A few months ago, around the time Googlebooks were announced, I became really interested in Chromebooks again, partly because I’d completely lost touch and partly because I heard good things about the new ARM chipsets, namely the MediaTek Kompanio Ultra. But Chromebooks are not a common thing in my neck of the woods.
And, even then, there aren’t a lot of manufacturers that ship interesting hardware in this field. I’ve been seeing ARM laptops slowly take over the lower tiers of the Windows laptop market, but I was very curious about the current value proposition of either Googlebooks or vanilla Chromebooks.
Disclaimer:Lenovo sent me this review unit free of charge, and this article follows my review policy.
For years now, the elephant in the room has been that both Googlebooks and Chromebooks are Google creatures first and foremost as far as user experience and software are concerned–and despite OEMs trying to figure out how to sell laptops that shipped palatable alternatives to Windows, it’s never been very clear if ChromeOS is a viable alternative outside of education.
I think that overshadows what the actual hardware can do, especially now that the ARM variants are becoming really good.
Another question is what the Chromebook experience is like from a software perspective–not technically, but in terms of actual usability and productivity in a world where the web has pretty much “won” (even if you don’t like that idea).
And yes, both flavours of “-books” share Google’s lack of consumer product focus and inability to coherently segment their offerings. Googlebooks are meant to be a premium tier when compared to today’s Chromebooks, yet both platforms run Android apps in a desktop context–which is actually one of the things I wanted to experience properly after years of various hacks.
The 14-inch OLED panel and full keyboard make for a comfortable writing machine.
The machine I was sent is the Lenovo Chrome 14M9610, which in typical Lenovo fashion is not a single model but a platform reference they unfold into multiple retail configurations. Having recently bought a brand-new Lenovo Yoga for my eldest’s college work, I can tell you it is very hard to pin down exact specs from PSREFs… But I digress.
The PSREF covers several regional combinations, but I got Lenovo’s UK 12GB/128GB configuration:
MediaTek Kompanio Ultra 910 (eight cores: one Cortex-X925 at 3.63GHz, three Cortex-X4 at 2.8GHz and four Cortex-A720 at 2.1GHz), with an Arm Immortalis-G925 MC11 GPU
MediaTek NPU 890, rated at up to 50 TOPS
12GB of soldered LPDDR5X-8533 RAM and 128GB of soldered UFS 3.1 storage (the upper tier has 16GB and 256GB UFS 4.0)
14-inch 1920x1200 60Hz OLED display (non-touch), rated at 400 nits and 100% DCI-P3
5-megapixel webcam (with the classic Lenovo privacy shutter)
60Wh battery (charging via USB-C, of course, and I got the 45W power adapter)
On unpacking the machine, I immediately noticed how light it feels compared to its size and metallic finish (I suppose Apple and Microsoft have conditioned me to expect meaty, weighty slabs of aluminium). Lenovo build quality for low- and mid-range machines has always been one of the reasons I favoured them (I have an IdeaPad Flex 5, which was a great investment), and the fit and finish do not disappoint.
The overall feeling is somewhat like a MacBook Air, although at 314.2 x 219.1 x 15.79mm this non-touch configuration is roughly one centimetre wider and deeper than the 2022 13-inch Air, even as it is a little lighter–one of the reasons for that is that the lid is aluminium but the bottom is ridged PC-ABS, which also gives it a good grip:
The ridged plastic base keeps the weight down and gives fingers something to grip.
One thing I liked immediately (especially considering my recent experience with the Surface Laptop) is that the small lip above the webcam makes the lid easy to open by feel alone (and yes, it passes the one-finger opening test). This serves the same purpose as the MacBook’s classic under-trackpad bevel, although it is a little annoying considering it sticks out just enough to cause friction when sliding the laptop into a bag.
Connectivity is basic but sensibly laid out:
A 5Gbps USB-C port on either side (both supporting charging and DisplayPort 1.4)
An additional 5Gbps USB-A port on the left (next to a charging LED) and a 3.5mm audio jack on the right
There is no HDMI socket or card reader, but being able to charge from either side is definitely useful, and the Chrome reminded me of how plain nice it is to have a charging LED. Seems like a feature Apple would value, right?
I got a machine with a UK keyboard layout, which has been amusing because I had forgotten how awkward it is to use some common symbols on it (which I mostly fixed by switching to a US international layout).
Accessibility and Dictation each get a dedicated key, the latter being uncomfortably close to the power button.
The keyboard itself is simultaneously classic Lenovo (slight bottom-edge rounding, tactile and bouncy) and classic Chromebook (with a Google logo “G” key and the Caps Lock key optionally doubling as an “action” key for emojis, search, etc.). It is also backlit, although at a single (dim) brightness level.
Sadly, this model (being the lowest in the range, as far as I can tell) does not have a fingerprint reader or any other biometrics, which meant some awkward moments when I kept reaching for the power button to log in.
The trackpad is a generously sized 120 x 75mm glass-like Mylar surface. It is smooth and accurate enough that I did not reach for a mouse during normal use, although the mechanical click is more conspicuous than the rest of the machine (I’m a tap-to-click person, so I never really pressed it).
Curiously, I initially thought pinch-to-zoom was not supported, but after a ChromeOS update it suddenly started working.
The same update also made three-finger overview and four-finger desktop switching more responsive, which is why I remember it distinctly.
The display is, in a word, gorgeous, and I didn’t expect it to be this good. It’s a very crisp, 14-inch 1920x1200 OLED panel, with the slightly taller 16:10 aspect ratio I prefer and thin bezels, at the top of which is mounted the 5 MP webcam:
A physical privacy shutter for the webcam, and a raised lip that makes the lid easier to open.
Lenovo rates the display at 400 nits and 100% of DCI-P3, and the combination of deep blacks and vivid colour is immediately obvious without looking at a specification sheet. The great contrast and legibility are something I’ve taken for granted on MacBooks, and it’s been one of the main reasons I kept picking up this machine for writing.
The only drawback I can see is that the lack of touch support makes it somewhat awkward to interact with Android applications, but I’m perfectly fine with that (the higher-specced models do seem to have touchscreens, though).
As to external monitors, Lenovo claims support for the built-in panel plus two external displays, but in practice I could only try one, and that would “only” go up to 3840x2160 on the Kuycon P20–which is pretty good for an ARM device (although my Surface can achieve more, this is likely enough for most people).
The speakers were one of the most enjoyable (and unexpected) parts of using the machine. There is a Dolby Atmos logo right under the left Ctrl key, and the speakers sounded unusually full for a Chromebook–music has enough warmth and stereo separation that it was a pleasure to listen to, and the bass was, if not MacBook-grade, at least serviceable for the eclectic mix of jazz and baroque that I favour these days while writing.
