroid 9M2PJU Mod: The Enhanced APRS Companion for Android, Built by Hams, for Hams — Hamradio.my — A ham-built mod to APRSdroid. Given the LoRa APRS iGate work on EtherHam right now, this is a natural companion piece.

DIY / Gadgets

• Flash a $2 Pi-hole From Your Browser: The 9M2PJU ESP32-C3 AdBlocker — Hamradio.my — Browser-flashable ad-blocker on a $2 ESP32-C3. Not radio, but exactly the kind of cheap-hardware hack that fits the workbench.

History & Culture

• How Amateur Radio Built the Indonesian Internet: PaguyubanNet, Packet Radio and the Pioneers of 1986–1994 — Hamradio.my — Deep history on how ham packet radio bootstrapped Indonesia’s early internet. Genuinely excellent long-read, easily the standout item this week — from both the curated and mega feeds.

Reference

• DXing Techniques for Amateur Radio: A Complete Guide to Working the World — Amateur Radio Today — Solid DXing primer, good bookmark-and-share material.

Community / Silent Keys

• DXer and DXpeditioner Kan Mizoguchi, JA1BK, Silent Key — ARRL News —Passing of a notable DXpeditioner.

Also Noted

• San Diego Section of IEEE Invited Talk Intersects ORI’s Antenna Work — Open Research Institute

• VHF for Parks On The Air? — Amateur Radio Weekly

• The W9IMS Brickyard 400 Special Event: Third Time’s a Certificate — The SWLing Post

• Records with Soul and a WAVE — The SWLing Post

7. Ported Scripts to My Proxmox Server

This week, I moved a few scripts to my Proxmox box. The backstory? The Raspberry Pi where those scripts lived…died. And it wasn’t just any Pi — it was the one that handled my home lab tracker and processed the Digital Radio News Gatherer script every week.

I had been planning to move those scripts to my Proxmox server — someday — so the failure of the Pi changed that to right now. Fortunately, I was keeping regular backups of the scripts and data on a USB thumbdrive, in ext4 format. Proxmox can read ext4, so my backups provided the smoothest way to transfer scripts, settings, and data to Proxmox containers.

The Pi’s death wasn’t quiet. First it was a service crash-looping on the lab tracker. It turned out a routine apt autoremove had quietly pulled Flask out from under a process that had been running fine for weeks, and nobody noticed until the next reboot forced it to reload a dependency that was no longer there. In other words, it was already broken, I just didn’t know it.

Then the real trouble started: the case ran hot, SSH sessions began dropping mid-command, and one by one, services stopped responding. Cockpit, gone. Docker, gone. The web apps stayed technically alive but started throwing I/O errors the moment they tried to touch the disk. That’s the signature of a filesystem giving up, not just a CPU getting warm. A fan and a reboot bought some time, but a few hours later it came back worse — even basic commands like tail and sudo started failing. The NVMe drive, a budget no-name unit, may well be on its way out…or is possibly underpowered by the HAT. Pi 5 NVMe bases are known to be stingy on current under sustained load, which can look exactly like a failing drive without one actually being bad.

By the time I went to grab the backup drive, every way I had of reaching the Pi remotely had stopped working — the usual command-line connection, the web-based control panel, even the official remote-access tool. I pulled the thumbdrive and plugged it into my Proxmox box. After that, recovery was almost anticlimactic: I could read straight from the drive without disturbing anything, and there sat every night’s backup, the program and its database both untouched. From there, I built a new home on the Proxmox server: a small, self-contained slice of the machine, running its own private copy of the software it depends on instead of sharing whatever happens to be installed system-wide. That last part matters more than it sounds like it should — it’s the difference between a routine cleanup silently breaking your service, and your service carrying its own dependencies with it, immune to whatever else happens on the machine.

The Digital Radio News Gatherer had its own separate gremlin — a Groq API key that looked fine in its config file but wasn’t actually making it into the running process, because the cron job launching it never sourced the environment file, and even after fixing that, the file itself was missing an export in front of the key. Two small, easy-to-miss mistakes stacked on top of each other, neither one loud enough to point straight at the cause. That is sorted now, and worth remembering for anyone else running scripts out of cron with secrets tucked into a separate file.

These are lessons for anyone running similar infrastructure off a Pi — and I know plenty of you are: back up more than just your data. A solid backup process is what saved me here, not a database export. (All of my Proxmox containers are backed up weekly to an external target.) And if you can move a service that matters onto something with real redundancy, consider doing that before the hardware makes the decision for you.

8. Open Notebook on Proxmox Server

I’m starting to play with an open source project called Open Notebook. So far, it looks interesting and seems to run fine on my Proxmox server. I’m going to use it for capturing meeting agendas and notes, and then see if that kind of content proves useful.

I installed it several weeks ago and already an update was waiting. Updating it was very easy — much easier than I expected. It runs in a Docker container, so just three commands took care of the version upgrade:

docker compose down
docker compose pull
docker compose up -d

Docker Compose makes it pretty painless — three commands and your data stays untouched since it lives in the mounted volume, not in the container. I wish all updates were so easy!

9. QRT: End Transmission

Plumbing, part 2. We’re still dealing with the plumbing leak I reported last week. We now have a new faucet that works well. I had to add a plug to the unused hole in the sink top, but that was not much money and only a quick trip to Ace Hardware. The continuing issue is the water that seeped under the kitchen floor tiles. That lifted several tiles along the edges. I’ve been running a fan and a dehumidifier constantly since then to try and dry the moisture out from under the floor. A new moisture meter is arriving this week: pinless, so I can take relative measurements, comparing dry, unaffected spots with the parts that did receive some seepage. I need to make sure that wood won’t grow mold when I fix some of those tiles.

