Doc Searls's Blog, page 11

Departments of Correction

Fortunately, we've already got it here: unlimited 2GB/s symmetrical service for $59/month. Bloomington's city fiber rollout has been paused by the mayor. Here's square's story about it, which is also in Bloomdocs—an example of a news commons at work.

One more reason to move off Chrome? A URL that begins with chrome-extension://efwhaddfugisallthisjiveepwnj/ before it gets to http:// is not a URL.

My record is about 20 feet—in opposite directions. And under furniture that's hard to move. You drop your AirPods case on a hard floor. How far do the pods fly away from their popped case?

Also wanted: knobs and switches. Buttons are coming back to dashboards.

Their advice: If you want mail to be postmarked today, hand it to a human at the window. Letters we dropped off through a slot at a Los Angeles post office on April 9 were not postmarked and sent until the 15th. The reason: they were slammed and shorthanded.

Re-reading Material

On our new digital age:
Some New Ways to Look at Infrastructure was the first draft of
What does the Internet make of us?, but is worth reading because the stuff about infrastructure mattered and was dropped in the second piece. (2017)
Will Our Digital Lives Leave a Fossil Record? (2020)
How early is Digital Life? (2020)

On the decline of over-the-air broadcasting:
The new business of free radio (2008, with additional remarks by Jonathan Marks)
What happens after TV's mainframe era ends next February? (2008)
What happens after TV's mainframe era ends next February? (2008)
Doc Searls on VRM and the future of radio (2009)
The past and future of radio (2010)
March Madness and radio's future (2010)
Why music radio is dying (2011)
How to rescue radio (2013)
The holy grail of radio (2013)
AM radio declared dead by BMW and Disney (2014)
A conversation with Doc Searls on the future of radio (2014)
How will WMAL survive losing its transmitter? (2015)
DIY radio with podcasting (2016)
The slow sidelining of over-the-air radio (2016)
FCC is just making a bad thing worse (2016)
Where Public Radio Still Rocks (2019)
How long will radio last (2020)
How the once mighty fall (about WMAL, above, in 2020)
A look at broadcast history happening (2023)
The Tall Tower Market Collapses (2024)
Radio's Death Knells (2025)
When TV Went Digital (2025)—a reposting of the second item from the top of this list, with some updates

TGI-Fi

Whole Lotta Badshit Going On. The latest 404 has a weekend worth of it.

Surprised this one didn't come sooner. Want the feds to stop funding public broadcasting? Fine. There's an argument for that. (I made one, way back in 2008.) But bias, which is everywhere (because the voice from nowhere is insincere and boring), claimed by POTUS, is the wrong reason. Woulda been better to just say public broadcasting should stand on its own. (Which, for the most part, it has been, for decades.)

Visitations

It's still creepy. World showed off its Orb a few years ago at IIW to approximately no applause at all. A lot seems to have happened since then. Here's what Wired says about it.

Such as the "@ username" in the Profile setting. I'm in a group zoom-like conference call (is it a "call"?  a "session"? not sure) on Signal, where I'm learning (partly on my own) how Signal is changing. For example, one has up to three IDs: the handle you use in conversations, your phone number, and an @username, which is new to me. There's an appearance feature that lets you hide it in case somebody gets unwelcome access to your unlocked phone. Other stuff too. My main issue with Signal is how different it is on the phone and on the computer. Also, the conference (zoom-like) thing needs chat. I don't see it there. Some things are still not clear to me, which is okay, I guess.

Old writing style, new habit. Wordland is easier than tweeting. Postings also go straight onto my blog, not onto some platform that (trust me, or Dave) that has to work hard not to enshittyfy itself.

That aren't by an AI. Writings about AI: ChatGPT, Claude, Gemini, & Co.: They Are Not Brains, They Are Kernel-Smoother Functions, "Attention", "Transformers", in Neural Network "Large Language Models", and everything here.

Same nightmare. David Strom expands on the privacy perils of the connected car. 

As I recall, anyway. Just learned from Bill Wendel that firstpodcasts.org exists, and is mostly right. 

The other Durant. Got some nice comments to The best 3-point shooter you never saw, which I posted four years ago.

Shall I try Kagi? Google search is woefully enshittified.

Grifting away

According to Joseph Cox at 404, the U.S. government registered the domains

thetrilliondollardinner.gov,dinnerforamerica.gov, andthetrillion.gov.

All of these (correct me if I’m wrong) are about enriching the U.S. president, his family, and favored friends through a memecoin scheme by which anyone (say, Putin) can buy influence.

