Technology: Weekly Summary (November 03-9, 2025)

I would describe this week in tech blogs as a messy kitchen where everyone insists their spice is the real reason the soup tastes good. To me, it feels like three or four big conversations are simmering at once — AI’s money problems and strategy, Apple trying to square privacy with big partnerships, and a continuing tug-of-war between convenience and control (your data, your apps, your defaults). There’s also a quieter undercurrent about old things — retro hardware, simple watches, and the pleasure of getting software to behave — that keeps showing up like a familiar neighbour with a thermos.

AI: money, agents, and the Balkanization

There was a lot of finger-pointing at OpenAI this week. The tone in several posts is sharp. Will Lockett calls recent moves desperate; thezvi.wordpress.com peels back the corporate decisions and the demand for a federal backstop; and Alex Wilhelm notes how firms keep chasing growth even when investors twitch. I’d say the thread running through these pieces is: the AI business still looks like a fast-expanding mall with many empty stores and huge electricity bills.

There are some concrete drama moments too. Charlie Guo writes about Amazon suing Perplexity over agent shopping bots. That one really sticks in the mind because it forces a practical question: when an AI agent acts like a person, who gets blamed or paid? To me, it feels like watching a courtroom drama where the main act hasn’t yet learned the script.

Then there’s the strange money dance: Snapchat is accepting a $400 million check from Perplexity to put AI into chats, according to Brian Fagioli. It’s easy to picture teenagers asking an AI for homework help, or for a pizza order, and the advertisers rubbing their hands. But another post asks: will users even care, or will it be wallpaper — something ignored but always there? The answer is probably both. Humans are fickle; we add, we ignore, we come back.

A recurring idea is fragmentation. Dave Friedman keeps returning to the ‘great AI Balkanization’ and to why Chinese labs spend less but sometimes get more out of constraints. The takeaway: there’s no single path to useful AI. Some labs blow cash on scale and ambition, others optimize for local needs. It’s like two chefs making the same soup — one buys every fancy gadget, the other sharpens the knife and wastes nothing.

On the ethics and social side, posts range from alarm to weary realism. Simon McGarr and others worry about extractive models and unemployment. Mark McNeilly and a few newsletters stitch ideas about AI in healthcare, advertising, and research. And a curious little study — rude prompts sometimes produce better answers — popped up in [Mark’s roundup], which is funny and slightly terrifying. Humans are still very much in the loop, and often are the weak link.

Apples and oranges (mostly Apple news)

Apple was the topic of many small, sharp posts. There’s rumor-level stuff, product-level complaints, and regulatory shimmies.

The rumor mill: a low-cost MacBook powered by an A-series chip got resurfaced by Jonny Evans. I’d say this rumor is like a neighbourhood whisper: plausible, tempting, and likely to make a few people start saving in secret. Relatedly, Corning is looking at microbatteries with Ensurge, which Jonny Evans covers — a tiny detail with potentially big consequences for wearables and edge AI.

Apple’s design choices and messaging drew criticism. Callum Booth complains about Liquid Glass — he’s not enjoying the new shiny. Michael J. Tsai went through macOS 26.1 and iOS 26.1 updates and spotted a raft of fixes, plus a Tinted mode to soften that Liquid Glass look. Reading those two pieces back-to-back, it’s like someone swapped the kitchen tiles and is now surprised the floor is slippery.

Privacy-meets-partnership shows up too. Michael J. Tsai and Nick Heer cover Apple’s rumoured use of Google’s Gemini for a better Siri, and the curious move of Apple building private cloud compute for Gemini. This is one of those compromises that smells good on paper — Apple pays Google a lot (cited around $1B) but claims on-device privacy. I’d describe them as a tug-of-war: money on one side, privacy brand claims on the other. It’s a bit like hiring your rival’s chef to make you dinner and promising you won’t taste the salt.

Elsewhere, Michael J. Tsai and Nick Heer squinted at Apple’s Live Translation roll-out in the EU and asked what “additional engineering” really meant for DMA compliance. The web App Store update that doesn’t let you buy apps? Nick Heer points out Apple made a pretty catalogue, not a store — which is roughly the digital equivalent of opening a bakery, but only letting people look through the window.

Tools, workflows and the small ergonomics that matter

There are a surprising number of practical posts this week. Little things, but they accumulate. Themes here: defaults, PDFs, file sync, and the endless tweaking of personal setups.