Frustrated with bad UI. I’m trying to configure an events plugin for my work website, and the user interface is full of labels that don’t mean what they say. Example: “configure calendar” is actually a shortcode generator. Or two labels that sound identical, one above the other. A “configure” link hiding under the “other data” label. An unlabeled, bright orange button in the usual “submit” position with an arrow on it…that doesn’t appear to actually do anything. If you couldn’t tell, I really, really don’t like poorly designed UIs!

And that’s enough to close Random Wire 192. I hope you had a productive, interesting radio week.

73, and remember to touch a radio every day!

1. QRV: Are You Ready?

I started a series on DMR radio this week, because my coverage of DMR has been a bit light. Why? Because I don’t really understand it. The best way to learn DMR is to use it, and I find that organizing what I’m learning and writing about it really helps to cement that knowledge in my mind.

I fixed a bug in my AllStar node running on an Arduino UNO Q board. It is reliable again. Speaking of reliability, I also implemented a backup routine for one of my primary AllStar nodes that is driven by a Dell Wyse 3040 thin client PC.

There are a few write-ups about artificial intelligence in Random Wire 191. I had an interesting experience with two different models in Claude. I also gave Hermes Agent a try and found it is too powerful and complex for what I need, but that doesn’t mean it isn’t right for you.

Following up on my deep dive into the push-to-talk over cellular space, I ordered a PoC radio. This is a leap of faith. I want to experience what a typical buyer experiences, and I’m hopeful the new radio will be fun to use. This week I touch on some purchasing friction points. Next week I should have the radio in hand.

I learned about CascadiaMesh this week…and it is interesting enough that I ordered a solar-powered mesh device to go on the roof. Fingers crossed it works well.

I signed up for QSO One and tested it for a few days. It has great promise, but right now, the UI is still a little rough around the edges, at least in my experience.

And I saw a strange glitch in my APRSdroid display. It was odd enough that I did some research on it and documented it in a bug report filed on GitHub.

I also tried a speaker-mic on the Uniwa F400 smartphone with DVSwitch and DroidStar installed. Did it work? Yes as a remote speaker, no as a microphone. I’ll be exploring button mapping as a way to try to get it working with that platform.

That’s Random Wire 191 in a nutshell. I hope you enjoy reading about these amateur radio and technology topics this week. I’m always open to learning and trying new things, so please do send your ideas and suggestions to me!

2. Thank You

I want to thank all the innovators and all the folks who form the solid backbone of our hobby. Last week, I wrote a brief piece about crossing the chasm, in the context of adopting amateur radio technology.

This week, I’m leaning in on this concept. I usually occupy the “early market” part of the curve — innovators and early adopters. Trying new things, or things new to me, is what I enjoy most. The makers, designers, builders, and coders keep the magic alive in our hobby. I doubt we’ve ever seen as much innovation and creativity in amateur radio as we’re experiencing now.

But people like me are actually a small percentage of the hobby, compared to the much larger group: the “mainstream market” folks. The early majority, late majority, and laggards (not my choice of term!) form the spine that holds up amateur radio, year in and year out. They buy the gear that keeps major players in the game, and that benefits all of us.

The “early market” folks are one side of the amateur radio coin; the “mainstream market” folks are the other. We can’t have one without the other — both are needed to keep the hobby healthy and vibrant.

Thank you to the innovators and early adopters. Thank you to the larger group that keeps the hobby together. No matter which side of the coin you prefer, you’re important to all of us.

3. New on EtherHam

Some fun radio-related topics this week.

• Introduction to DMR: The Four Wheels on the Car — This is Part 1 in a short series on DMR, written for those of us who find DMR, well, daunting. DMR asks new users to understand codeplugs, timeslots, talkgroups, and networks before they’ve even keyed the mic — and most guides bury that in jargon before you get near your first contact. This piece cuts through it with one idea: DMR is like a car, and those four concepts are its four wheels. Get all four turning and you’re on the air; skip one, and you’re not going anywhere. Part 1 of Tommy’s Guide to DMR builds that mental model — no gear list yet, no codeplug, just the four ideas you need before any of the rest makes sense.

• AllStar on Arduino UNO Q: Part 2 (The Bug That Killed My Node…Until I Fixed It) — If your UNO Q AllStar node runs for hours and then drops dead — heartbeat LED red, lsusb empty, nothing but a full power cycle bringing it back — the culprit is a documented bug that silently flips the board’s USB-C port out of host mode. The fix is a one-line command, made permanent with a small systemd service, and it’s all laid out below. The QNode is now rock solid, and yours can be too. Part 1 is at How to Run AllStar on the Arduino UNO Q.

• Backing Up Your AllStar Node: Why I Do It Even Though “eMMC Never Fails” — Every AllStarLink node builder has an opinion on whether the storage in their node will ever fail — and one popular take is that the eMMC in a Dell Wyse 3040 thin client basically never does. That might even be true. But I learned decades ago, the hard way, what it feels like to lose weeks of irreplaceable work to a single bad disk, and I’ve never wanted to relearn that lesson on a node that also connects to my ham phone. Here’s why I back up my node anyway, what eMMC actually gets right (and wrong) compared to microSD, the one AllScan file I almost forgot to include, and a free, MIT-licensed script you can use so a dead drive is a shrug instead of a crisis.

• Hardening SSH Access to Your AllStarLink Node: Moving Beyond Port 22 — Your AllStarLink node isn’t just a simple appliance — you’re running an unattended Debian server with a direct line to your radio and your home network, and automated bots are scanning for it right now. A compromised node isn’t just an inconvenience: it can mean unauthorized transmissions attributed to your callsign, or a foothold into every other device on your LAN. This walkthrough shows how to close a common gap in an afternoon — moving off the default SSH port, adding real rate-limiting, and locking down Fail2Ban correctly on Debian, including the exact reason it silently fails for a lot of people.

