Innovation: Weekly Summary (October 20-26, 2025)

I would describe this week's chatter about innovation as a messy garage of ideas. Some tools are shiny. Some are dusty. Some smell like fresh coffee, and some smell like burned toast. To me, it feels like everyone is trying to build the next thing that changes the kitchen, the city, or the way we talk to machines — and they’re arguing about whether you need a master chef or a neighborhood cook to pull it off.

Old companies and lost mojo

Phil McKinney tells a story that reads like a family photo album with edges yellowed. He goes back to 2007 and a conversation about "20% time" — that famed Google idea where engineers spend part of their week on side projects. At first glance it’s a policy argument. But he ends up with something else. He learned from Art Fong that HP had once done the same thing. The twist is a cultural one: innovation wasn’t a policy it was relationships, trust, and clearing the stuff in people's way.

I’d say his piece feels like watching an old neighborhood slowly change. People who knew each other made things happen. When those people left or stopped talking, the magic faded. He built an Innovation Program Office and it worked. Then he retired and, sadly, the company drifted back. You can almost smell the paint peeling off the innovation posters in that tale. It’s a useful reminder that rules can’t replace trust. Rules can nudge, but they don’t bake the bread.

This one made me think of a family workshop where grandpa taught you how to use the lathe. You learn by being around the tools and the people. If grandpa stops coming, the lathe gets dusty. That’s not a unique story. It’s human.

If you want to poke at the spot where corporate memory frays, read Phil McKinney. It’s practical, and a little sad.

Tools, chips, and the hardware sprint

There’s a big thread this week about tools that actually do the work. Nate did this sprint through 100+ AI tools and picked 15 that matter. He isn’t romantic about AI hype. He’s looking for tools that replace whole apps, or at least do parts of work differently. His categories — data proximity, determinism over vibes, artifact ownership — are blunt and useful. I would describe them as three ways you can tell if something will actually help you get a job done, rather than make you feel busy.

Data proximity is an old idea dressed in new clothes. If your model can touch the right data quickly, it beats a fanciful model with no data connection. Determinism over vibes is my favorite blunt phrase. It’s about whether the tool gives predictable outputs you can rely on. Artifact ownership is about who gets to keep and own the thing the model makes. That last one matters more than people think. If your AI helps you write a contract, who owns the draft?

Then there’s the hardware angle this week. A Paderborn start-up called One Ware — yes, a small German town — is promising hyper-individual AI models in seconds on a couple of chips. Georg Kalus wrote about them. To me, it feels like someone trying to make a pocket microwave for intelligence. You don’t need a data center. You need tiny, focused smarts near the edge. Sounds neat.

At the same time, Europe is trying to not be left behind. Georg Kalus also wrote about the Munich Advanced Technology Center for High-Tech Chips (MACHT-AI) at TUM. Bavaria is ponying up money to teach chip design. That feels like a village deciding it won’t import everything from the city anymore. Chips are a long game. They need fabs, talent, and time. But training people is the first step. If you’ve ever seen a small-town bakery try sourdough for the first time, you know the enthusiasm. It might be messy at first. But it can work.

Microsoft’s new Copilot companion, Mico, came out with a lot of theatrical color changes. Brian Fagioli called Mico a "major doofus." I’d say the companion looked a little too much like a safety doll and not enough like a real partner. The tech under the clown paint is important: long-term memory, privacy toggles, health and education features. But the face matters. If it looks like a toy, people won’t trust it with serious stuff. It’s like buying a Swiss Army knife that has a tiny plastic blade. Cute, but will it hold up?

Smart glasses made a clear push this week too. Keenen Charles wrote a thoughtful piece about them, especially now that Meta’s Ray-Ban Displays are out. The hardware finally looks useful. But will they be as good as phones? Maybe not. My take: smart glasses are more like a really handy pocketknife. They’ll change specific chores — navigation, quick lookups, overlays — but they won’t replace the stove.