Two posts on app defaults and reading habits caught my eye: Luigi Mozzillo wrote twice this week — once about moving to Unread from Reeder Classic, and separately about narrowing a blog’s focus toward tech. The Unread post is the kind of thing that makes you nod if you’ve ever swapped a default app and then felt smug. The blog-evolution note is quieter: writing in English pushed topic choice, apparently, and that’s the sort of real-world friction authors rarely discuss.

PDF tools and mountain-of-files problems get attention from AppAddict. One post argues users are picky about PDF tools. Another champions Mountain Duck 5 (on sale), which maps remote storage into Finder. These are not glamorous, but they are the nuts, bolts and glue of day-to-day tech life. If you spend time wrestling with PDFs or slow remote collections, these posts are like the neighbour lending you a torque wrench — useful.

Defaults and app lists show up in Leon Mika and Denis Defreyne. Leon’s list of defaults is practical: Obsidian for notes, Google Keep for grocery lists, small music purchases from Qobuz and Bandcamp. Denis rambles with personal tech weeknotes, but you get the drift — people still care what fills that tiny slot on their dock.

On syncing music metadata and managing personal clouds, Rewiring walks through Hetzner plus Picard workflows. That’s a very specific playbook for folks who keep music neat and offline. The tone is: privacy plus control equals a tidy music collection.

And for those who like to tinker with Linux and old hardware, there’s good reading. Joelchrono and The Ubuntu Incident show that resurrecting old laptops or installing Rockbox still makes people happy in a way new gadgets don’t. One author gives away a Dell D410 after installing LMDE 6 so a friend can use it; that’s the sort of quiet generosity you don’t see on flashy product reviews.

Hardware, chips, and the cost of making silicon

Chips and fabs keep hogging headlines. There’s sober, nerdy analysis about lithography and capital allocation.

On technology claims and skepticism, Vikram Sekar dives into Substrate’s pitch for X-ray lithography (XRL) as an EUV alternative. The piece reads like a lab report and a detective story: the tech could be a game-changer, or it could be wishful thinking. The advice — be skeptical until you see measurements — is good old-fashioned engineering sense.

Layer that with Lawrence Lundy-Bryan on “Consensus Capital”: how the state once coordinated tech development through DARPA and NASA, then ceded ground to VC, and now seems to be creeping back in. This is macro-level plumbing for national strategy: when chips matter, the state re-enters the kitchen with a new recipe.

On the corporate side, there are posts about how American labs burn money while Chinese labs squeeze more out of less. Dave Friedman and others point out that capital intensity isn’t the whole story. It’s an almost Shakespearean theme: big spending doesn’t guarantee good theatre.

And then there’s the mundane but human side: Pierre Dandumont tests a PCI USB 3.0 card and reports real-world throughput on older PCs. That is the kind of post that helps you decide whether to buy or not, and is far more valuable to many readers than another benchmark graph.

Privacy, data centers, and the environmental angle

Two themes mingle: privacy (and the messy idea of personalization) and energy use in data-heavy services.

Privacy: James O'Claire writes about how personalized ChatGPT responses can feel creepy. I’d say his discomfort is the sharp end of a broader debate. People like tailored experiences, until the tech remembers the one stupid thing they told it three months ago. The Apple-Siri-Gemini coverage ties in here: Apple wants better AI but also wants to keep its privacy halo. Not an easy balance.

Energy and environment: Brian Fagioli highlights a report saying Netflix-style streaming beats ChatGPT on carbon footprint. That was one of those counterintuitive facts that don’t get air time. Streaming video uses a lot of power and, in aggregate, produces surprising emissions. So if you’ve been blaming AI for all the climate sins, maybe take a breath and check the carbon spreadsheet. It’s a reminder that the greening of tech is a network problem — the data centre, the grid mix, the whole stack.

On community trust, Andy Masley argues that data centres can bring real local value to communities, and that distrust of big institutions is often misplaced. That’s less flashy but important. It’s like reminding people that the new factory on the hill might also fix the local potholes.

Government, regulation, and geopolitical theatre

Several posts trace how power and policy chase tech. There are veterans of the policy wars, and there are worried analysts.

Lawrence Lundy-Bryan and Dave Friedman both explore the changing role of the state in funding tech. The message is: when tech matters to national security, markets alone don’t set the table. There’s also the story of Trump’s cyber czar, Joshua Steinman, covered by Ashlee Vance, which reads like part memoir, part policy brief. It shows how geopolitics and telecoms have bled into each other.

And on the legal front, the Perplexity-Amazon showdown matters. If courts decide AI agents can act like people (or can't), the whole shopping-as-service industry changes overnight. Lawyers will love the feast; the rest of us will watch prices and convenience shift.