• Push-to-Talk Over Cellular: The Actual Buying Experience — I ordered a PoC radio from Talk2Me USA on July 11, expecting to click “buy” and wait for a box. Instead, the next four days were an inbox: a $15 billing mix-up that was later reversed, a new Terms of Service that arrived the day after I paid, and login credentials for three separate radio networks — before the hardware itself had even shipped. The ad copy sells “buy now.” The actual experience is “buy now, then wait” — and it turns out most of what a buyer needs to know happens in that gap, not at checkout.

4. Claude AI: An Hour with Sonnet, Ten Minutes with Fable

I reported last week that the default engine on Claude had changed from Opus to Sonnet. I also reported that Fable had been released, then blocked, and then made available again. I got a chance to use the high-powered Fable engine this week, and therein lies a story.

This relates to the UNO Q item directly above in this newsletter. As I tried to figure out what was going on with my Arduino UNO Q-powered AllStar node, I eventually turned to Claude for some help. Using the default Sonnet engine, we went round and round for an hour. It got to the point where I would tell Claude: no, we’ve tried that twice already, I do not want to repeat that process…and Claude would push it forward anyway.

Eventually, I asked for a written summary of that conversation, closed the thread, and opened a new one with Fable as the engine. Ten minutes. That’s what it took to do the proper research and lead me to a solution that worked.

To be fair, this comparison isn’t perfectly clean. It’s tempting to blame Sonnet for an hour of churning — but that churning is what produced the diagnostic evidence, including the exact kernel log signature of the failure, that let Fable’s first web search land directly on a known bug report. A fresh session armed with documented dead ends gave Fable a real head start.

There’s a workflow lesson here that applies no matter which AI you use: when a session starts going in circles, ask for a written handoff summary, close the thread, and start fresh. That technique rescued this project as much as the model swap did.

Honestly, Opus was a capable engine. I’m not particularly pleased with Sonnet, at least for sleuthing code and hardware problems. Fable burns through credits faster, but I was very pleased with how thoroughly and surgically it developed a solution. As noted, though, the work I did with Sonnet gave Fable a leg up as we reached the finish line.

5. Hermes Agent: Complex and Powerful

Some weeks ago, I set up an instance of Hermes Agent on my home server. I started by powering the reasoning side with Grok, but it burned through my credits faster than expected, so I downscaled to Gemini (free), which worked well.

Hermes Agent is an open-source AI agent framework, and “agent” is the operative word — it’s not a chat window, it’s a system meant to run unattended and act on your behalf. It can execute multi-step tasks autonomously across your terminal, your IDE, and more than ten messaging platforms (Telegram, Discord, Slack, WhatsApp, Signal, email). It remembers things across sessions — your preferences, your environment, lessons learned — and it can save reusable procedures as “skills” so it gets better at your specific workflows over time. You can point it at almost any LLM provider (OpenRouter, Anthropic, OpenAI, DeepSeek, local models) and swap between them mid-workflow, run multiple isolated instances with their own memory and config, and extend it with plugins, custom tools, webhooks, and cron jobs. On paper, it can handle coding and debugging, research, system administration, data analysis, content creation, and even home automation.

That list alone tells you something: this isn’t a “download an app and go” tool. Getting it running means configuration files, API quotas, provider swaps, and troubleshooting that assumes you’re comfortable poking around under the hood.

I spent a couple of days putting it through its paces — switching models, testing its WordPress and GitHub integrations, watching how it handles tool-calling. It’s genuinely capable, and I came away with real respect for what it can do. I also hit real friction: a crashed gateway, a free-tier quota that ran out faster than expected, and copy-paste quirks that took some digging to sort out.

Here’s my takeaway. If you have a specific need for an autonomous, always-on agent — something that runs unattended, chains multiple tools together, or watches your systems continuously — Hermes Agent may be worth the investment of time it takes to learn. But that’s not the problem I have. I mostly want a smart assistant to review my writing, help me code, or think through a problem. For these tasks, a standard AI chat interface gets me there with a lot less overhead. Hermes is a great solution for a problem I just don’t have.

6. PoC Radio Followup: I Ordered One

I’ve had some interesting exchanges about my push-to-talk over cellular deep dive (see: Buying Push-to-Talk Over Cellular Without Getting Taken). Even though I’ve previously purchased three PoC units, those were generally low-end devices. I’ve ordered a modern unit at a higher price point that meets these criteria:

• Cellular connectivity

• Wi-Fi connectivity

• Can use the HamLink PoC app by Talk2Me

Why? Two reasons. First, I want to put myself in the shoes of a typical buyer who is trying to find the device that meets his/her needs. Second, I want a PoC device that will also let me connect to America’s Kansas City Wide Network, and HamLink PoC offers that capability. I ordered through Talk2Me specifically because it appears to be one of the leading vendors in this market space, which gave me more confidence going in than I’d have with an unknown seller.

Worth saying up front: I trusted Talk2Me enough to send them $200 for a radio I hadn’t yet held, on the strength of that reputation. That’s the lens for everything below. These are observations from a buyer who chose them for good reason, not grievances against a vendor I regret choosing — and precisely because I went with a name I trusted, the friction points I ran into are worth knowing for anyone buying a PoC radio, from any vendor.

So far, the buying experience is about what I expected. Details about each radio choice were hard to find. The HamLink PoC app was supposed to be free with a radio purchase — instead, I was charged for it, and later told that charge will be refunded. This kind of behavior (“it’s free” but then you get charged) makes it harder to trust vendors. To be clear, these observations are necessarily about Talk2Me, since that’s who I bought from, but I have no reason to think Talk2Me is worse than other vendors in this space. If anything, the fact that friction shows up even with a leading vendor is the point: these are things any PoC buyer should watch for, regardless of who they order from.

I was pleased that for the HamLink PoC app, I was asked to provide a copy of my FCC-issued amateur radio license. That gives me confidence that app usage will be limited to licensed operators.