A close-to-browser riff came from John Lampard. He’s wary of OpenAI’s ChatGPT Atlas. He argues it’s not a browser but an AI aggregator with a lot of bells and whistles. It makes me think of a roving salesman who claims he has everything under the sun in his suitcase. Sometimes that’s useful. Sometimes you miss the old shelf of trusted books.

And then there’s Leica. Fatih Arslan wrote about the Leica M EV1, the first M-system with a built-in electronic viewfinder. For gear heads this is big. Leica’s pace is slow but deliberate. That’s not flashy Silicon Valley speed. It’s more like a craftsman carefully sharpening a blade. The EV1 fixes a real annoyance — close-up focus — and it’s a real innovation for a niche audience.

Put these together and you see a split. A chunk of the conversation is about making AI portable and useful where people actually live. Another chunk is about looks and trust. The hardware and interface choices matter just as much as the model in the box.

People, demographics, and network states

A whole other set of posts zooms out to the people side. Heather Flanagan lays out how demographics shape the Internet’s future. Her point is plain: technology doesn’t float. It rides on human talent. If a country’s population shrinks or ages, that affects who builds the web and what they build. Africa’s demographics get special attention. Young populations there could drive huge change — if investment and standards work include them.

I’d say that her piece feels like a town meeting where half the chairs are empty unless someone brings snacks. Demographics are the chairs. Talent needs space and incentive. If you don’t seat young people at the table of standards and protocols, you get an Internet shaped by older tastes.

That ties to the network state idea from Peteris Erins. He writes about Network States as digital communities that compete for residents. Trust and acquisition strategies are key. Some of his critique warns about niche networks becoming exclusive. Which is a real risk. Ask any club that started with good vibes and ended up with a membership list. The appeal of a network state is seductive: choose your rules, pick your neighbors. The danger is ending up in a gated cul-de-sac.

There’s also a more civic angle. Santi Ruiz covers the Recoding America Fund — $120 million to make government systems actually work. This is not glamorous. It’s plumbing. But the premise is simple. Better recruitment, better systems, and a culture that tolerates experiments could matter more than another shiny app. If the public sector can get nimble, it changes the whole field. It is, as they say, less about unicorns and more about seatbelts.

Finally, Anil Dash pushes back on the media habit of celebrating funders over founders. His argument lands like a friendly shove: coverage that worships VC term sheets instead of the people inventing things paints a warped picture. I would describe his tone as impatient in a good way. He wants readers to notice the folks doing the work, not just the ledger.

All of these stories are linked by one idea: people matter. Talent pools, community rules, civic capacity, and storytelling that centers makers — they shape what actually sticks. Innovation without people is like a bakery with no customers.

Alchemy, levers, and the promise of turning little into a lot

A couple of writers leaned into metaphor this week. Shawn K wrote a lovely, slightly weird piece asking: what if alchemy had worked? He draws a line from chrysopoeia — trying to turn lead into gold — to what AI can do today. The point is not mystical. It’s practical: AI lets people create outputs with tiny inputs. Design mockups, repairs, writing drafts — that’s modern alchemy.

To me, the alchemy metaphor works and also trips over itself. Making things out of little is exciting. But that magic needs rules and incentives. An AI that makes value from nothing could upend markets. It’s like getting a windfall of tomatoes every week. Suddenly the local farmer’s market gets weird. You either eat the tomatoes or figure out how to sell sauce.

Rob Snyder’s piece on "divine levers" gets at the motive side. He worries that entrepreneurs often chase selfish goals and forget the demand they should be serving. His 'divine lever' is a philosophy that serves demand without selfishness. It’s a bit lofty. But I’d say it’s grounded: products succeed when they meet real needs, not when they serve an ego.

Link those ideas to the hardware and tool conversations and you get something interesting. If tiny chips let you make bespoke intelligence, and if that intelligence is used for selfish scalping or for useful work, the results will differ. Motivation matters. Intent matters. A tool is just a tool. It becomes alchemy in people’s hands.