The speakers fire upwards through the grilles beside the keyboard, but there are two discreet bottom slots as well, which helps explain why they sound larger (and weightier) than the machine would lead you to believe. And since the Chrome is completely fanless (one of the nice perks of modern ARM machines), there was no fan noise to spoil the effect.
The Chrome worked seamlessly with my Wi-Fi 6 network, reporting nearly 1Gbps on the 5GHz band, confirmed both by my internal OpenSpeedTest instance and the OpenWrt stats:
OpenSpeedTest over the home LAN: 912.6Mbps down, 1022.6Mbps up and a 2ms ping.
Steam Link, which is particularly finicky with network bandwidth on Android, worked perfectly–and the game streaming experience was excellent, with very low latency and good video quality. In fact, I ended up playing Control and Hades 2 over Xbox Cloud Gaming a couple of evenings as well, which was a delight on this screen.
The Kompanio Ultra 910 is the interesting part of this machine, and was the reason I sought out this particular model. It is marketed as a very power-efficient eight-core ARM SoC with an Immortalis-G925 MC11 GPU and a 50 TOPS NPU. In ordinary use it feels very snappy even with multiple tabs, Android applications and several virtual desktops open. And given my hatred of fan noise, I was quite pleased to see that the whole package is passively cooled–there is no fan whatsoever, even if the underside gets a bit warm when running Linux.
The 60Wh battery turned out to be another benefit of pairing ChromeOS with ARM. Curiously, the diagnostics page initially showed that my unit had already accumulated 21 charge cycles and was at 97% battery health:
Battery health was still at 97% with the cycle count up to 33.
I thought that was due to it being in storage, but after a few weeks of my using it, the figures are consistent with this sample having been in use for a while. In practice, though, the battery has been excellent.
The first time I used the machine, straight out of the packaging and after minimal charging (reporting 100% full, but really just a top-up), the Lenovo Chrome lasted me easily two days of partial use (roughly 3-4h a day) before I started using developer features.
Over one weekend I charged it late on Sunday evening; over another I charged it on Monday–and both times it took under an hour to top up. And then I realised I could turn on battery optimisation (i.e., “low power mode”), which nearly doubled the machine’s runtime for writing and research.
This means that I was able to get nearly two full working days from this (3/4 of that if I used any Linux apps, and around a day of intense local development).
This is where Lenovo hands over to Google, and where things were a little stranger.
It’s been a while since I set up a Chromebook, but I noticed that you are now asked during setup whether you need Microsoft 365 support, which effectively enables OneDrive access.
However, I decided not to cross the streams and connect OneDrive directly. The consent screen grants the ChromeOS “Connect OneDrive” extension full access to read, create, update and delete my files, as well as ongoing access when I am not using it, and… I just don’t want that to happen:
Full read and write access to OneDrive, including when the app isn’t in use, was more than I wanted to grant.
It’s not clear how the connection is handled, who owns the keys or whether anything running in Google’s cloud can access OneDrive, so I skipped it.
This kind of integration is, sadly, something nobody seems able to get right–or make understandable enough for regular people to use, let alone people like me…
I was, however, able to use Windows App (the new name for the Android Remote Desktop app) to connect to my workplace resources just as I do on the Mac–including passkey support, via a QR code that I could scan with my iPhone. This, and the fact that Teams was able to use the onboard camera even when running from Azure Virtual Desktop, instantly promoted the Lenovo to the category of something I can reliably take with me on trips.
Even though I am not a big fan of the relatively plain look of ChromeOS window decorations (it feels like Openbox), I can’t argue against their utilitarian usability, or complain that there aren’t up-to-date creature comforts like multiple desktops and split window management:
Browser windows, Android apps and a Linux terminal share the same desktop overview.
In fact, I found the desktop experience (such as it is) very much in line with what I use daily on Windows, even down to the little layout helpers when you hover your mouse over the maximise button:
Equal and unequal splits are available alongside full-screen and floating layouts.
Gemini works OK, but I just had to do the pelican stunt…
Yes, Gemini is bundled in. Yes, I am entitled to a year’s worth of Pro usage if I put in my credit card information. No, I did not do anything with the offer–as you’d expect, given I roll my own AI assistant, I did not feel like committing to it. But I did use the built-in Gemini Notebook application (which is a web view into NotebookLM) for quite a few things, and I did try all the Chrome-baked assistance for searching, editing, etc., even though it could get pretty intrusive–I would get one of these every time I clicked in an input field, for instance:
Writing and image-generation prompts kept appearing where I just wanted to type.
Then there were the things I expected to work but just didn’t, like translations:
Help me read couldn’t handle the selected Chinese text.
But what I found most useful, given my penchant for testing accessibility features, were camera-based head tracking to move the mouse cursor (which worked very well indeed) and dictation support, which was instantly available everywhere by hitting Google+D or the dedicated key right next to the power button. That placement led to some finger fumbling, and I occasionally suspended the laptop by mistake…
Dictation worked consistently better than what I get on my iPad–it doesn’t have the new Rambler AI-powered post-dictation clean-up, but the speed of recognition itself and tolerance for pauses and changes in inflection were impressive.
From what little I tried in the various bundled apps, I’d say that if you’re committed to Google, you’ll find the Gemini experience useful, even if I ended up installing the Android versions of both Copilot and ChatGPT for my personal workflows.
One of the first things I did was to install Syncthing, Obsidian and Tailscale, as well as a few other staples:
Syncthing, Obsidian and Tailscale sit alongside Google’s defaults in the launcher.
Tailscale was immediately recognised as a valid VPN app and seamlessly integrated into the system (something I didn’t know Chromebooks could do):
The Android version of Tailscale integrates with ChromeOS’s own VPN controls.
However, after a while, I realised I needed to sideload apps to take full advantage of Android–ChromeOS Developer Mode requires you to completely wipe the device, but enabling ADB debugging is a separate option that doesn’t require Developer Mode. What I eventually did was to enable ADB debugging and use the Linux environment to install apps locally–that works perfectly, but still plasters a red warning under the login screen that reads This device may contain apps that haven't been verified by Google.
Things have improved tremendously since the days of my Acer C720, when I first installed crouton and eventually replaced ChromeOS entirely, dealing with assorted hardware tweaks along the way.
The built-in Terminal has its own profiles, themes and SSH settings.
Enabling Linux development support is now trivial, and it creates a Debian-based Crostini VM that is, for all intents and purposes, a “normal” Linux environment–the Terminal application is integrated with ChromeOS (you can mount any folder on the system into it) and has its own appearance, keyboard, mouse and SSH settings, while the VM exposes the usual command-line tools and package management, as well as an interesting assortment of virtualised devices:
An ARM64 Linux environment, with Virtio devices handling storage, networking and sound.