There is nothing conservative about this. Nothing republican. Nothing that will make America, or anyone, great, much less again.

It’s a something with an intrinsic worth of nothing, and a scam straight out of the old playbook.

There is no pale beyond which shit like this cannot go.

Flyings

Neither here nor there, yet. At BWI, about to board an SWA flight to IND. There are thunderstorms between the airports and more on the way, but no delay so far. Meanwhile, this gives me  chance to talk weather apps. My current fave is Windy, which visualizes winds at all altitudes, real time lightning (with visuals and sounds), temperatures, satellite and radar imagery, changes past and future, and more. Second choice is Weatherbug.  I'd say more but need to board.

Of possible interest

In reply to the question "Which blog post took the most work?" the answer is From Hollywood Park Racetrack to SoFi Stadium, which spanned 2005 to 2024 and involved taking many dozens of aerial photographs on approach to LAX, doing research, and writing about it, hoping it helps historians along the way.

Two angles on one story. Noozhawk says Tri-County Produce, an independent local landmark store, has been saved by a longtime employee and his son. MSN says the purchase was made possible by local customers Eric and Wendy Schmidt.

Arendt called it terror. Steven Weber on radical uncertainty, in Barron's.

Signs of enshittification. SiriusXM's app sucks. It used to have Favorites, which were like buttons on an old car radio: a collection of channels you chose to put on a simple collection that you could arrange your own way. That's gone. Now the front of the app has Based on your morning listening and Channels for you below a bunch of tabs: For you, Music, Talk & Podcasts, Sports, and Howard. What used to be Favorites is now "Library" on the bottom right of the screen. These are topped by "New episodes," and then go down a list that changes based on… I don't know. Probably guesswork by an algorithm. Also on the bottom are Discover, Channels (remember those?), and Search. Imagine a grocery store where at the entrance you have to get past a pile of stuff that's some algorithm's guess at what you like, and everything you come for isn't where it was before, and won't be in the same place tomorrow. That's SiriusXM now.

The pain is coming. Am I seeing less marine traffic to the U.S. than ever? Looking for historical data. I'm sure many are digging deep into intermodal and logistics stats to gauge the effects of tariffs and economic uncertainties caused by White House wackiness. Got links? [Readers answered! Logistics Managers Index (and the latest stats), CNBC on falling traffic in Los Angeles and Long Beach. HT to old pal Chip Hoagland.]

Where safety is a slogan. Wired says your car is snitching on you. As Mozilla made clear almost two years ago, privacy is an inconvenience: for the car makers, for their commercial "partners," and for the cops. Consumer Reports says the same is true for TV makers.

Nice! I just noticed that Wordland now lists all my (too many) categories. Taking advantage of that now.

Today's title is how I view everything I write.

Terminations

Looks like The Terminator took out a bunch of Starlink satellites. (SpaceWeather.com)

Pasts and Futures

The Airbus 390 is gigantic. Like a cruise ship for the sky. Except, I gather, it's not real.

Pope Francis dies on Easter Monday. For a sense of what Francis embodied and meant, I recommend Jonathan Rauch's Cross Purposes: Christianity's Broken Bargain with Democracy.  I'm in the midst of listening to Rauch on Russ Roberts' Econtalk podcast, and I think his case is sound, caring, brilliant, and timely. (Arg: The episode isn't listed yet.)

Ad growth is decelerating.

An Inadequate AI Solution

Let’s take the example of printers, which tend to suck. Here is one example of suckage at work, and the instructive hell it put me through:

The Canon MG3600 is my wife’s. It sits by her desk and does a good-enough job.

The Epson WF-3520 is mine. It sits by my desk and leaks black ink. It has been doing that since I bought it in a hurry for $35 at a Goodwill store when we first got to town in the summer of ’21. The drip runs at about a drop per month, but otherwise it’s been okay. I use it mostly for scanning. It’s good at that.

Or was, until I encountered the problem above.

Answers to questions about the error code proved fruitless in a series of searches (such as this one) and at the Epspon link above.

So, naturally, I went shopping for a new printer. My only requirement was that it would scan at least as well as this old Epson.

Alas, scanning is a checkbox for “all-in-one” printers. Here’s an example from PC Magaziine. Nothing much about scanning. This makes sense, because demand for high-quality scanning is relatively low, and most printers are at least okay at it. But Consumer Reports (bless them) does rank scanning quality, and it gave me this in reply to an Ask CR question:

The Canon PIXMA G3262 is not one of CR’s recommended printers, rating a middling 47 out of a possible 100 score. Its problem areas are low predicted reliability, low versatility, and slow text speed. The only thing it’s good at is scanning. The printer page on Amazon is also not encouraging. It has only two user ratings, and both call it junk.

