Technology: Weekly Summary (January 12-18, 2026)

The week in tech reads like a crowded market street. Lots of shouting. Lots of bargains and folks trying to sell the same thing three different ways. I would describe the tenor as two parts excitement, one part worry, and a stubborn thread of people trying to do things differently. Below I sketch the main themes I kept tripping over while reading. I’d say these are the bits worth poking into if you want to know what people are arguing about right now.

The AI story that won't shut up: models, partnerships, and the ad tax

There were three big splashes about AI this week. First, Apple quietly agreed to lean on Google's Gemini for a chunk of its new "Apple Intelligence" and Siri work. That was written up in multiple places, like Benjamin Mayo, Greg Morris and Lucio Bragagnolo. To me, it feels like watching a proud bakery admit it bought its croissants from the supermarket. People noticed. People care. Some called it an admission of defeat. I’d say it’s more of a practical truce. Big tech rarely dies of pride alone.

Second, Anthropic's Claude Cowork kept bouncing around the blogs. Early impressions from people like Simon Willison and coverage by Conrad Gray treated it like a new kind of office assistant. Folks like John Hwang argue it’s a dishwasher for startup playbooks—suddenly the old ways of building moat and hustle look shakier. I would describe Claude Cowork as a tool that strips away a lot of pipeline friction. To some, that’s liberating. To others, that’s terrifying.

And then there's ads in ChatGPT. That announcement lands like a wet sock in the camp that still treats these models as public servants. Brian Fagioli and others pointed out how the move changes the vibe. To me, it feels like your local pub deciding to hang a TV over the dartboard. The service is the same, but the mood shifts. People keep saying ads won’t influence outputs. Sure. But humans notice cues. Ads change trust, and trust is the currency these systems still live on.

These items are connected. Partnerships, new tools, and monetization moves tell the same story in three chapters: the era of building-for-show is ending, the era of buying-and-assembling is here, and the bill will arrive in a way people don’t like. If you want a close read, check Nick Heer and Mark McNeilly for good takes on what the Apple and ad moves mean in practice.

Agents, "vibe coding", and the future of getting things done

A noisy subtheme: coding is becoming orchestration. People wrote about agents, coding loops, and the idea that engineers will spend more time steering software bots than typing the lines themselves. There's a neat thread from Peteris Erins about "Ralphing" and running prompts in loops. Simon Willison and John Hwang both dug into Claude Cowork as a general agent and how it changes startup playbooks.

I would describe the shift as similar to going from cooking every meal to running a home where people cook. You still need someone who knows the pantry and the timings. The person changes from being a stove jockey to a head chef. It’s an odd mix of more reach and more fragility. If your prompts or your plumbing fail, nothing cooks.

This is tied to money and attention. Dave Friedman noted that as AI inference costs fall, overall spending will climb. Less expensive tools don’t save money so much as encourage people to use five of them instead of one—like buying half-price sausages and suddenly building a sausage tower. The same logic makes the agent era inevitable: cheap compute plus good UX equals more automation, not fewer bills.

Jobs: who gets replaced, who gets remixed

A recurring worry appears in several posts. People are watching layoffs and flawed automation. Will Lockett wrote a sharp piece about Salesforce and the human cost of overhyping AI. The headline was brutal: the company leaned too hard on AI as a replacement and then had to hire parts of the workforce back. It reads like a cautionary tale. It also echoes comments about translators and drone photographers losing gigs to cheaper AI versions in Thord D. Hedengren and wider coverage about blue-collar impacts in Mark McNeilly.

But there’s another take. Christopher J Feola argues AI is destroying old jobs but creating new ones. The new jobs are messier and often demand different skills. I’d say the net is unclear. Imagine a small town where the factory shuts but a dozen workshops pop up. Some people adapt fast. Some people don’t. The gap is where the politics and training programs must go.

There’s also an interesting human-versus-tool theme in "Developers are the new middlemen" by Stephan Schmidt. He wonders whether developers risk becoming obsolete if tooling gives direct access to capabilities formerly gated by developers. It’s the old middleman story—scribes, brokers, and now coders. I’d describe it as a mirror. We’re looking at the same social shape. Which side you come down on depends on whether you like being a craftsman or running the factory.

Privacy, tracking, and the slow march to local-first

Privacy keeps appearing in odd places. There was a thoughtful technical sketch about putting ads into end-to-end encrypted video while preserving privacy from Gojiberries. It’s the kind of engineering that makes you raise an eyebrow. On the one hand, the authors want to preserve revenue streams. On the other, they're trying very hard not to break encryption. I would describe this as trying to thread a needle while riding a bicycle.

Closer to street level was the adtech check on a cooking blog by Simon Rumble. He mapped all the servers and trackers and came away annoyed. Folks who love the internet often love it because it’s messy and free. But messiness turns into harvesting when it’s industrialized.

There’s a practical countertrend too. Vivaed laid out steps to build a surveillance-free block. That post reads like a survival guide. Turn off the Wi‑Fi, rely on mesh networks, form local agreements. It sounds a bit like a co-op vegetable garden, but digital. Lucio Bragagnolo and Duncan Geere both wrote about moving away from US cloud stacks and toward European alternatives for privacy reasons. To me, those posts feel like people choosing small independent shops over the mall.

This week’s privacy pieces also intersect with the Cloudflare acquisition of Human Native reported by Brian Fagioli. That’s part of a thread where companies are trying to make sense of how AI scrapes and uses content. Cloudflare’s move is pitched as giving creators leverage. It’s a slow battle for who owns what and who pays whom. Think of it like the farmers finally asking the supermarket to stop buying strawberries from the roadside stand without a receipt.

