Innovation: Weekly Summary (December 01-7, 2025)

The week’s pieces on innovation felt like a messy, useful toolbox left open on the kitchen table. Some tools are shiny and new, some are rusty but trusted, and some are half-finished prototypes that you keep poking at because they might just work. I would describe them as a jumble of hope, worry, and stubborn craft. To me, it feels like a mix of people trying to build better things and people worrying that the rules and the culture are changing underneath them.

Hardware and the stubborn joy of reuse

There was a neat thread about hardware that kept coming back. On one end you’ve got very physical, big stuff — ships and rockets and trailers — and on the other end you’ve got the smaller but equally concrete: motors, charging hubs, and new steel mills.

Jack C. writes about China’s first autonomous booster recovery ship. The ship is huge — 144 by 50 meters — and it’s built to sit there and catch a part of a rocket so it can be used again. I’d say it feels like watching someone build a clever fishing net for expensive hardware. You can almost hear engineers swapping stories over tea about how many launches this will save. It’s not glamorous. It’s plumbing. But it matters. Reuse is the small, practical change that makes big things cheaper.

On the land side, Tom Moloughney reports on a mobile, off-grid EV charging hub deployed at a Zipcar facility in Massachusetts. This trailer can push out up to 320 kW. That’s not a sexy new battery. It’s a way to get power where the grid refuses to behave. To me, this reads like the food truck version of charging infrastructure: you bring the juice to the people rather than waiting for city permit paperwork to move at glacial pace. A useful hack.

Then there’s something more hidden but potentially game-changing: axial flux motors. Nacho Morató explains why these motors could be transformative for EVs — they’re compact, they pack more punch for their size, and they cool better. But, like many good-looking ideas, their future depends on making them cheap at scale. I’d say the story here is familiar: great tech, but the battleground is manufacturing. If you can’t make a lot of them cheaply, they stay boutique — the sports car of the motor world.

And close by, Ashlee Vance profiles Hertha Metals, a new American steel startup using hydrogen and gas instead of coal. That’s a big, clunky industrial problem being solved by chemistry and stubborn engineering. It’s like turning an old coal oven into a cleaner, faster one. If it works at scale, it’s quietly massive: better steel production without the carbon hangover.

A common note here is practical improvisation. These posts are less about glamour and more about how to actually make things work: catch boosters, charge cars, build motors, melt steel with fewer emissions. The theme is craft plus the conviction that the small, persistent fixes add up.

Cars, platforms, and the vertical push

Rivian gets the long look from Austin Lyons. The piece digs into vertical integration and what it means for product experience. To me, Rivian’s approach reads like a restaurant that controls the farm, the kitchen, and the front of house. You get menu cohesion and fewer excuses when something is off. Their zonal architecture and simplified computing strategy are clever ways to reduce complexity.

But there are the usual trade-offs. Vertical integration costs cash and makes you less flexible. Lyons points out Rivian’s cash burn and competition. I’d say the post asks: do you want neat, controlled products or do you want to bet on an ecosystem of parts that someone else will put together for you? It’s a choice, and both have costs.

These automotive posts also tie to the hardware pieces above. Axial flux motors, battery chargers, and platform design are all part of the same conversation about how to make EVs better not just in labs but on the road. If you think of innovation as cooking, some folks are inventing new recipes while others are fixing the oven.

Institutions, risk, and the slow death of mojo

There’s a darker chord in the week’s reading about institutions losing their edge. Naked Capitalism writes critically about DARPA, saying it used to be this wild, creative workshop and now it’s hamstrung by politics, conservative contractors, and perverse incentives. The argument is blunt: DARPA still makes prototypes, but the system that should scale them into actual capabilities is clogged.

That complaint crops up elsewhere, in different forms. Dr. Colin W.P. Lewis covers Gillian Hadfield’s work on AI governance, and the point there is institutional mismatch: laws and old rules don’t fit fast-moving tech. Hadfield says we need institutions that learn and adapt, not ones that just try to slap a rule on top. I would describe this as a plea for smarter plumbing in the public square. The tech moves fast; the institutions don’t.