The radio was originally tracking to arrive Saturday, July 18, at the early end of the vendor’s 7-11 day estimate. Since then I’ve received a delay notice pushing delivery to July 20. Another friction point worth flagging for anyone shopping this category: the delay notice didn’t explain a cause, and there’s still no order-status page or tracking link that gives more detail than “delayed.”

I also received a new terms of service notice before the radio or the app were even in my hands. The timing felt odd — as though placing the order triggered it, rather than it being a routine notice sent to all customers. If it’s the latter, the coincidence is remarkable. I can’t say which it is, and I’m not accusing Talk2Me of anything here. However, a buyer in the middle of an active order is not in a great position to evaluate a new ToS agreement on its merits. That’s worth naming as its own category of friction, separate from the pricing and shipping issues above.

For a purchase already made opaque by pricing and add-on confusion, a shifting delivery date without explanation and an untimely ToS update amount to small trust deductions. None of this is disqualifying on its own, but it’s the kind of accumulated friction that shapes whether a buyer sticks with a vendor for a second purchase.

I’ll have more on this once the device arrives and I get it configured — and with a radio in hand, issue 192 will shift away from the purchasing process and focus on the device itself. Fingers crossed this is not a $200 mistake. I’m actually hopeful the device and service will prove useful and enjoyable!

7. Have You Heard of CascadiaMesh?

Thank you to subscriber Joe who brought CascadiaMesh to my attention.

CascadiaMesh is a volunteer community project (not a company or nonprofit) that’s building out a regional mesh network across British Columbia, Washington, and Oregon, using solar-powered nodes for “grid-down” resilient comms. It’s grown fast — over 2,300 active nodes as of today. The network runs on MeshCore, an open-source LoRa mesh protocol, and the site provides onboarding guides, device recommendations, node-building tutorials, a live map, and links to sibling regional meshes (Puget Mesh, Salish Mesh, etc.) across the US, Canada, and Europe.

My take? CascadiaMesh appears to have positive momentum. Here’s a 30-second screen recording of the live map to give you an idea of the vibrancy of the network. It is operating not just across town, but across state and provincial lines.

Important: MeshCore has its own iOS/Android app, its own web flasher (flasher.meshcore.dev), and its own config tool. A MeshCore node can’t talk to a Meshtastic node — they use different protocols and encryption.

OK, color me interested. I ordered a small solar-powered node from an Etsy vendor. I’m hoping to put it up on top of the galvanized steel chimney cap which sits above the peak of the roof. Maybe there are other MeshCore nodes near me!

If MeshCore interests you, too, you should probably be aware there is a battle going on over trademarking the MeshCore name. Ultimately, I don’t think this will affect the codebase or the community, but it could result in the project getting renamed. I hope not, because MeshCore is perfectly descriptive.

8. QSO One, Part 2 — Testing

QSO One‘s pitch is bold: “Amateur radio, reimagined.” The marketing promises a single modern client that puts AllStarLink, EchoLink, DMR, and M17 all in one app, on Windows or Android, with no node to build, no hotspot to flash, and no radio interface to wire up — twenty-five dollars and five minutes between you and being keyed up on any of it. It’s framed as the client nobody else had bothered to build: a clean, cross-platform answer to IAXRpt’s abandonment, DVSwitch’s complexity, and the fragmented, one-app-per-mode reality most of us have been living with for years.

If this is real, it will have real value for radio amateurs. So I downloaded the QSO One app and installed it on my Windows laptop. This is what I like doing: poke at something to make sure it lives up to its billing. I’ll have a full write-up on this in the near future because the experience was a bit rocky. Essentially, the UI doesn’t really lead you through how to get set up and connected. I spent quite a bit of time poking around, trying to figure things out.

But once I got it connected with my AllStar node 588418, it worked well. The audio had a bit of choppiness to it but I suspect that is my network, not the app. To test that, I shut down QSO One and used the venerable iaxRpt app to connect to my node, and heard exactly the same choppy audio.

I was not successful at getting M17 to connect. It did find my reflector M17-PNW but then errored out. This appears tied to a connection-handoff bug in the app rather than something wrong with M17.

At this point, I would describe QSO One as a very good “built from the ground up” new app that is still rough around the edges, in active development, and shows promise. This one is definitely worth watching.

9. Chasing a Weird APRSdroid Display Glitch

A few weeks ago I noticed something odd in APRSdroid’s Hub view on my Android smartphone: my mobile’s SSID was showing up as “KJ7T-9_1” instead of “KJ7T-9.” At first I assumed I’d fat-fingered something on the radio, so I changed the SSID to -11. Same thing happened — “KJ7T-11_1.” Then I saw Jeff’s portable, W7NEE-9, and it had its own additional underscore and number: “_3.” Whatever this was, it wasn’t just me, and it clearly wasn’t an SSID problem, since the SSID itself was displaying correctly in the title. The mystery character was showing up as a second line, like a comment.

The next step was to stop guessing and go look at the actual data. Pulling the raw packets for KJ7T-9 from aprs.fi’s raw packet view showed the underscore-digit baked right into the transmitted Mic-E packet. APRSdroid was not inventing this — it was real. And critically, aprs.fi wasn’t displaying it as a comment at all. That told me the “_1” was actual data that one piece of software was interpreting correctly and another wasn’t.

It turns out this is a documented part of the Mic-E spec. Bob Bruninga WB4APR (SK) produced a reference table of Mic-E “device identifier” bytes — a two-character suffix that radios append to the end of the Mic-E info field to identify themselves, positioned right after the base-91 altitude data. Per that table, “_1” decodes to a Yaesu FTM-300D and “_3” decodes to a Yaesu FT5D. The spec is explicit that receiving software should recognize these bytes and strip them from the displayed text (or use them to show the radio model) rather than showing them as if they were part of a comment. aprs.fi does this correctly. APRSdroid, at least as of the current release on my phone, doesn’t.

