Internet: Weekly Summary (October 13-19, 2025)

A hurried week on the net, in a few pieces

A lot of voices this week circle the same small set of worries and longings. I would describe them as a mix of caution and homesickness. To me, it feels like people are staring at the same map of the internet but pointing to different hazards. Some shout about technical cliffs. Some worry about who gets a ticket for the ferry. Others remember a porch light that used to be on and want it back.

I’ll walk through the threads I kept seeing. I’ll nudge you toward the original posts, because each of them is worth opening like a tin of biscuits. Click the names if you want the meat. But if you just want the taste, read on.

The hype about AI agents — and why they might not bail us out

Dave Friedman has that blunt, almost cranky kind of piece titled AI Agents Won't Save You. He pokes at the shiny idea that smart agents can run errands for us online — book a flight, manage subscriptions, do our online banking. I’d say his point is simple and stubborn: the internet today is not a free, open marketplace of services waiting for agents to roam. It’s more like a gated mall where each store keeps its back door locked.

He argues that agents need interoperability. They need to be able to move between services smoothly. But the big platforms are more often building walls than doors. To me, it feels like trying to teach a dog to use a subway in a city where every turnstile needs a different token. Maybe the dog gets in at some stations, but not all. Friedman keeps coming back to that image: promising tech meeting closed ecosystems. He doesn’t deny the cleverness of the agents. He just warns that cleverness without access is a neat trick with nowhere to go.

That skepticism crops up elsewhere too, but in different forms. Which is interesting — people keep circling the same problem from different angles.

Internet access as the real gatekeeper to AI

The Microsoft-ish study summarized by MBI Deep Dives takes a more global, statistical look. It introduces a neat metric called AI User Share — basically how many working-age people in a country actually use AI tools. The headline is almost boring in a useful way: internet access matters more than curiosity.

If someone in Lagos or Lima or a small town in India has a decent connection, they adopt AI tools fast. But get rid of connectivity, and adoption collapses. The paper ties AI use to GDP per capita too, which is no surprise. Wealthier countries are ahead, and they might even be approaching a ceiling. But the surprising bit is how much of the gap is about infrastructure, not attitude.

I would describe the takeaway as practical: if global AI is a party, then the internet is the bus. Fix the bus, and more folks can show up. The research quietly reminds you that a lot of the AI conversation — all the metaphors about agents taking over work — assumes a baseline level of connectivity that doesn’t exist everywhere.

It’s like expecting everyone to join a phone call when some people still wait outside for a payphone.

Supply chains: the hidden, boring, terrifying limit

Then there’s Heather Flanagan with a cold dose of reality. Her piece, Why Tech Supply Chains, Not Protocols, Set the Limits on AI and the Internet, pulls the camera back from dreamy software metaphors to dirt, mines, factories, and power plants.

Flanagan’s argument is the sort you read and then notice at dinner: chips, minerals, power, skilled hands — these are finite. Protocols and standards are negotiable. Supply chains are not easily waved away. She draws attention to geopolitics and demographics. A country can have brilliant engineers but no raw materials. Another can have materials but no fab capacity. The global network we lean on is fragile and oddly local at heart.

It’s a useful reminder. We talk about the internet like it’s weightless. She calls out the myth that software scales freely. I’d say she’s the one saying, plainly, that AI is not manna — it’s logistics plus energy plus people. It's a humbling image: think of data centers as farms needing water and fertilizer.

And that begs another thought: if supply chains are the bottleneck, then regional resilience matters. Not everything can be global if you want it resilient. That’s a theme that threads through several posts.

Hooks, projects, and the old school's stubborn hope

Doc Searls writes in a quieter, older register. His piece My Three Hooks is part memoir, part manifesto. There’s this ache for the internet to be a place that enlarges personal agency rather than squashes it. He talks about projects — MyTerms for privacy, a personal-AI project, and News Commons for local journalism.

Doc’s tone is steady. He worries that journalism and broadcasting used to be the weight anchors for public life, and now they’ve been leeched by social algorithms. I’d say his writing feels like someone cleaning out an attic and finding a few useful tools. He’s trying to stitch new tools into old values.

There’s a strange comfort in that. The projects are small and practical. They don’t promise to rewire the internet overnight. They’re more like lanterns you carry down a dark hallway, one at a time. People who like tinkering with real fixes will find his pieces warm. People who like fireworks will find them slow.