And Ben Werdmuller worries about what it takes to build a culture — a Silicon Valley-style serendipity — and whether you can legislate it. He’s sceptical of copying infrastructure without copying the messy, social stuff that made Valley work. That’s a useful take: innovation needs both capital and social glue. You can buy servers and tools, but you can’t buy the bar where people argue badly and then build something brilliant afterward.

There’s overlap: DARPA’s problems are not purely technical. They’re cultural and political. Hadfield suggests the same for AI: governance needs to be adaptive and distributed. Werdmuller adds that culture and neighborly accidents matter. Together they paint a picture where systems, not just gadgets, make innovation happen.

AI, coding, and the anxiety about deskilling

AI is everywhere in these pieces, but not in the way the press usually says. It’s less about chatbots coloring pictures and more about what AI does to craft and judgement.

Anil Dash’s piece on vibe coding is a standout. He argues that LLMs help people write code and that this democratizes software creation, but it also creates new risks: devaluation of skilled jobs, shallow understanding of systems, and insecure or brittle code. I’d say he’s asking for a reality check: tools that make creation easier also tempt us into skipping the hard parts. That’s like giving someone a chainsaw and then not teaching them where it’s safe to cut.

That ties into an older fear in Naked Capitalism about the decline of deviant thinking: if we lean on AI for analysis, do we shrink our ability to question? They compare modern users to those who fought against Copernicus — the idea being that generative models can reinforce the comfortable story rather than upset it. The piece points to studies that show reliance on AI can dull critical thought. It’s a sharp and slightly gloomy note.

Chris Armstrong has a friendlier take on tool-making. He notes that developers build their own tools out of necessity and curiosity. That habit keeps people engaged and inventive. I’d describe them as the folks who don’t wait for the shop to open; they make the wrench when they need it. That culture of building your own is itself a kind of immunity against the worst effects of AI: you still shape the tools rather than just being shaped by them.

So there’s a push-pull. AI expands who can build. That’s good. But it also risks creating shallow builders. The week’s posts suggest the solution isn’t to ban the tools but to maintain norms that value deep understanding, tool-making, and critical thinking.

Makers, mavericks, and the myth of “self‑made”

Two historical pieces explore solitary genius versus communal support. Afiq writes about Alfred Lee Loomis, who used money from finance to build a private lab and collaborate with top physicists. It’s a romantic story: a private citizen playing with big ideas and nudging a whole field. I’d say it reads like a reminder that resources and connections matter — even for mavericks.

Along similar lines, Jamesin Seidel pushes back on “self-made.” He was at a Demo Day and saw how community — the introductions, the follow-up, the shared sweat — amplifies success. He argues, well, you’re not an island. Talent plus community is how you grow.

And then there’s nostalgia and institutional memory. John Buck retells Mike Liebhold’s Apple stories — the guy who pushed media research inside a company and even imagined an iPad decades early. That’s the other side of the story: sometimes innovation needs someone who says no at the right moment, someone who knows the tech and speaks the language of management so ideas don’t die in the wrong meeting.

I keep coming back to the same point in these pieces: individual brilliance is useful, but the social scaffolding converts brilliance into things that last.

The design shuffle: fashion, reality labs, and skepticism

Design shows up as an angle too. Victor Wynne notes that Alan Dye leaving Apple for Meta Reality Labs is the kind of move that makes headlines. A top designer heading into XR and AI-driven experiences sounds promising, but Wynne is cautious. He suggests that even brilliant designers struggle without a supportive ecosystem — a product becomes real when the whole house helps, not just the painter.

This ties back to the older theme: talent matters, but systems matter more. Design can be a spark, but manufacturing, distribution, and context determine whether the spark lights a kitchen or just a match.

Old tools, new uses — building your own

Chris Armstrong and some others circle back to the very human habit of making your own tools. It’s a quiet but potent strand. When developers build editors, scripts, tiny apps, they create islands of productivity that can later become bridges. It’s like keeping a pocketknife in your back pocket: you might not need it for everything, but when you do, you’re relieved it’s there.