Rather than just grumbling about it, I filed a bug report on APRSdroid’s GitHub page. I made sure to include the actual raw packet as evidence, a link to the spec document that defines the correct behavior, a clear statement of what I observed versus what should happen, and a specific suggested fix. I also included my app version and Android version, since that’s the first thing any maintainer will ask for anyway.

I’d encourage anyone who runs into a quirk like this in ham radio software to consider doing the same thing. Many of these tools — APRSdroid included — are maintained by one or two volunteers in their spare time, often for free, and often for years. A report that says “this looks broken” is easy to ignore. A report with raw data, a spec citation, and a proposed fix is something a maintainer can act on in five minutes instead of an hour. It also means the next ham who hits the same weirdness can search the issue tracker and find an answer instead of struggling to figure out a diagnosis from scratch. Bug reports help the maintainer, but they also help the community.

10. Speaker-Mic for Uniwa F400 Phone

Will it work? That has been on my mind since I ordered this through Amazon (this is an affiliate link):

Portbale Microphone for UNIWA F400 and Anysecu W400 4G Network Radio

(Yes, portable is misspelled. That’s the way it is listed.)

It certainly makes the smartphone look more like a radio, but will it work? It arrived, I attached it, and it works well as a remote speaker. However, the push-to-talk button works the same as the PTT on the phone itself, which is to say it doesn’t work with DVSwitch Mobile or DroidStar. I didn’t really expect it to, but I also haven’t explored button mapping apps to see if I can get the PTT to work with one (or both) of those apps.

Two resources I’ll be exploring to try to get the PTT functioning the way I want.

• Button Mapper: Remap your keys (by developer flar2) — the most established option, free, no root required for most actions, works via Android’s accessibility framework. Nearly 11,000 reviews on Google Play. Handles volume keys, camera buttons, headset buttons, and “custom buttons” (which is the category a PTT accessory would likely fall into). Find it in the Google Play Store.

• Key Mapper (open source, by sds100) — more technical/granular option, popular in the XDA/hobbyist crowd. Supports multi-key combos and sequences, and unlike Button Mapper it’s fully open source if your readers want to see exactly what it’s doing under the hood before granting accessibility permissions — which is a reasonable thing for a security-conscious ham audience to want. Find it on GitHub.

Note that both of these apps rely on Android’s accessibility service to intercept the keypress. Also, remapping only helps if the target app already accepts input from some other bound key (like volume-down) that these tools can redirect the PTT press to. It’s a promising rabbit hole, not a guaranteed fix.

11. What’s On The Bench

• Antenna Launcher — I still have my pneumatic antenna launcher to complete. The wiring diagram for the electrical trigger was not clear, so I requested (and received) additional information from the vendor. But that was months ago, so I need to review it all again before I commence with the last phase of soldering. Considering that my end-fed halfwave wire fell out of the tall Douglas fir tree, completing the launcher has become more urgent.

• UHF/VHF dipole — A small dipole antenna is coming from a vendor in Oregon. I’ll strap this to a telescoping painter pole to get my LightAPRS Gateway antenna out of the garage and higher in the air. That should improve performance considerably.

• T-Beam ordered — I plan to set up an APRS-over-LoRa digipeater on 70 cm. This is different than normal APRS on UHF because LoRa uses spread spectrum radio — the two types of APRS on UHF are not compatible. Maybe I’ll discover a few more people are doing this; I hope so. The device was shipped on July 16 and may arrive in time for Random Wire 192.

12. Recipes for Radios

I had a problem with APRS on my Yaesu FTM-300DR mobile radio. VFO B is set to use SmartBeaconing while I’m mobile…but it stopped working right after I made a backup copy of the frequencies and settings to a microSD card. Thinking about it, I realized I must’ve fumble-fingered a setting while trying to remember how to write a copy of the radio to the card. It’s also possible that a voltage spike scrambled some settings.

Short version: I got it working again. SmartBeaconing had been turned off and the beacon interval was set to manual. Enabling SmartBeaconing and setting the interval to Smart solved the problem, as shown in the track below for KJ7T-9:

You can tell it is using SmartBeaconing because the dots are farther apart on the lengths of road with higher speed limits. Closer to town and in town, at lower speeds, the dots are closer together. (The other thing I’m happy to see is at the end of this track when I returned home, my own KJ7T digipeater picked up my beacon.)

The takeaway: I’m again reminded of the value of recipe cards. In the IT space, we call the step-by-step instructions to accomplish tasks not done very often by a special term: runbooks. However, I think of them as recipe cards. Even though I know some of my mother’s favorite recipes by heart, I always refer to the old recipe card while I’m making one of those dishes. The same kind of system — a recipe card for a particular configuration you want to work right every time — also works for your radios.

If keeping a notebook or computer file of particular steps seems too onerous, try some 3×5 cards contained in one group with a rubber band. This system does not need batteries or an internet connection.

For important settings you rarely change — or perhaps as memories fade with time — recipe cards can save you from an immense amount of frustration.

13. Short Stack

Community

• Join Hams Over Video Conference 24/7 — What is this? It’s a Jitsi Meet server hosted by the same ham who runs AmateurWire (Roger Radcliff KE8LCM), and it appears he has a partner in this (Murray Green K3BEQ). I tested it in a browser on my laptop and in the Jitsi app on my Android phone, and both work fine. Ham radio is about connecting with others, and this service helps you do that.

Podcasts

• HRWB 267 – Designing The Pebble Radio — “In this episode we meet Becky Spiceland N4BKY, Mike Spiceland N4FFF and Barb Asuroglu WB2CBA the amateur radio operators behind the Pebble radio project.”