Sadly, that does not include Vulkan graphics acceleration. It is still perfectly usable, although I did not push it to the limit–my goal was to try the regular Chromebook experience rather than turn this into another Linux laptop, but all of the basics worked, including setting up and running VS Code:
The Linux version of VS Code runs locally alongside ChromeOS applications.
In fact, I ended up doing a fair amount of development on this, mostly to prove to myself that I could take it on a trip and have all the essentials:
VS Code for full-stack development (and a full copy of this site, including all the Go back-end and rendering code)
Enough of the Android SDK to build, deploy and debug an Android app completely locally, even on this relatively low-powered machine–something that is still impossible on an iPad.
I am currently developing gi (my new TUI harness) on it, because it’s more than fast enough for it–again, something that was most definitely not on my bingo card.
I should mention that despite some of the bundled apps being games, the MediaTek chipset does have one drawback: there is native graphics acceleration for Android games, but most emulators and off-label gaming hacks currently swear by Qualcomm’s Adreno GPUs, and most of the “alternative” GPU drivers you’d normally rely on for Vulkan graphics acceleration are not compatible. This meant that currently trendy hacks like GameNative and DroidDeck would simply not work, although to be honest that is a very niche thing indeed.
However, like I pointed out above, Steam Link and Xbox Cloud Gaming work perfectly fine, allowing you to play more demanding games via streaming rather than relying on the local GPU–and considering that you can probably do that for a long time with this combination of battery life and lightweight hardware, that’s a perfectly viable way to enjoy gaming on this device.
But that only makes me quite curious about how Googlebooks will fare in this regard, since there is a pretty big market for mobile gaming on Android devices right now, largely driven by emulation and the availability of decent performance GPUs in relatively affordable devices (plus FEX and Proton on ARM are evolving very quickly).
Android apps update automatically via the Play Store, but ChromeOS is also frequently updated (I’ve gone through at least three updates in roughly two months), and even though things like trackpad responsiveness were definitely a part of the experience, the last few release notes were almost exclusively about new AI features:
The release notes give Lens shopping searches more prominence than everyday usability fixes.
Surprisingly, I have very few, and they’re mostly related to the somewhat visible gaps in this combination of ChromeOS, Android and a desktop system. Some Android apps didn’t like being used in a desktop environment (not really news for anyone playing with Samsung DeX routinely). There are two Settings apps (one for the embedded Android, which will pop up when you are doing vaguely off-label things like developing on it), and there is a clear lack of unified visuals–Android was never the most consistent thing visually, but throw in a “normal” desktop Chrome and you have yet another UX flavour adding to the dissonance.
And yet, all the slightly weird basics I tried worked: Japanese input, mouse scrolling, dragging, etc. It’s arguably better than a GNOME desktop simply because I could seamlessly copy/paste text, rich text and even images across Chrome and Android.
File management is still a pain, though, so even though I could sync my Obsidian drafts using Syncthing and resort to VS Code, I had to struggle with the somewhat awkward Android-ish-but-not-quite filesystem arrangement.
But by far the biggest annoyance for me has been the somewhat arbitrary choice of things that the built-in Chrome browser refuses to download by default–SVG images, ZIP files from my LAN machines, the occasional (seemingly random) plaintext source file, etc. It seems to be fixable via security settings, and when I find out exactly how I’ll update this post.
Even considering it’s designed to a cost and unapologetically plastic, I think this is a great machine, and Lenovo has executed it very well. Although I prefer the 3:2 display ratio of the Surface Laptop I have been using for a couple of months now in parallel, I invariably reached for the Lenovo at the end of the day instead of my iPad because of the larger OLED display, full keyboard and overall lightness.
In fact, I think the best thing I can say about this laptop is that I literally had to force myself not to reach for it.
Although it is true that I am in a fairly unique position these days since the vast majority of my tooling only requires a browser or a terminal and (thanks to my preference for using an iPad) I moved all the specialist apps like Blender, OrcaSlicer, etc. to remote VMs over the years, this is a great writing machine–to the point where I ended up creating a web-based writing workshop to get around the need to sync my drafts to it.
Writing Workshop keeps the formatted draft and editorial suggestions side by side, entirely inside Chrome.
Yes, Writing Workshop was created because I wanted to have a proper WYSIWYG experience inside Chrome that was as close as possible to what the device was intended to do (and use AI to improve my writing, not generate it wholesale). I might have gone a bit overboard here, but I regret nothing.
As to committing to Google… No, I did not move my personal e-mail onto the device (well, I do have one Google account I regularly use, but it’s not my primary), and no, I could not quite commit to Gemini, but being productive on this machine was far less hassle than I expected: everything Android that I really needed worked. Even if the UX polish wasn’t there, the basics worked: I could sync my files to it, use Android equivalents of a few iOS apps, run VS Code, and even connect to corporate resources with… zero hassle really.
And it must be said that I was able to do some of those uniquely because Google, unlike Apple, lets me develop and install my own apps locally and gives me a perfectly usable Linux sandbox. What a novel concept, to be able to develop applications for your hardware… on your hardware.
As an Apple (and primarily iPad) user, the overall experience was… sobering. Yes, Apple has (temporarily, at least) gone “down-market” with the MacBook Neo to address the low-end laptop market, and I know that Googlebooks have a lot of challenges to overcome (with Google itself probably the biggest one), but I now find the prospect of them tantalising, as long as I am still able to install and run what I want.
I definitely want to try a Googlebook as soon as possible. There’s a there there for sure, and no matter how people feel about it (and Google), I think it has the potential to be a genuinely compelling device for my workflow.
Oct 3rd 2026 · 2 min read
·
#ai #dev #gherkin #life #tdd
This week was a bit different. For starters, it rained. And I visited the local office. And I consciously set aside hours to catch up on my writing instead of herding AI or doomscrolling, and again, failed at a bunch of that because–well, I lied.
Work has been… odd, discouraging and, above all, draining, so I’ll just get that out of the way: Every time there is a reorganization, a bunch of wheels are reinvented. That is fine; it’s part of the transition process. But the kind of wheels that people focus on during those transition periods is, I think, the most telling signal of whether that reorganization actually needed to happen.
That said, after a late-night stint last Friday pouring a bit of my soul into an internal project that might actually be fun but ultimately pushed the “buffer zone” between work and real life into the red, the week was finally over and I could sit down and focus.
I’ve been trying a new tactic for doing small projects, which is to have AI take my SPEC.md, generate a set of Gherkin files (which I then revise) and then develop the code from there–either as pure TDD or using the feature files as oracles for the actual tests.
And guess what? It’s mostly worked (I have three or four working examples, and I’m actually writing this on one of them), although retrofitting it to an existing project can be quite messy. As an extreme example, I took my Python MCP server for office files and my Go OOXML library, moved all the fixtures and test cases to a separate repo, and left a few agents to hash it out with the current test cases as grounding. They are still at it.