The Brother MFC‐L8905CDW is $699.99 at Amazon, which is way out of my range.

The Canon Pixma MG7720 is now in the used market, and (far as I can tell) available only on eBay, and none of the offerings there look interesting.

So I went looking just for scanners, and narrowed my choice down to the Canon CanoScan Lide 400, which is just $78.86 on Amazon. I put it in my cart, then I thought, Wait a minute. What if I asked an AI what that error code means?

So I did, with ChatGPT. The dialog:

Me:

What does “Failed to open connection to device (-21345)” mean on a Macbook Pro running Sequoia 15.3.1 that is trying to connect to an Epson WF-3520 printer?

ChatGPT:

“Failed to open a connection to the device (-21345)” on your MacBook Pro running macOS Sequoia 15.3.1, when attempting to connect to an Epson WF-3520 printer, typically indicates a communication issue between your Mac and the printer, particularly affecting scanning functionality. Here’s how you can troubleshoot and resolve this issue:

Step-by-Step Solutions
1. Enable Local Network Access for the Printer/Scanner
macOS Sequoia introduces enhanced privacy settings that may block devices from accessing the local network. To ensure your Epson WF-3520 has the necessary permissions:​

Navigate to System Settings.Select Privacy & Security.Click on Local Network.Locate your printer or scanner in the list and ensure it’s toggled on.​
Apple Support Community
Reddit

This step has resolved similar issues for many users experiencing the (-21345) error.

It listed other solutions, but that one worked. Three levels down in System Settings, the printer had been toggled off, presumably by Sequoia’s enhanced privacy settings. I am now back to scanning normally with my lousy old Epson.

I could take my learnings here in at least two directions:

I could carp about Apple failing to tell customers about potential hassles created by enhanced privacy settings. Ask ChatGPT “What kinds of things do Apple’s enhanced privacy settings (introduced with Sequoia) break or at least inconvenience?” and you’ll get a long list of answers, starting with the one I encountered with the printer. To my knowledge (always subject to correction), Apple hasn’t bothered with the same. (Wouldn’t providing customers with that kind of list be nice work for Apple Intelligence, which so far is an embarrassing fail?)I could compare how ChatGPT, Claude, Gemini, Meta, CoPilot, Perplexity, and other early AI leaders perform on challenges like this one, but that would be a waste of time. That’s because, for all the good they do, they are not ours.

See, AIs of the usual sort don’t know anything more about our personal stuff than we tell them, and we may not want to do that, just for privacy reasons. Also, they tend to forget. And they are not ours. At best they are personalized, rather than personal.

A truly personal AI (or pAI) would be all mine and trained on what is under my control. For example, all my property, including these two printers. It would have data about my purchase and use of these printers. But it would also have access to data about identical or similar printers out in the world, and ways to engage the corporate AIs so we could solve problems together in far better ways than what we have now. In the process, the likes of Epson would have better intelligence about product usage than they’re getting now.

So we are talking here about agent-to-agent stuff, with the right capabilities and resources located in the right places.

Toward making that happen, there was lots of talk earlier this month at IIW XL (our 40th in 20 years) about MCP (the Model Context Protocol), A2A (agent-to-agent protocol), and GNAP (Grant Negotiation and Authorization Protocol), all of which hold promise. I’m not here to debate or advocate any of them, just to note that there are protocols out there, as options with which we can work.

But what about the entities using those agents—companies making and selling products on one side, and customers using those products on the other? How can they learn from each other, and not just make transactions and then get frustrated with each other when problems show up?

One answer is Market Intelligence That Flows Both Ways. I wrote that in 2016: nine years ahead of its time, so far.

Looking back, I don’t think that kind of flow will happen without AI agents, and full agency by those agents operators, on both sides.

We need to be able to lead in our dances with companies in the open marketplace. But first, we need personal AI to help us gather and manage intel about our lives, our things, our relationships (including the medical and financial kind), our travels, and the services we use. We are not close to having any of that, and not helped by the Apples, Microsofts, Amazons, and Googles of the world, all of which want to confine our capacities to their walled gardens.

We don’t need to knock the walls of those gardens down. The revolution we need will simply obsolesce them.

That’s why it will not be supplied by APIs, or by anything that requires that we have an account. We need to be independent and self-sovereign: first parties who speak for themselves in first-person voices.

I’ll say more about all that in my next post and others to follow.