The human side: education, attention, and small joys

There’s a steady set of posts about how tech shapes daily life. Teachers are warned not to outsource connection to apps (WHY EDIFY), bloggers talk about escaping phones and regaining focus (Jana), and one author berates the share button for wrecking attention (Angadh Nanjangud).

Some pieces go soft and wise: the Timex owner who loves a simple watch (Noisy Deadlines). Others are practical: a guide to keeping games on NTFS with Linux (Popcar). There’s a clear line: not all tech is about disruption. Much of it is about little comforts and saving time.

Education posts remind us of the obvious but easily forgotten: tech is a tool, not the goal. That’s a line you’ll see in the classroom posts and echoed in pieces about personal knowledge systems and the method of loci for memory (jerlendds). The PKM folks want our notes to behave more like a map and less like a pile of laundry. Cute idea, very practical if you actually use it.

Retrocomputing, hardware nostalgia, and the joy of fiddling

There was a gentle undercurrent of nostalgia. Ruben Schade and John Paul Wohlscheid take readers back to Compaq’s first portable and to Jack Tramiel’s Atari era. Pierre Dandumont found an old ROM that boots Mac OS on an Apple Network Server — which is the kind of find that makes museum curators whisper and hobbyists grin.

These posts remind us that tech stories have long tails. Old machines get new life, not just from emulation, but from people who tinker and write about it. It’s comforting, in a way: while the giants bicker over GPUs and cloud bills, someone is still soldering a joystick and smiling.

Odd little threads worth following

• An argument that computers are easier to hold accountable than humans — a contrarian riff from Andy Masley. It sounds cheeky, and it is, but it nudges at something real about audits, logs and blame.
• A walk-through of the Pebble 2 Duo software roadmap by Eric Migicovsky. For people who like small hardware with heart, this one’s a treat.
• A practical sale alert for a Google TV streamer and a note on app compatibility from Elias Saba. If you’ve been watching Fire TV antics, this is a timely dip into alternatives.
• A deep technical ramble about lithography and whether X-rays can unseat EUV by Vikram Sekar. If you like your reading with a side of physics and skepticism, that’s a late-night page-turner.

Patterns and disagreements I kept catching

1. Money vs. meaning. A bunch of authors poke at the same tension: who benefits when AI improves? Is it users, platforms, researchers, or investors? Posts disagree violently about whether the current spending is rational.

2. Privacy theater. Apple’s moves show a company juggling privacy claims, partnerships, and regulatory pressures. Many authors don’t trust the neat messaging. That distrust matters because it shapes how people accept new features.

3. Fragmentation is the new normal. From federated Mastodon updates (Simon Willison) to competing AI models and national strategies, most writers think we’re moving toward many partially interoperable systems — messy but resilient.

4. Practicality still wins. Despite all the high-level hand-wringing about AGI and geopolitics, posts that teach you how to move files, clean up music metadata, replace a browser, or get a watch that lasts seven years draw a lot of quiet appreciation. People want tech that works, not just slogans.

5. Nostalgia is not dead. Retrocomputing and careful hardware posts suggest a minority still finds delight in the physical and the persistent. That’s perhaps a cultural counterweight to the ephemeral cloud.

A few small surprises

• Streaming video’s carbon cost beating AI chatter. I didn’t expect that headline to land so many times. It’s a good reminder that the most visible technology isn’t always the most expensive in energy.
• The intensity of legal questions about AI agents. The Amazon vs Perplexity suit feels like a turning point that could decide whether agents stay curiosities or become regulated market participants.
• The sheer number of tiny, useful posts about changing defaults, switching music services, or learning MySQL. It’s oddly comforting. People still write to help each other optimize small corners of life.

If you want specifics, poke into the posts I’ve hinted at. There’s real meat in the long reads by Vikram Sekar on lithography, the policy lenses of Lawrence Lundy-Bryan, and the practical setups from AppAddict and Rewiring. The tech gossip and product rumors come from Jonny Evans and Michael J. Tsai — both tidy places if you like your Apple news with a side of doubt.

There’s a lot more in the links than I can fold into one afternoon’s reading. The pattern that settles in my head is this: big forces (money, state policy, and geopolitics) shape the contours, but the daily life of technology is still decided by the small, stubborn acts — choosing an app, changing a default, swapping a battery, or teaching a kid to look up from the screen. Like my aunt used to say at family dinners: the main things get shouted about and the small things keep the house standing. Read on if you like the shouty bits, and if you’d rather learn how to fix a stubborn PDF, there are helpers for that too.