Math, reasoning, and the illusion of miracle progress

Several posts wrestled with the limits of AI’s intelligence. The cheers for systems solving parts of Paul Erdős’s problems—covered by Political Calculations—are real. Machines are starting to push into formal math. But most authors insist these are low-hanging fruit. Tom Renner had a provocative take calling LLMs a long confidence trick—harsh, but it got people talking about trust vs. spectacle.

There’s also a neat balance between celebration and skepticism in pieces like "AI Trends: 2025 Lookback and 2026 Outlook" from Michael Spencer. He cataloged growth and the impending hard questions—jobs, governance, and measurement. I’d say the math wins are interesting, but they don’t mean the rest of the house is in order. It’s like someone mastering the front door lock but leaving the windows open.

This dovetails with ethics writing. Susam Pal offered "Inverse Laws of Robotics"—a small manifesto reminding readers not to anthropomorphize these systems and not to trust them blindly. It’s the kind of common-sense you wish had a louder megaphone.

Governance, big-power tech, and the geopolitics of code

A couple pieces went large. Nick Heer wrote a sharp line about data centers acting like U.S. military bases. It’s an image that sticks. Tech infrastructure is not neutral. Another story: American entrepreneurs apparently scouting Greenland for loose rules and high-capacity fiber, covered in brief by WARREN ELLIS LTD. These are the kinds of things that look exotic but actually shape where innovation flows.

Then there’s the policy angle. Jason Steinhauer argued that democratic inaction on AI sometimes reflects process, not paralysis. It’s a useful reminder that fast fixes often break things. Reading that felt like watching a city council argue about whether to move a bus stop. Nobody is happy while it’s being debated, but the decisions matter.

Palantir’s ICE tool also showed up in Ben Werdmuller as a reminder that tech choices have real civil consequences. Surveillance and enforcement tech isn’t an abstract debate. It’s how people get profiled and stopped in their neighborhoods. That one made my stomach drop a bit.

Small things that add up: tools, formats, and the small hardware delights

Not everything is tectonic. Some posts were refreshingly small and practical. Michael J. Tsai wrote about Markdown’s continued life and how it keeps spreading into places like Docs and messaging. I would describe Markdown as the duct tape of the web—simple, weirdly sticky, and used for everything. There’s also a gentle revival of desktop Linux, with pieces from Kev Quirk and distro news from Brian Fagioli (MX Linux, Linux Mint). If you like tinkering, this felt like a cozy corner of the internet lighting up.

Hardware posts ranged from whimsical to useful. A surprisingly touching review of the Nüborn Baby by Josh Collinsworth was a product post that read like a parenting trial. It’s the sort of hands-on sinking-slow feeling you get when a baby monitor is glitchy. LEGO’s Smart Bricks, covered by Nacho Morató, brought back childlike wonder—smart toys that actually play without your phone. And then there was the sad little tale of the Fire TV Blaster dying on purpose, reported by Elias Saba. Companies remotely bricking devices? Feels a bit like a landlord changing the locks and not telling you.

Other small wins: Tailscale got a plug from AppAddict as a simple VPN that just works. Matthew Brunelle wrote about the Ploopy Touch open trackpad and the joys of open hardware. These are the pocket tools people actually use, not the ones on a keynote stage.

Creative, cultural, and tangential reads that kept showing up

The week wasn’t all tech roadmaps and ad debates. There were thoughtful tangents. Anil Dash had a warm piece on Wikipedia at 25. Vox Meditantis wrote about the difficulty of "un-inventing" technology. Duncan Geere shared notes from sonification workshops and a move away from US tech stacks. These posts are the kind that sit with you. They’re small mirrors held up to larger systems.

Also, nostalgia played a role. Posts about rotary phones and Easter eggs in Office 98 by Pierre Dandumont reminded readers that tech is layered. Newness sits atop oldness. I liked that. It keeps the internet human-sized.

Recurring patterns I kept seeing

• Push-and-pull between centralisation and local control. Big clouds get bigger. Small local stacks try to breathe. Writers on both sides argue passionately.

• Monetization creeps into everything. Ads in ChatGPT. Hardware getting bricked to force upgrades. Cloudflare buying a data-licensing company. Money shapes what features get built.

• Tools change roles before people do. Coding agents make devs into orchestrators. AI writes proofs and translates jobs. The social arrangements lag behind the capability curve.

• Privacy keeps reappearing in quieter, clever ways. Not always the loud policy fight. Sometimes it’s a how-to: use a VPN, run a mesh network, or switch to an EU provider.

• Skepticism is back. People who cheered for every model now ask what the real value is. That skepticism is useful. It keeps people honest.

Little threads I want to re-read

If you like the idea of a practical deep-dive, try the agent and tooling pieces from Simon Willison, Peteris Erins, and John Hwang. If you want policy and geopolitics, Nick Heer and Ben Werdmuller are sharp. For the moral and human side, Will Lockett and Vox Meditantis are worth the time.

I kept thinking of images that tie it all together. The week felt like a vintage train station. New lines added. One or two signals fail. Vendors selling strange gadgets. A conductor announcing platform changes. Some passengers change trains. Some sleep through it. You can moan about the announcements, or you can notice the map and pick a route.

If you're curious about particular posts I mentioned, the linked names will take you right to the authors. Read the whole posts if you like the smell of fresh ink. There’s more detail there than I can give in one wandering stroll. Some of these pieces will age into obvious truths. Some will look embarrassingly short-sighted. That’s the fun part. I’d say keep the ticket stub from this week’s station. You’ll want it later.