Remembering the early net: modems, newsgroups, and a porch light

Nostalgia is alive this week too. Phil Gyford takes us back to 1995 in My first months in cyberspace. The modem sounds, the fiddling, the real joy at discovering newsgroups and chat rooms. It’s tangible — the clunky steps you had to take to get online. There’s a sense of community in those stories that is not romanticized; it’s practical. You learned certain rules, you met the same faces, and sometimes you felt like you belonged.

Ruben Schade echoes that nostalgia from another angle in Rubenerd: Putting the www back in URLs. It’s a small point — the presence or absence of 'www' — but Ruben uses it as a proxy for a bigger loss. The web used to be a collection of places, each with its own address and identity. The removal of 'www' and the rise of unified, branded domains feels to him like a consolidation of personality. That is, websites feel more like storefronts owned by big chains than mom-and-pop shops.

Both pieces tap the same chord: the early web was more porous. You could set up a page, and the address mattered in a different way. I would describe their mood as mild mourning. It’s not all gloom. The memories are practical, like old recipes. They remind you that decentralization was not just a political stance; it was everyday plumbing.

The little communities that still try to keep the lights on

On that note, Justin Smith-Ruiu writes Keeping House, which is a newsletter-ish peek at community efforts — the Hinternet livestream, an essay prize, and a little absurdist game where folks become secret mayors of remote towns. There’s grief tucked between the jokes: lost TypePad archives, the pain of content disappearing.

This post reads like a neighborhood noticeboard. It’s messy and human. The movement Justin writes about — the Attention Liberation Movement and the Hinternet — wants to push back against attention-hoarding platforms. The strategies are small and sometimes goofy. But they’re real attempts to preserve different ways of being online.

The message I kept getting was: people are building in public, imperfectly. That’s where hope lives for some of us. Maybe it’s naive. Maybe it’s stubborn. Either way, it’s happening.

Writing for an audience that might not be there

The tone of unease shows up in Olu Online in upload. Jay Hoffman — or the author behind the site — is plainly wrestling with the category of writing online. Who reads what you write? Does it land? Is it worth the bother? It’s an old question with a new sting: we have more publishing tools than ever and, paradoxically, less certainty about connection.

There’s also a small digression into an alphabet superset challenge and the idea of long-form things like books or documentaries. The writer doesn’t fix any grand theory here. He just worries and speculates in a way that feels familiar and honest. I’d say that many bloggers will nod along. It’s the mood of an older craft in a faster marketplace.

The problem with treating the internet like a person

Martin Haehnel writes Hub 'The Internet Is Not A Person', and he’s annoyed in a constructive way. He argues against the habit of treating abstractions as if they had intent. You hear this a lot: ‘the internet decided’ or ‘society wants’. Haehnel says that’s sloppy thinking. The internet is a tangle of systems and people, not a single actor.

That matters because if you want change, naming that actor — usually a politician, an executive, a business model — matters. The danger, he says, is that blame floats away into the ether. It’s like yelling at the weather. I would describe his critique as a useful practical correction: don’t mistake the map for the territory.

There’s also a political edge: thinking of the internet as an organism can lead to performative fixes that don’t hold up. He pushes us to put names on things and then push those named actors. It’s a small test but a necessary one.

Recurring themes and the places they meet

After reading all these pieces, a few themes kept popping up like stubborn weeds. They’re worth listing because they show where readers and writers are concentrating their energy.

• Access versus hype. Several posts return to this: you can dream up AI agents and spectacular services, but if the basic question of who is on the other end of the modem isn’t solved, the dreams are gated. There’s a pattern: big promises meet small, practical limits.

• Physical constraints. Flanagan’s piece is the clearest example here, but you feel the same idea in other posts. Software lives on hardware and in supply chains. That’s a boring sentence that matters. Think about it like this: you wouldn’t expect soup to make itself. The broth needs pots, water, and a stove.

• Local, small, and stubborn projects. Doc Searls and Justin Smith-Ruiu are writing about methods of small-scale repair. They’re not trying to overthrow platforms in one tweet. They’re trying to build tools and rituals that let people keep publishing, keep local journalism alive, and keep archives from vanishing.

• Nostalgia as critique. The remember-when pieces aren’t just wistful. They’re diagnostic. The internet used to allow different forms of identity, address, and publishing. Now consolidation concentrates power and flattens variety. That’s a loss some folks are trying to quantify and some are trying to repair.

• Naming and responsibility. Haehnel’s point about personification is practical. If you want change, you have to name the actors and nudge them. Blaming the internet rarely moves anyone who actually makes policy or product.

• Writers’ uncertainty. Several posts show the emotional labor of writing. There’s a thin, aching question: why write, when attention is extractive and fleeting? People answer that in different ways: some keep building utilities, some keep telling stories, some keep trying little experiments.