However, I see great potential in doing this for real TDD:
Humans take literally forever to agree on what a user story should be.
Gherkin gives me a compact, almost Python-like and formulaic take on the desired outcome(s) for a feature, action, etc. In short, it makes things a lot more deterministic and predictable.
I can match each feature to a part of the SPEC.md without losing the opportunity to refine it (and break it down if needed) in a machine-readable format.
And the LLMs do not need to reinterpret the features. They can help draft them, but once Gherkin scenarios are wired to tests, you can stop spending tokens reinterpreting them and just… run the tests.
Looking back, I am starting to realize that most of my hacks for herding AI are actually about removing it from the equation–my MCP designs bake in workflow guidance, my original development approach tried to set things on rails from the start, and now I’m just putting blinders on it.
As a way to sneak some minimalism back into my computing, I’ve gone “back” to TUIs and spent a little while working on gi, which is now more pleasurable to use and much more pi-like (same commands, same UX, mostly because I very much like its minimalism and you can’t go wrong with that).
Oct 2nd 2026 · 1 min read
·
#apple #macos #photography #raw
Redlamp was developed by one of the most talented Mac developers I’ve worked with in the past, and I’m very happy to see it–especially considering the way Apple completely neglected Aperture, and the shambles Photos still is if you have an inkling about real photography workflows.
It’s a native RAW editor for Apple Silicon, written in Swift and Metal, that opens folders directly and keeps edits in sidecar files. RAW development, masks and local adjustments, reusable Recipes and film looks are already there–although it’s still pre-alpha, with crop, healing and lens corrections on the roadmap.
Library management is further down the road too, so it isn’t a full Aperture replacement yet. I’ve added an app page with a screenshot and a few notes, and I look forward to using it from now on.
Oct 2nd 2026 · 1 min read
·
#apple #exchange #macos #mail #microsoft #outlook
The writing has been on the wall for EWS in Exchange Online for ages, and Apple’s promised Graph support still hasn’t shipped; M365 can temporarily keep EWS enabled until April 2027, but right now some tenants may effectively kick out Apple Mail/Calendar, which is just stupid because it is definitely not Microsoft’s fault that EWS is being phased out in favor of a better, much more flexible approach.
I have zero idea about why it’s taking Apple this long to get their act together, since EWS has been feature-frozen since 2018 and Graph has been meat and potatoes stuff for ages. Even though I do run Outlook locally for my (very rare now) consulting/advisory work, I am not looking forward to being locked out of my accounts just because Apple effectively stopped maintaining its cloud account integrations and has been phoning it in for the last 3 or 4 macOS releases–not to mention that, as I’ve found over the past couple of years while trying to automate stuff, it is effectively impossible to have a unified local API for contacts, calendars and to-dos that actually works, let alone e-mail…
So I guess it’s a moderately good thing I’ve been using Graph wrappers for years now to get around Apple’s inability to maintain things, no?
I’ve been watching the launch of Meta’s Muse and OpenAI’s Dots somewhat… bemusedly, if you will. Being a regular Verge reader/listener, I am partial to Nilay Patel’s take that Silicon Valley’s “software brain” mentality has obscured both a lot of the actually practical uses of AI, and the unvarnished truth that the vast majority of people simply don’t care—or that they do care about the privacy and productivity impacts.
Being too close to the coding agent madness (I am still poking at various agent harnesses, and feeling guilty about spending too much time yak shaving rather than actually using them for achieving more in other domains), I think that we’re at a weird plateau where:
We definitely have far too many interpretations of what an agent harness is
There is zero standardization on workflows (skills and MCP servers are still duking it out at both ends of the extensibility spectrum, but every user ends up creating their microcosm of weird little agent integrations)
I feel we’re coming to another plateau of decreasing returns, because even as models keep getting (arguably) smarter, what you can do with them beyond flashy demos hasn’t really gelled into useful outcomes yet.
Or maybe there are now just too many useful outcomes lost in the noise—it’s actually hard to tell. But if you thought we had far too many note taking apps (and workflows) in the world, this continued Cambrian explosion of agent-related things will seem very familiar.
However, unlike coding agents and all the hustle currently pervading the startup world, mainstream consumer agents target the way you go about managing your life—which may or may not involve note taking, but is quite likely a very personal, somewhat inconsistent and occasionally messy process–just like note taking.
And I think AI is not going to be much help with that, other than remove toil (yes, it will sort out your inbox for you and research your travel itinerary). But I keep seeing funky lifestyle demos, agents that shop online (why the heck would anyone trust an agent to buy clothes for them?) and other fairly contrived examples that I (and I suspect most people) would never delegate to another person, let alone an AI that would love to have access to my credit card number and tell advertisers I would love to know more about their products.
Somehow, I don’t think there is a good fit here. I think there’s a pretty decent chance that the idea of a Her-like assistant, however seductive, neither fits the way most people think about AI nor the sales and advertising business models that I can glimpse just underneath the surface.
I’ve come across most of these, but I’m still stuck on the papercuts I wrote about a few months ago. I have other priorities these days, and repeated exposure to Apple’s lapses in UX and QA judgement has somewhat dulled my senses.
iPadOS, meanwhile, still has weird bugs involving Spotlight and external keyboards, and those have been driving me nuts constantly.
Work was gruelling this week, with the rather unwelcome result that I didn’t exercise at all. I also fell asleep on the couch most afternoons/evenings just after closing shop–which ought to give you some idea of how much energy I had left for anything else.
I managed to, somewhat on autopilot, do a few updates on various projects (or yell at AI to do so, depending on the level of guidance required), but I’m late with a bunch of reviews and need to take some time to rest and do nothing even vaguely productive, so this is about as deep as this week’s update gets.
As an encore, the ageing RTX 3060 in borg has been repeatedly acting up at the worst possible time, and dropped off the bus for the third time this week just as I was drafting this. If anyone has a spare DGX Spark they want to get rid of (ha!), now would be a very good time.
I do have a very nice ARM64 workstation that I haven’t yet found time to add to my cluster, but I probably need to burn a heap of money on a Mac mini or a DGX clone as a Christmas present to myself, so I’m saving up. If you’d like to chip in, it would be much appreciated.
For now, I’d just like to get through the rest of the weekend without finding something else to fix…
Update: If you’re a fan of SuperGiant or their Hades games, this concert recording of their music is an excellent way to spend an hour or so. These people deserve every single bit of the success they achieved.
In Nilay Patel’s Decoder conversation with Mark Gurman about John Ternus and Apple’s next big thing, Gurman describes something camera-equipped AirPods might eventually do: look at an email on your computer screen and help add an event to your calendar. And that was when he lost me.