The value here is cultural more than technical. It’s the habit of tweaking, of being slightly unsatisfied with default options. That attitude is what creates the “we’ll-just-make-it” solutions — a ship to catch boosters, a trailer to charge cars, an axial motor prototype.

Contradictions and repeated beats

There are a few repeated tensions across these posts. I can’t help but notice them.

• Prototypes vs. Production. Many posts praise prototypes but then warn we lack the pathway to turn experiments into durable products. DARPA’s bennies turned into dusty demos. Nice. Useful. But not always used.
• Decentralized creativity vs. centralized scale. Silicon Valley’s accidental social glue is contrasted with large firms and state-backed projects. Both create innovation, but in different ways. The Valley vibe is messy and social; the large projects are organized and resource-heavy.
• Tools that empower vs. tools that deskill. LLMs and other tools let more people create, but they can hollow out deep practice. The remedy suggested is stubborn craft: keep building your own.

I’d say these are the hums in the background. You hear the same note in aerospace, in car design, in code, and in governance.

A few small human stories

Some posts are less about systems and more about people. Ashlee Vance on Laureen Meroueh at Hertha Metals is one. Here’s a young founder, prodigy background, doing something industrial and stubborn. It’s refreshing to see a metal-smith story written with curiosity rather than condescension. It’s industrial romance — messy, hot, and loud.

Larry was a No — the Apple recollections — gives a personal view of how one person’s persistence nudged a company into new territory. It’s the small human sparks that often get retold like campfire stories.

These human notes are the glue. They remind you that innovation is not just policy briefs and spec sheets. It’s coffee and arguments and late-night soldering.

Governance and the need for adaptive rules

I want to pull the thread on Hadfield again because it matters. Dr. Colin W.P. Lewis shares her call for institution-aligned AI governance. She argues that preference alignment — trying to make models agree with user preferences — is not enough. Instead, we need institutions that are designed to provide distributed judgment and feedback loops. That’s a mouthful. What I’d say is: laws should be like good gardens, not walled estates. They should be tended and adjusted, not locked down.

It’s a subtle but important shift. The posts here suggest we should stop pretending a single law will fix fast-moving tech. Build flexible institutions. Teach them to learn. That’s easier said than done, of course, but worth saying.

A few small grumbles and a sideways look

There were bits that made me laugh or wince. The DARPA piece is almost elegiac — people fondly remembering the golden age and then sighing. The Copernicus-AI analogy in Naked Capitalism is dramatic: can a machine be deviant enough to suggest a heliocentric model? It’s a good question, but a bit theatrical. Still, the underlying anxiety is valid: will reliance on models dull our weird, contrarian streaks that drove big breakthroughs?

And the Meta hiring story had that familiar cynicism: hire top design talent and hope the rest follows. It’s like getting a Michelin-starred chef and asking them to cook in a food court. Talent helps, but the kitchen needs infrastructure.

Threads worth tugging further

If you’re in the mood to dive deeper, these are the threads I’d pull:

• How do you turn a clever prototype into a robust product? The DARPA and Rivian pieces both touch on this, from different angles.
• What’s the role of culture in innovation ecosystems? Ben Werdmuller’s point about serendipity is worth some thinking. Can you really build it? Or is it like trying to bottle a pub’s atmosphere?
• How do we balance AI democratization with preserving deep craft? Anil Dash and Chris Armstrong give you two sides of that tug-of-war.
• Who funds the patient years of industrial innovation — not apps but steel and motors? Hertha Metals and axial flux motors show the capital question up close.

I’d say these are the useful pull-strings. They’re the ones that make your thinking move from neat summaries to actual messy action.

There’s a lot more in each essay than I’ve hinted at. If a line here catches you — the booster ship, the axial motor, the “vibe coding” worry, the plea for adaptive institutions — go read the pieces. They’ve got the color and the receipts. This was, for me, a week of reminders: innovation isn’t one thing. It’s hardware and culture and messy human networks. It’s prototypes that need pathways and designers who need partners. It’s the risk that tools make us lazy and the hope that tools make us more people. Read them, poke at the ideas, and maybe build a thing or two — or at least argue about them in a pub or on a long walk, the old-fashioned way.