Where people disagree — and where they quietly agree

Disagreement is mostly about tone and scale. Dave Friedman is skeptical in a practical, almost pessimistic way about agents. Doc Searls is stubbornly hopeful in a small-project way. Flanagan is pragmatic and sober about limits. The MBI piece is quietly technocratic: it treats the problem as measurable and possibly fixable through infrastructure.

But underneath those differences there’s a modest agreement. Nobody is saying the web as it is is perfect. Nobody is offering a single, giant fix. Instead, people are pointing at particular levers: better access, resilient supply chains, local projects, clearer naming of responsibility. It’s less a chorus and more a neighborhood meeting where everyone brings a different casserole.

Little disagreements that matter

• Agents and openness. Friedman sees checked doors; others see potential if infrastructure and policy change. That’s a real fork. If platforms stay closed, agents are toys. If they open up, agents could be useful — but then you bump into supply chains and business incentives.

• Centralization versus resilience. Ruben and Doc both worry about consolidation. Flanagan says regional supply chains are necessary for stability. You can agree on the problem but differ on whether the fix is policy, commerce, or community action.

• Storytelling and craft. Olu’s worry about readership is shared, but some say keep going because the craft itself is the point. That’s less a disagreement than an emotional variation on the same theme.

Small notes, side streets, and odd little delights

• The secret mayor game in Justin’s post is dumb in the best way. It’s a reminder that not all responses to the internet’s ills have to be heavy-handed. Sometimes you need a laugh to keep going.

• Ruben’s plea for 'www' is charming because it shows how tiny technical choices become cultural markers. It’s like arguing the merits of tea in a specific cup. The argument is partly symbolic, but symbols matter.

• Phil’s modem stories read like a family album. They’re not a policy treatise, but you don’t need policy to remember how community feels.

What to read if you want to dive deeper

• If you want a skeptical look at agent hype, start with Dave Friedman.
• If you’re curious about how many people actually use AI and why, read the MBI summary by MBI Deep Dives. It’s the most empirical piece.
• For a reality check on hardware and limits, Heather Flanagan is the one to read.
• If you want patient, practical projects and a certain old-school internet ethic, Doc Searls offers concrete ideas.
• For nostalgia and early web texture, Phil Gyford and Ruben Schade are the tiny, warm lamps you’ll want on.
• For community experiments and a messy, human take on keeping archives and attention alive, check Justin Smith-Ruiu.
• For the writer’s mood and the quiet doubt about readership, Olu Online has a thoughtful piece.
• For a useful corrective on how we talk about the internet, read Martin Haehnel.

Little analogies that stuck with me

• The internet as a gated mall versus a neighborhood of shopfronts. Friedman and Ruben live in different apartments in that mall, but both complain about locked backdoors.

• AI adoption as a party and internet access as the bus. The party can be great, but the bus matters first. The MBI study makes this concrete.

• Supply chains as farming. Flanagan’s reminder that chips need power and ores is like saying you can’t have fresh bread without a mill and a farmer.

• Local projects as lanterns in a hallway. Doc Searls’s projects are small lights, not floodlights.

A few tangents that matter

Sometimes the conversation drifts into small bits of policy or play. For instance, the idea of naming actors (Haehnel) feels slightly bureaucratic, but it’s a practical tool. It’s like calling a plumber by name when a pipe bursts instead of moaning that ‘the house is broken’. Slightly less romantic, more useful.

And the nostalgia pieces — they’re not just longings. They’re models. When someone says they want the web to be more like 1995, they often mean: smaller identity silos, easier publishing, more clear ownership of your corner of the web. That’s a specific thing to design toward, not just a wistful picture.

Final little worry and a faint hope

There’s a through-line of worry this week. Hype about agents, the limits of supply chains, consolidation of identity, and shaky attention economies — they all point to a fragile present. But there’s also a steady pulse of people doing repair work. Sometimes the projects are twee. Sometimes they’re practical. Sometimes they’re stubbornly local.

I’d say the mood is realistic rather than fatalistic. People are naming problems, sketching small fixes, and reminding us that access and infrastructure matter as much as algorithms. If you want the juicy debates, start with the posts I linked. They’ll take you from the high talk about agents down to the practical dirt where things actually happen.

Read the posts if you want the long versions. Each author brings a different hand to the same set of questions. Some hands are patient, some are blunt, some are fond. They all matter in a neighborhood that keeps changing its fences.

If you do click through, you might find a rant, you might find a plan, you might find an old modem booting up in the margins. That’s the internet right now — messy, stubborn, and a little bit alive.