• Solder Smoke episode 265 is available — listen to it or watch it on YouTube.

APRS & Digital

• aprs.world — “APRS World is a modern, open-source client for the global APRS-IS amateur radio network.” Accessible through a web browser and has an Android app.

• DroidStar 9M2PJU Mod — I installed this customized build of DroidStar on my Android phone and found it works fine. I tested it on YSF, M17, and IAX. For those of us who have been waiting for DroidStar connections over IAX to be fixed, in this build, it works!

• Learn About FlexRadio’s Aurora Transceiver from the Team that Developed It (Video) — “A few months before Hamvention, DX Engineering’s Michael, KI8R, caught up with Tony, K1KP, Steve, N5AC, and Gerald, K5SDR, from FlexRadio during Orlando HamCation for a deep dive into this breakthrough rig.”

Computing & AI

• Microsoft Just Patched a Record 570 Flaws in Windows — The takeaway? Patch now as this release included patches for three zero-day vulnerabilities.

• Someone connected a Raspberry Pi 4B to an old, unsupported Kindle, and now it’s an e-ink dashboard — Probably because I have old Kindles and some Raspberry Pi boards, this looks like a great way to combine old and new tech.

• The People Who Will Thrive in the AI Age — “The general pattern that the research points to is that many people don’t use the time they save using AI to do less; they use the time to take on new tasks.”

14. QRT: End Transmission

I thought it would be an easier week but a plumbing leak in the kitchen dispelled that notion. As I type this, I have a fan going to keep air moving and a dehumidifier operating to try and dry out the wood under the kitchen linoleum tiles.

The good news is we have a new faucet. The bad news is the cost of the plumber to install the faucet was about what a new mid-range HF radio costs. When I heard the price, I was gobsmacked. But it works well and that is what is important.

I added a few nylon mooring cleats to my little dinghy to make it easier to tie to the dock. When I say little, I mean tiny: it is a six-foot-long fiberglass dinghy I bought for dragging behind the sailboat. Right now, though, it’s at the lake house with a small electric motor on it. As long as it’s not too windy or choppy on the lake, I feel perfectly safe out on the water. On the weekends when the wakeboard boats are generating big, sharp waves? The dinghy stays tied to the dock.

We have been settling back into a regular routine since last week’s ER visit. That gave me a little more time to work on the lawn and flower beds. My lavender plants really took off this year, making the bees very happy. I have a long line of nasturtiums coming up along a border and am looking forward to those flowering, too. The bees and hummingbirds are enjoying my flower pots on the deck, and that is where I sit with my Kindle e-reader and keep an eye on my spouse through a window.

Closer to radio, I realized it’s been a while since I logged into my Mastodon account: @kj7t@mastodon.radio. I think I need a weekly reminder to do that, so I don’t abandon folks who are waiting for a response.

I have much to be thankful for and appreciate my amateur radio hobby and radio friends very much.

73, and remember to touch a radio every day!

1. QRV: Are You Ready?

This week has raced by. It’s Thursday as I type this, and I’m behind schedule because of, well, life.

On the good news front, I got my LightAPRS Gateway device working. The story is on EtherHam. What I didn’t write about is learning about APRS over LoRa 70cm radio. While that has nothing to do with the LightAPRS device, it does touch APRS. I’m curious about this so I ordered a device (T-Deck Plus with GPS on 433MHz) to experiment with. Here’s the kicker, though: when I look at a map of APRS stations using LoRa, I find exactly one in Washington State. It’s an APRS-over-LoRa desert here. However, if I get my device running properly, maybe I can make a contact with the other station about 30 miles away. I might have to find a hill with line-of-sight to Seattle to make this happen. We’ll see!

For the LightAPRS device, I have a small dipole VHF antenna coming. I intended to set that a bit higher in the air on a fiberglass painter pole because I’d like to get a better signal in and out of the digi. I’m feeling pretty good about one watt through a small mag-mount, though:

2. Thank You…Supporters!

I have several people to thank for their support over the past few months. Most support comes in through the “buy me a coffee” system but there have also been EtherHam store purchases. Everything helps and is greatly appreciated.

• Coffees: Thank you to Rich, Ray, and David.

• EtherHam store: Thank you to Jeff, Rich, Joe, and Ren.

3. New on EtherHam

• Buying Push-to-Talk Over Cellular Without Getting Taken — I researched the current market, and this sent me down a deep rabbit hole. The push-to-talk over cellular field is confusing, and that confusion leans in favor of vendors, not buyers. If you’re thinking about buying one, or you already bought one and discovered it wasn’t quite what you expected, this guide walks through what nobody tells you upfront — using a car-lot analogy that will feel uncomfortably familiar. Read more in this EtherHam-exclusive guide.

• Concept: Using the Arduino UNO Q for Home Power Monitoring — What really drew me to this idea was the Arduino UNO Q’s split-brain design: one board handling both real-time sensor capture and higher-level computing. These are jobs that used to require two separate devices. I sketched out (conceptually — not yet built) a whole-house power monitor with non-contact current clamps on the split-phase legs and an isolated voltage sensor. These are all read by the microcontroller side while Linux handles the math, logging, and dashboarding — no direct contact with bare conductors, ever. Full write-up, with parts list, firmware split, and dashboard options, is up on EtherHam.

• LightAPRS Gateway iGate Beacon Success — Troubleshooting Log — A week of ruling out CGNAT, firewalls, servers, and a suspicious libc6 upgrade led nowhere. The real fix turned out to be something I hadn’t considered: the callsign’s SSID itself. This post walks through the manual nc tests, the reboot comparisons, and the variable-isolation process that finally got KJ7T’s iGate beacons landing on aprs.fi — even though the server-side root cause remains a mystery. Bottom line: the digipeater is now working well, and I feel comfortable recommending the LightAPRS Digipeater, available from QRP Labs, as a fill-in digipeater and igate.