The email is already on a computer, and by all accounts of how computers have worked for the past decade, the software running on it ought to be able to add an event to a calendar. Introducing a camera seems like a roundabout way to get past the data silos I’ve been complaining about for years.
Not that it wasn’t a good Decoder episode (I’m a regular listener), but it was a very mixed bag and I have… feelings about it, especially because a lot of the discussion revolves around finding Apple’s next huge business. But I am more concerned about the now than vague hypotheses–I expect something different from the devices I already own.
And I also expect Apple to get its act together in many other ways.
Gurman thinks Ternus will need a new centrepiece for Apple, something that takes over from the iPhone. He traces a progression through the Mac, iCloud and now AI, with glasses and screenless devices as possible next steps. This is a very typical “next big thing” expectation and something that would be natural for Apple to pursue, but it feels off (and a quintessentially outsider take) to me considering their trajectory so far.
Cook built an extraordinary manufacturing and logistics organisation, and I’ve given him credit for that. Maintaining growth at that scale is an unenviable task, but I’d like to hear more about improving the products it already sells. Near the end of the episode, Nilay observes that he doesn’t really need a new Apple TV, and guess what, a device that does its job for years is a perfectly good outcome for me, too.
But I just don’t get the yearning for another centrepiece to Apple’s portfolio. As an example, I wanted something like the iPad in 2004, and was delighted when it arrived. My requirements were mostly about reading, writing and getting at my information without carrying around a laptop. A folding screen might help with some of that, sure, and the Duo might be an answer, but the way Apple has stuck to artificial software restrictions still puts me off.
And yet, I think there is a more fundamental issue that industry pundits keep harping on and circling around, which is that Apple really doesn’t know what to do with AI–it’s like some sort of fundamental impedance mismatch.
Patel and Gurman question Apple’s failure to build a competitive frontier model given how much effort it put into owning its silicon and modems. Later, Gurman argues that Apple could follow a successful OpenAI device quickly, because models are becoming commoditised and another supplier would work with Apple. But why should Apple? What is the incentive for them to do that, and how would they differentiate?
The huge gaping hole in Apple’s in-house model strategy isn’t new (and, to me, the biggest indicator of that impedance mismatch, since Gurman says he has seen no evidence of major AI acquisitions or huge pay packages to attract researchers), but I think Patel and Gurman are conflating two different things.
If a third-party model would let Apple compete with that hypothetical device, why is owning a frontier model essential to making a useful assistant? Relying on Google gives Apple a dependency to worry about, yes, but we’ve seen that before in search, maps, etc. That dependency doesn’t explain why Apple still can’t use the models properly, or why so much of the software Apple does own is still so difficult to use together.
Siri needs to carry out a request correctly, using the right information, without leaving me to spend more time checking the result than it would have taken to do it myself. And yes, it’s improved, but it’s still hampered by Apple’s single-user mindset.
Ironically, this does tie into the “home hub” concept, too. Bear with me.
When I wrote about the family assistant I wanted Apple to build, I was concerned about the immediate circle: shared calendars, reminders and keeping a household running. Smaller models can handle the intent parsing for those jobs–provided the application code checks which data the family member making the request may access and which actions the agent may take on their behalf. A more capable model doesn’t fix inaccessible task lists or the way iCloud treats a family as separate accounts sharing a payment method–and this is just my particular concern; there are dozens of integration surfaces that Apple just keeps ignoring.
The home hub Gurman describes would recognise who is standing in front of it and show them personalised content. I’d like to know how it handles shared calendars and reminders, including who can see or change them. Those are the things I already find frustrating about iCloud, and putting a screen on a robotic arm won’t fix them.
The gap that neither Nilay nor Gurman addressed is that Apple owns enough of that software to have an absurd advantage. It also controls the APIs that would let other people fill in the gaps, and not just for AI. I’ve been building personal agents on hardware vastly less capable than what’s in my iPad, but still need another machine for my container-based development tools. Apple doesn’t expose a hypervisor on iPadOS, so I can’t run those containers in a hardware-accelerated Linux VM on the device.
Yes, I know it’s a use case Apple doesn’t care about. But their hobbling of the iPad and refusal to address touch on the Mac are two core examples of things that they need to fix and that cascade into their entire approach to integration and software in general.
Then there’s the literal tone deafness about things like the Watch’s newfound always-on transcription features, and the fact that Apple has pretty much failed to read the room regarding how regular people will think about these new AI features.
Nilay asks at one point how people around these devices are supposed to feel about being recorded, and that’s the key issue for me. Apple’s ability to protect stored information doesn’t settle whether people wanted it captured in the first place, and I don’t want a future where my Watch (which is a medical device) will be banned because of this kind of idiocy.
The entire thing is just ill-conceived, and as bad as Meta glasses and their LEDs. A small icon visible to the wearer does very little for someone sitting across the table, and regular people won’t care that local processing and discarding the original audio can limit exposure. There’s an entire can of worms around whether the other person still needs to know that what they said is being transcribed, but my key point is that these things did not need to exist.
There is no timeline in which any of this is anything but product managers trying to see what sticks without understanding how regular people will think about the feature.
Gurman’s explanation of the proposed AirPods cameras felt weirdly inconsistent as well–they would apparently be able to read labels without producing ordinary photos or videos (and I completely get how that can work), but he acknowledges that he can’t explain how the system would limit what it recognises about people. Without that explanation and taking into account basic human concerns, I see an entire Flock-like controversy arising, and I just don’t understand how Apple is literally walking into that by purposefully including audio and visual features nobody asked for in their products.
Hands-free visual assistance could be useful, especially for accessibility, and Gurman argues that Apple should have shown more demos and explained its privacy approach more coherently, but I still think Apple just doesn’t get how to do any of these things–not technically, but from a purely acceptable perspective.
The bit that I was more interested in was when Gurman reported that Ternus and Eddy Cue are looking for more ways to make money from the App Store, with Schiller more wary of squeezing developers and attracting further regulatory scrutiny.
Phil has bucketloads of common sense. Him stepping away does not convey a good message.
I keep coming back to the same fundamental objection: I want to run my own software on hardware I paid for, without paying Apple to stop it expiring after a week. This has nothing to do with running a rival app store. My Swift editor project is something I’m building for myself, and I keep thinking that I would love to run it on my iPad, but can’t keep it there without either paying for the privilege or re-signing it every week. And the entire Apple toolchain assumes I want to distribute a product, with all the ceremony that entails, when I just want to use it on my iPad.
And that is, at a personal level, the thing where I wish Ternus would make allowances, especially now that Apple has stupidly powerful personal hardware–more capable hardware makes those restrictions harder to defend, and app-signing rules need to change.