Snippets, misalignment, and small but potent details

Luna Nova dropped a grab-bag called "202506 Snippets." It’s scattered, in a good way. Laser weeding to cut pesticide use. The elimination of lead in developed countries. Carbon sequestration projects that read like sci-fi. A heads-up about AI misalignment scenarios. A note on Rust tools for data viz and type checking.

These snippets are the kind of thing you glance at and then tuck into the back of your brain. Laser weeding made me think of the farmer in the next county who can finally stop reaching for poison and start reaching for a laser pointer. The environment stuff sits like a pebble under your shoe. You can ignore it. Or you can walk differently.

Misalignment is the shadow in many of these pieces. It’s the sense that models at scale can behave in ways we don’t want. It’s a low hum in the background. Not a siren yet, but you notice it on quiet nights.

Notes on hype vs reality

There’s a persistent skepticism this week about the hype cycle. One voice says we might be leaving the peak behind. Pessimism like that often smells a bit like realism. Georg Kalus cautions that many startups will struggle to deliver. The heat is on to produce real chips, real models, and practical value. That’s an echo of what Nate saw when picking 15 tools: the useful tools are those that replace an existing workflow, not just make it look fancier.

The skepticism shows up in small ways too. Microsoft’s Mico looks like a companion but feels sanitized. OpenAI’s Atlas promises a new way to gather knowledge, but John Lampard points out it’s not a classic browser. These kinds of things make you wonder: are we inventing actual helpers or just repackaging old habits with new paint?

A few human-sized analogies

• Innovation as a garden. You don’t grow tomatoes by buying soil and leaving it in a bucket. You talk to the neighbors, plant, weed, and water. HP’s lost mojo is a neglected plot. The MACHT-AI center is a community garden starting from seeds.
• Tools as kitchen utensils. Some are knives that change everything. Some are fancy gadgets that look good on Instagram. Nate’s list is a drawer of actual knives and wooden spoons. One Ware’s tiny AI chips are the pocket can opener you didn’t know you needed.
• Network states as neighborhood clubs. Some clubs are open and noisy. Others close up and gate themselves. They all start with a handshake. Trust decides whether they last.

What kept popping up

If you skim the week, a few themes recur like a chorus you can’t shake:

• People and culture matter more than policies alone. This is not glamorous. But it’s true. Phil McKinney makes this plain. So do the pieces about public sector talent from Santi Ruiz.
• Hardware is back in the spotlight. Chips, smart glasses, and edge devices matter. Europe wants a seat at that table. That could change the balance of power in some industries.
• Tools that actually replace work — not just make it look slick — are the ones to watch. Nate said that best.
• Demographics and who gets to speak at standards tables will shape what the Internet looks like next. Heather Flanagan and Peteris Erins nudge this forward.
• Hype has a shelf life. Some startups will make it. Some won’t. And the ones that do will usually be useful in a specific way, not magical for everyone.

Where to poke next

If you want the practical hooks, start with Nate for tools and prompts to evaluate them. If you want stories about corporate memory and why things fade, read Phil McKinney. For Europe and chips, Georg Kalus is the one to follow. For a human-centered warning about who builds the web, Heather Flanagan is sharp. If you like a quirky historical take about what it means to “make gold,” go with Shawn K. For a practical civic angle, read Santi Ruiz. The Leica essay from Fatih Arslan is small and satisfying if you like gear-talk.

I’d say the best way to use this week’s reading is like grazing. Don’t try to swallow it all at once. Pick a strand — tools, people, hardware, policy — and follow it for a while. You’ll see how the pieces fit.

That’s the shape of it. There is energy. There is worry. There’s hardware and soulful complaints about culture. There’s the smell of progress and the echo of the hype machine. Read the pieces if you want the detail. Or don’t. Either way, the week shows innovation as something you do with people, parts, and patience. It’s not a launch button. It’s a slow, stubborn practice.