• Crossing the Chasm: Which Amateur Radio Technologies Actually Made It — Why did DMR and FT8 become fixtures in the shack while D-STAR and Fusion have not yet achieved that status? This piece maps ham radio’s real technology winners and still-rising projects — CW, SSB, FM, AllStarLink, AREDN mesh, and more — onto the classic “crossing the chasm” adoption curve. Borrowing a framework shared by an amateur radio friend, it asks what actually separates a solved problem from a brilliant piece of engineering still waiting for its moment. Along the way, it builds context to help you place your own favorite radios, modes, and projects on that same curve.

• Weekly Report: July 9, 2026 — The weekly report covers band conditions, digital radio news, and a digest of Groups.io messages. The Repeater Builder group was the most active this week with 142 messages posted.

4. QSO One: A New Way onto AllStarLink

There’s a new app making the rounds in the AllStarLink and EchoLink space, and it’s worth a mention here even though I’m not ready to give it the full workbench treatment over at EtherHam. It’s called QSO One, and the pitch is simple: get on AllStarLink, EchoLink, DMR, and M17 from a Windows PC or Android phone, with no node to build, no hotspot to flash, and no radio interface to wire up. Twenty-five dollars, five minutes, and you’re keyed up.

I’ve spent some time this week reading through the developer’s own posts on the AllStarLink community forum, and the backstory is a good one. A ham out of Willoughby Hills, Ohio — Frank, KD8JKK, node 46369 — got pulled back into the hobby by a bad storm earlier this year, found his local repeaters quiet, drifted over to EchoLink, heard AllStarLink had grown, went looking for a modern Windows client, and found nothing. IAXRpt has been abandoned since 2019. So he built his own, reverse-engineering the IAX2 protocol frame by frame rather than leaning on an existing library.

That’s the kind of scrappy, solve-my-own-problem origin story I have a soft spot for. It’s also exactly why I’m not writing this up as a proper EtherHam review yet.

The open beta only went live in early May. As recently as this month, the developer rebuilt the entire audio engine from scratch — giving each network its own dedicated audio lane — because testers were reporting stuttering and garbled audio, the kind of thing one beta tester colorfully described as “wrestling the alligator to the ground.” Early reports since the rebuild sound genuinely improved, but that’s a few weeks of data, not a few years. There’s also been at least one credential-storage bug that could silently drop saved logins, but that has since been patched with a self-healing backup.

None of that is damning for a young project. But it does mean the honest way to describe QSO One right now is “promising and moving fast” rather than “proven.” Worth noting, regardless of how the audio engine matures: this is, as far as I can tell, a one-person operation. One developer wrote the IAX2 stack from scratch, runs the beta, answers the forum posts, and — per the latest update — now runs the backend account system that stores your saved nodes and favorites.

It’s also a closed codebase, so there’s no way for the community to audit the IAX2 reverse-engineering work, verify what happens to saved credentials, or fork it if Frank stops maintaining it. If Frank steps away for any reason, there may be no one else who understands that code well enough to keep it running. That’s a real single-point-of-failure risk for anyone thinking about leaning on it for a regular net check-in.

So this is an app worth watching, built by someone clearly solving a problem, at a price low enough that trying it costs you little. I find it a very interesting project and hope it continues to mature and grow. I’ll keep an eye on it, and if it’s still standing — and sounding good — in a few months, it’ll earn a proper look on the EtherHam workbench.

5. Claude’s Wild Two Weeks: Export Bans and a New Default

I’ve been seeing changes over the past few weeks in the models made available in Claude, so I naturally wondered what was going on. I learned that Anthropic just had one of its busiest stretches yet. If you use Claude for anything, here’s what you should know.

I’m on the Pro subscription, so the default model change described below affects me, and that change is what caused me to dig into this topic.

The default model quietly flipped. On July 1, Claude Sonnet 5 became the default model for Free and Pro users, replacing Opus. That’s not a small swap — Sonnet 5 ships with a 1 million token context window and performs close to flagship-level Opus 4.8 on many tasks, at a fraction of the cost ($2/$10 per million input/output tokens through the end of August, then $3/$15). Opus is still there if you’re on Max, Team, or Enterprise, but it’s no longer what most people get by default. Anthropic is clearly betting that “good enough and cheap” wins more users than “best but pricey.”

If you are a Claude Pro user, three things change in practice. For those using a Claude Pro subscription, you’ll see three basic changes when you use Claude with the default Sonnet 5 engine:

• Speed and quota: Sonnet 5 is cheaper to run than Opus (roughly a fifth the cost per turn), so on a usage-capped plan like Pro, the same 45-prompts-per-5-hours allowance now stretches much further before you hit a limit. Responses also tend to come back faster since Sonnet is a lighter model.

• Capability: for the bulk of everyday tasks — writing, coding, research, general Q&A — Sonnet 5 performs close enough to Opus 4.8 that most people won’t notice a quality drop. Its 1M token context window means it can also handle very large documents or codebases in one go, which Opus couldn’t always do as cheaply.

• What you lose: for the hardest problems — deep architectural decisions, gnarly debugging, subtle reasoning tasks — Opus 4.8 still has an edge. On Pro you’d now need to manually select Opus for those cases rather than getting it automatically; Max, Team, and Enterprise plans keep broader access to it.

Net effect: most people get faster, more plentiful, “good enough” responses by default, with the option to reach for Opus deliberately when a task actually needs it.

The government forced Anthropic to pull its own flagship model offline. This is the wilder story. Claude Fable 5 and its sibling Mythos 5 launched June 12 — and three days later, the U.S. Commerce Department issued an export-control directive ordering Anthropic to cut off access for any foreign national, anywhere, including Anthropic’s own foreign employees. The trigger, reportedly: Amazon researchers found a jailbreak that got Fable 5 to write exploit code for a real software vulnerability, which was enough to get it treated like a controlled technology. Anthropic complied, access came back starting July 1 once the controls were lifted, and the company extended Fable 5’s free trial as an apology of sorts. It’s a preview of a regulatory fight that’s probably not over — frontier models getting treated like export-controlled hardware is new territory.