I said earlier this month that I’d reserve judgement on Ternus for a year or so. Like Nilay and Gurman, I acknowledge that much of what he launches will have been in development under Cook, and being reportedly more willing to make decisions doesn’t tell us whether he’ll make better ones–but he can change priorities without waiting for a new hardware cycle, and the list of papercuts I put together in May has plenty of candidates, most of which are close to drinking age.
What I want isn’t rocket science–Mail needs to find and let me manage my mail (and it’s marginally better now). Calendar and Reminders need to work reliably across the people and tools that use them (and Calendar is still broken when it comes to accepting and handling meeting requests). I’d like automation interfaces that survive OS updates, and documented access to the information I’ve put into iCloud.
And letting me use more of my iPad’s ample processing power would be welcome. At the very least, I’d like to spend less of the next year writing workarounds for some of those things–assuming I can keep them on my devices for more than a week…
Sep 20th 2026 · 4 min read
·
#agents #ai #books #codex #gitea #hardware #notes #social media
It’s a bit insane that we’re past mid-September and my living room is still sitting at 29oC at 7AM, but such is life in the late Holocene, I guess. Regardless, the shift towards Autumn is starting to show, and I’ve had a couple of days where I suddenly realized I was working “late” into the evening and the lights started coming on automatically (never a good sign unless you’re having fun).
I have had to split my free time between a bit of stress at work and working on my health (neck, posture, back, exercise and other things). Yet, things keep happening.
Halfway through the week I decided (again) to do a couple of minor interventions and remove more social media apps from my machines–I have not been using Bluesky, Threads or Reddit at all for months, so those were easy decisions, but I am somewhat besmirched that I can’t really useMastodon (i.e., Ivory) for anything meaningful (and yet am paying a subscription for Ivory because it is the best iOS experience) and that Twitter (yes, I will keep referring to it as Twitter), sadly, remains the “best” window into the tech industry’s collective unconscious today, even as it continues to be a cesspool with no signs of improving.
Even forcing the “Following” tab with chronological order and using a web script to filter out ads is a compromise (ethical, political and mental health-wise), and yet I cannot look away.
Not for the dopamine hits, but for the edgier, more interesting hustles going on 24/7.
There is actual progress everywhere, but there is also a lot of tenuous hype, Silicon Valley bro culture, and subdued (yet sometimes misguided) pushback from the rest of the world.
It’s neither wholesome nor ultimately productive, so I’ve capped Twitter at 15m daily–the same as Hacker News, which still carries a semblance of authority even though the hustle is just as real there.
But I digress, again.
Consider this my usual yearly rant, which is steeped in tradition.
My Codex Open Source account is active again, so during the week I finished a bunch of low-impact polish work I had been meaning to do for ages now:
piclaw now has budget controls, marginally more consistent Settings UI across its two skins, model/session pickers, VNC panes, etc., as well as some timeline SVG fixes.
I took the time to do some “reverse TDD” and generate a shared Gherkin behaviour specification for the piclaw web UI, and then applied it in a consistent UX refactor for both vibes and tau-prime–I kept trying to use their bespoke interfaces and getting annoyed at the differences in agent steering, timeline attachments, session controls, etc., and at least now most of it is broadly consistent.
memento now has… Trash, somewhat saner bulk memory operations, and doesn’t “lose” older revisions when you’re looking at it as a reviewer; old proposals were being invalidated upon new revisions of a node, even if those new revisions didn’t address the proposals, and useful stuff was getting “lost”.
My piclaw agents can now talk across completely different networks using iroh (in fact, that is how I am porting memento to Go now, with two different agents coordinating between Intel and ARM hardware for testing).
Since I am still deep into agentic stuff at work, I decided to add an A2A connector to piclaw-addons. Its main goal is to provide me with all the enterprisey stuff I need for testing other agents: authenticated client/server, durable task mapping, pinned agent-card trust and authenticated push notifications.
There will be a few more additional features for these, but I think we’re hitting diminishing returns here, at least until I need to solve new problems…
Yes, I’ve been looking at TypeSafe’s Jev too, and trying to figure out how to emulate it with various techniques in go-pherence, because I really like the idea of smaller, focused models doing bounded jobs–I’ve never liked the fuzziness of LLM-generated JSON (there was a time when I was hacking on guidance to force schema compliance), and I think the principle is broadly right: most business AI I come across could be solved with a set of steps using BERT/Bayesian-like classifiers–but those classifiers would only work if we could “train” them with general knowledge, and that is what Jev “solves” to a large degree.
I’m poking at every open weights “implementation” out there right now, because I have dozens of scenarios where I can use something like Jev, but local.
For instance, it would have been great to have it back when I was running my feed summarizer instead of hacking my own classifiers, and Shelf would also benefit from a general purpose classifier for a bunch of things.
I got a new amazing KVM this week, which I am having too much fun with and will write about soon. Right now I am trying to clear out my review backlog, so the reason I mention it is that I can finally go back to setting up and testing more machines and clear some desk space, which is delightful since it has been a losing battle for months.
Now I don’t have to bother with anything but providing power for some of them, which makes things a lot simpler.
I’ve been keeping local replicas of my GitHub projects since… forever, but I’ve been neglecting the mechanics of it a bit, so this Friday I asked piclaw to:
Consolidate all my mirrors under a GitHub organization on my local Gitea instance, adding the missing ones from my recent projects.
Audit the inherited polling schedules against repository activity. Old projects now get quarterly checks or less, and we stopped polling things that got removed from GitHub (like emulators and other things I thought were worth preserving).
My old trick for managing polling overlaps is to set distinct intervals using a prime (or product of primes) number of seconds, but I told the model to add one-time jitter to clustered next runs, and this not only cut down Gitea polling from around 90 checks a day to around 20, but also made the overall distribution much nicer:
Longer polling intervals reduce the number of checks; one-time jitter spreads out the next runs.
This took, oh… 30 minutes, tops.
And I didn’t need to keep an eye on it, so likely less in practice.
My Writing Workshop thing is progressing well and has been quite helpful already in putting this draft together, so I am very happy with it so far–in less than an evening, I have a very helpful assistant (perhaps a bit too helpful and insistent, mind…) and the overall experience has been quite positive:
Writing Workshop showing the weekly draft with highlighted passages and a review findings panel
Of course there is a lot of polishing to do (my original intent was to add polish, not too much friction, and right now the 22 kinds of suggestions from the assistant are a bit overwhelming), but it’s a matter of tuning out what I don’t need and focusing on the improvements that truly impact the quality of the draft.
And then, who knows, maybe this will evolve into something even more substantial.
But what I’m most happy with is that I finished both Norse Mythology and Last and First Men this week, which is a positive data point in my return to reading.
I did get piclaw to go into GoodReads and mark them both as read for me from my Nomad, so it wasn’t a completely AI-free experience, but I am pretty sure my priorities are OK.
Still very much in line with the overall return to work mood of the past few weeks, I thought I’d post some notes about work gear.
You see, I have been using a “Surface Laptop for Business 13in 1st Ed with Snapdragon” (I kid you not, that is what it is called) for a couple of months now.
And it might be the best Windows laptop I’ve ever used–at least in summer.
Doesn’t look like a MacBook, or… does it?
Disclaimer: This is a work machine–a loaner, alas, not a permanent replacement–but I’m writing it up under my review policy all the same.
Although I don’t have a MacBook Neo on hand, I’d like to get that comparison out of the way because the general look is quite similar–except for the taller display (which I quite like, by the way) and the Windows logo.
Here’s my MacBook Pro vs Surface Laptop instead.
I looked up the Neo’s specifications and the Surface is a little narrower, slightly deeper front to back and a bit thicker, but they weigh practically the same.
And yes, this looks and feels a lot like a MacBook in general–it has the metal case, large trackpad and even the rounded display corners down pat, which helps establish an overall feeling of polish.
Even though this is a 16GB machine, it is actually one of the entry-level Surface configurations as far as CPU (and size) are concerned–I am using the 13-inch 1st Edition, with the eight-core Snapdragon X Plus. You can check the online specs for the business model, but this is pretty much it:
The basics, straight from Windows (IDs redacted).
8-core Snapdragon X Plus X1P-42-100 (clocked at 3.30GHz)
16GB LPDDR5x RAM (15.6GB usable)
512GB SSD storage
Qualcomm Adreno GPU and Hexagon NPU, rated at 45 TOPS
As you’ve probably gathered by now, I like tall displays, and I am quite fond of the Surface’s, whose 3:2 aspect ratio accounts for that slightly taller feel I mentioned earlier.
The panel has a density of 178ppi (decent, but not Retina), a refresh rate of up to 60Hz and a rated brightness of 400 nits.
The fact that it is a touchscreen is something I happily ignored for a long time with no ill effects, but it is quite useful, and colour reproduction seems quite good–I’ve found blacks to be nice and deep.
The keyboard is a bit hard to pin down–I quite like it, and it feels softer than a Mac’s, but with good travel and a slightly stronger “thock” when typing at speed.
It has a Copilot key (which I remapped to Search inside the first hour of using it), and the power button doubles as a Windows Hello-compatible fingerprint reader, so my hard-wired Touch ID reflex was quickly sated:
Power button with integrated fingerprint reader.
The backlight has three discrete levels, and if you’re used to a MacBook keyboard, you’ll need some adjustment regarding the Fn key (which is to the right of the left Ctrl).
As to the trackpad, it’s large, roomy and responsive, but the OS doesn’t drive it the way macOS does–there’s none of the gradual, weighted motion I’m used to, so gesture feedback feels noticeably different at first.
But the responsiveness is there, just not for scrolling (and yes, one of the first things I did was to enable “natural” scrolling), and Windows swipe gestures work nicely.
One relevant note is that since I’m a “tap-to-click” person, it took me a while to realise that clicking only really “worked” halfway down the trackpad–typically for gestures like clicking with your thumb and dragging windows with a finger.
The Surface has one USB-A port and a headphone jack on the left and two USB-C on the right, with both USB-C ports specified for charging, data and DisplayPort 1.4a, with support for up to two 4K displays at 60Hz.
Surface Laptop ports on either side.
They read out as USB 3.2 ports rather than USB4 or Thunderbolt.
On occasion, only the port farthest from the screen would drive an external display, and the documentation doesn’t explain the difference I saw between them.
That didn’t stop it from being very enjoyable to use for work–I was able to drive a 4500x3000 panel at 60Hz (and a Logitech Brio 4K hanging off its USB hub) with zero issues, and the experience was excellent throughout.
The “Omnisonic” (again, that’s actually what they are called) speakers are invisible–no grille or opening that I can find, and yet they sound surprisingly good, which justifies the Dolby Audio label.
The two “Studio” mics discreetly embedded in the bezel alongside the 1080p front-facing camera also worked fine, and over the past couple of months I had zero complaints during Teams calls.
Two of the reasons why I love this machine are that it is dead quiet (I’ve never heard a fan) and, most importantly, it stayed almost perfectly cool in normal use throughout late July and August.
I did start out by feeling some warmth under load in the first few hours (yes, Windows Update, thank you), but over longer use it has been confined to the underside.
The bottom case never seemed to go past about 42°C under sustained load, and the heat spread pretty uniformly–at least based on my spot IR thermometer.
One of the stranger bits of friction has been Wi-Fi roaming.
The machine uses a Qualcomm FastConnect 7800 adapter, and with my Wi-Fi 6 setup it kept dropping off the network often enough that I did an AP-side audit rather than blame Windows on vibes alone.
Qualcomm FastConnect 7800 driver settings.
The living room AP looked healthy, with continuous telemetry, no LAN or backhaul errors and good 5GHz associations.
But the Surface was the most frequently disconnecting client in the OpenWrt hostapd logs, repeatedly dropping off and reassociating across the access points.
That made client-side roaming or band steering worth investigating, especially with usteer in the mix, but the logs alone were inconclusive, and without Linux support doing any low-level troubleshooting on the Surface was out of the question.
I lowered the driver’s Roaming aggressiveness setting to Medium-low, and that seems to have fixed it.
Preferring 5GHz and disabling adapter power saving would have been my next steps, but it wasn’t necessary.
I am considering delving into WLAN AutoConfig logs further to help distinguish Windows/driver decisions from AP-side steering, but if it works, I’m not going to “fix” it further.
Spoiler: I had zero issues.
I did not think twice about this being an ARM device, nor did I do any weird compatibility tests–in fact, up until I started drafting this I hadn’t even installed WSL (it works perfectly for what I tried, and you get an ARM Linux userland, just as you would expect).
This is because I have been using it almost exclusively for non-developer work (Office, Teams calls, a lot of Remote Desktop–which is where I actually develop), and the occasional CAD session using Shapr3D, which supports Windows on ARM and ran flawlessly.
I also tested exactly zero games running natively on it (and would not use a laptop for gaming, anyway, unless I was streaming to it).
Rather than bore you with random performance figures, I will just say that I never felt it slow down in any relevant way.
And that is considering the massive use I make of an eclectic mix of Office web and native apps with dozens of simultaneously open documents in either flavour–I don’t keep around hundreds of background Edge tabs, but I do have pinned tabs with key documents for each project and use them, which is arguably more demanding.
Since most of my work is actually server-side AI, I barely noticed (or used) any of the AI features that ship with the machine; one I did use was simply swiping in from the edge of the screen and lassoing text to summarise, which was quick and easy enough to draft Teams messages from project documents. I didn’t dig into the mechanics, but it seems to be using a local model to do the summarisation (and OCR).