Opus 4.1 is on its way out. Deprecated on June 5, hard retirement on the API August 5. Anthropic wants everyone on Opus 4.8 instead, which also came with a real price cut — down from $15/$75 per million tokens to $5/$25, about a 67% reduction. If you’ve got anything hardcoded to claude-opus-4-1, you’ve got a month to migrate.

Cowork is leaving the desktop. Anthropic’s file-and-task agent tool, previously desktop-only, is expanding to web and mobile, with workflows able to keep running without an active browser session. Small item, but it fits the pattern: Anthropic is pushing agents further into everyday, non-coding work.

The takeaway: Anthropic is optimizing for cheaper defaults and broader access, while getting a real taste of what “AI as a national security concern” looks like in practice. It’s worth watching whether the export-control precedent becomes a recurring headache or a one-off situation. This is new ground for all of us.

Sources: Anthropic, TechCrunch, Forbes, Anthropic export control statement, Claude Platform Docs, Claude Sonnet 5 vs Opus 4.8: Price, Limits, and Benchmarks (SmartScope), Opus 4.8 vs Sonnet 5 vs Haiku 4.5: Which Claude Model to Use (The AI Career Lab)

6. An AI Price War Is Underway

While Anthropic was busy reshuffling its own model lineup, a bigger story has been playing out across the industry: the major AI labs are cutting prices against each other, hard.

It started at Google’s I/O conference in May 2026, where Google cut prices on its AI subscription plans — its top-tier Ultra plan dropped 20% — and rolled out a cheaper Gemini tier. Google’s Sundar Pichai pitched the math directly to enterprises: shift 80% of your AI workloads to Gemini and save over $1 billion a year. That was a shot aimed squarely at OpenAI and Anthropic.

Both companies responded. OpenAI has pricing pressure of its own to manage, and Anthropic followed with steep cuts on its side: Opus 4.8 came down from $15/$75 to $5/$25 per million tokens, and Sonnet 5 launched even cheaper at $2/$10. Meanwhile, cut-rate entrants like DeepSeek are underpricing all of them, competing with frontier-class models at a fraction of the cost.

The catch is this isn’t discounting from a position of strength. Running these models is genuinely expensive, and by most reporting, OpenAI and Anthropic are both still losing money at scale even before the price cuts. One data point making the rounds — Uber reportedly burned through its entire 2026 AI budget in four months after Claude Code adoption jumped from 32% to 84% of its engineering division. Cheaper per-token pricing helps users, but it raises the obvious question of how long labs can keep pushing prices down while costs stay high.

Sources: Welcome to the OpenAI, Anthropic, and Google price wars (Sherwood News), Following Google Gemini price cuts, OpenAI also set for price reduction wave (TradingKey), AI Price War: OpenAI and Anthropic Face New Economics (The AI Chronicle)

7. Retevis RA89R Radio: FM and Weather Radio

Retevis advertises that the RA89R includes NOAA weather radio and weather alerts. I did not find pre-programmed weather frequencies in the radio. You can enter those frequencies directly in VFO mode, or program them yourself into the radio. That seems different from other radios that advertise NOAA weather channels — at least in my experience, with most including those as pre-set frequencies in the radio and accessible with the press of a button or two.

Here’s where the mini-manual let me down. How to tune to FM broadcast radio stations and how to receive NOAA weather broadcasts are not made clear. Exploring the Retevis app on my Android phone makes it evident you can set FM broadcast radio frequencies in the app, and when you write the data to the radio, they are also saved there.

The only place I found to set NOAA weather radio frequencies is in the app, then you write the data to the radio and call up those frequencies as saved channels.

So: if you are looking for FM broadcast radio frequencies or NOAA weather stations in your new Retevis RA89R radio, they aren’t there. The easiest solution is to use the app over Bluetooth to program those into the app and then write those settings to the radio.

8. EFHW Came Down

Monday morning dawned, and since I was up early, I thought I’d fire up the IC-7300 MK2 transceiver and work some FT8 while my wife was waking up. First things first, though: check the antenna. And surprise: it had come down from the tall branch above the deck!

This could be the catalyst for me to finally finish soldering the last few wires on my pneumatic launcher kit.

9. QRT: End Transmission

Another interesting week is in my rearview mirror.

I’m on my high school class reunion committee. A few years ago — at my urging — we formed a 501(c)(7) nonprofit “social club” to govern the finances and provide activities for our surviving classmates. Why? Because some classmates have substantial net worth and did not want to be held personally liable if something went wrong at/after a reunion event. The nonprofit social club approach tackles that exposure (along with event insurance), and it also helps us not to lose track of each other in the years between events.

We actually have a great time when we get together. This year will be a picnic and barbecue, and we’ll take lots of pictures for our classmates who could not attend.

On the IT side of this, I’m in charge of the class database and the website. We had more than 500 students in our senior class, so it’s a bit of a lift to keep up with all of them. Many people have passed and we’ve lost touch with quite a few folks, but we remain in contact with more than 200 classmates. All in all, I think our news reaches about 300 classmates. While I wish it could be more, that’s pretty good, considering how many years it has been since we flipped our tassels and walked off stage.

The other big challenge this week was a medical issue with my wife, resulting in a trip to the emergency room. I always figure a minimum of four hours to get in and out of the ER — this one consumed eight hours. The ER was swamped with people, so my wife’s procedure was done in the hallway, and that’s where I stayed by her side that entire time. The important conclusion to this: she’s back home and feeling better.

73, and remember to touch a radio every day!