This class of Snapdragon NPU isn’t particularly useful for LLM work, but the camera effects got more use. Windows Studio Effects does automatic framing (which I found very useful), eye contact correction and portrait lighting (which I honestly didn’t find useful), with no noticeable performance overhead. Again, as a Teams machine, this is pretty much perfect.
The blurbs I’ve seen about the 50Wh battery say it’s nominally rated for up to 23 hours of local video playback or 16 hours of active web use, which feels about right.
Those tests typically use fixed workloads, so neither figure is really representative of how long it lasts with Teams running, but I can tell you that on a quiet week, I was able to use it for around three days with continuous stints of around 4 hours (I need to keep moving around on account of my back and posture, so I roam about the house a bit).
There were a couple of low points (typically associated with meeting-dense days), but I was assured a full day’s worth of battery life on occasional office trips. I didn’t measure charge times, but a 65W charger had it ready to use again over lunch (a Portuguese lunch, mind you).
I love this thing’s size (it is smaller and lighter than my MacBook Pro), simplicity and utter lack of fussiness in design language (even the feet are… cutely minimal), and the only thing I’d change about it would be making it easier to open (there’s no bevel under the trackpad, just a very slightly protruding, sharpish front edge).
Surface Laptop front edge and underside rubber foot.
Windows on ARM might still be a drawback for some people, though.
We actually skipped buying one of these a couple of weeks ago (and went for a Lenovo with a brand new low-power Intel Arrow Lake/Ultra 7 chip) because one of my kids needs to run SolidWorks for college.
But it was a close call, and there were other factors (like an OLED screen and being able to run Linux, not to mention being able to swap storage later).
But if, like me, most of your work revolves around Office and happens on remote machines, this is a very sweet laptop, and I am actually trying to replace my main work machine with one of these…
As my AI policy points out, I do revise my posts with AI, but until now, I’ve resorted to post-draft LLM passes that take a bunch of SKILL.md files and either fix outright typos and misspellings or add editorial blockquotes to my drafts. The entire process feels needlessly technical in the sense that doing it inside vim or VS Code (typically as I bring stuff together for a final draft) detracts–or, rather, distracts me–from the writing experience too much, and sometimes automated replacements slip through.
This is kind of huge, considering that I’ve always wanted an actual M8 (and built a headless one into a TrackerKB with its own keypad), but the hardware is extremely expensive. Having an iPad version that works as intuitively as you’d expect with an iPad keyboard is a very welcome surprise, and I’ll be trying out Bluetooth gamepads as well ASAP.
It feels good to have something like this pop out of nowhere and grant you some respite from depressing industry hype and even more depressing work stuff… Especially since it sounds amazing on the iPad Pro speakers.
Given how much time I’ve been spending on small native apps and my swift-app-template, Swift Build becoming the default in SwiftPM is probably the bit of this release I will adopt soonest. Having the same build system on macOS, Linux and Windows is welcome, as is Subprocess finally reaching 1.0–I still have plenty of uses for Swift outside SwiftUI. Being able to await inside defer and shield cleanup from task cancellation also sounds useful for the sort of audio and background work I’ve been doing in swift-smart-prompter, but to be honest that’s the kind of thing that I would prefer to do in other languages without weird idioms.
The WebAssembly and Embedded Swift improvements are tempting, though. The WASM SDK is now available directly from Swift.org, JavaScriptKit’s safe bridging is apparently up to 40 times faster than the older dynamic approach, and microcontrollers get more flexible types and error handling. None of this fixes my complaints about SwiftUI, but I would quite like to use more Swift without having to drag Apple’s frameworks along with it…
Following up on my testing of the LattePanda Mu and IOTA, this time I’m looking at the Sigma, which (as is becoming the norm with my recent pieces) I actually got before summer break.
Even though my personal calendar is now filled with a smattering of fresh events, I am quite happy about the ones that don’t involve health, legal and tax annoyances. My back and other ailments seem to be stabilizing thanks to daily exercise and my having dedicated some time to cooking my own meals, which, besides being cheaper (a concern I’ve been having of late), also forces me to stand and move about, not to mention giving me an excuse for handling very sharp knives in very satisfactory ways.
Besides the showmanship demonstrated by the keynote opening (nice touches there, Tim) and the visual effect when you open it, the iPhone Duo is… strangely familiar, and yet, somehow just strange as well:
Only yesterday I was discussing Mistral off-work and pondering what they might be up to, given both their EU lobbying – which fits rather neatly with the European AI pitch I wrote about in April – and the fact that they haven’t been doing a lot of actual model releases (in practice, nothing really new, or at the same pace as other labs).
I guess that they are positioning themselves squarely in the sovereignty arena, where a few other European companies have already started providing generic open-weight models for privacy- and US-sceptic customers that don’t want to be dependent on the whims of non-EU administrations. It’s a decent market to be in in Europe, but feels too regional, and I hope they get back to improving their own models, which are substantially lagging behind by any benchmark (if you care about benchmarks).
My LG TVs are too old/dumb to fall into this lot, fortunately, but what I want for my next TV is a huge, dumb monitor, and reports like this are why. Gamers Nexus and Level1Techs apparently found LG TVs recording audio with the screen off and storing it for later upload, besides snooping on other devices on the local network. I don’t know how much of this applies to European models (the article doesn’t establish that), but I would very much like someone to check rather than assume GDPR has taken care of it.
I’ve been complaining about TVs spying on their owners since 2012, and about the lack of EU scrutiny in 2024 and 2025, so this is getting a little tiresome. The European Commission keeps harping on Apple, and I suspect that has rather a lot to do with it being a much more recognisable brand–going after Apple has immediate political impact, whereas investigating what LG and the other TV manufacturers collect inside people’s homes seems to be nobody’s priority.
The desire to own an Apple Studio Display has probably been hanging over most Mac desktop users since time immemorial (well, since 1998 at least, but most people are more familiar with the “modern” 2022-era look) for two reasons:
As many people have remarked, the moat between ideas and technical execution is narrower than ever, and that has a number of consequences–both when it comes to investment (in this case, personal, although I am collecting war stories…) and focus.
And I think I need to be more selective about what gets my evenings, since I am back at work and already spending too much of my free time building things on a computer.
I have a long and somewhat embarrassing history with LISP that goes back to the years when it was fading from academia and LISP machines were giving way to Macs and DECstations (a transition that should be familiar to anyone who read the UNIX-HATERS Handbook), so I got properly hooked on Clojure when it came out.
This is indeed the end of an era. As I wrote when the transition was announced, Cook built Apple into what is probably the most formidable product manufacturing and logistics organisation on the planet (I remember when Nokia was considered the same, albeit only for mobile phones), and Patrick McGee’s Apple in China makes the sheer scale–and strategic cost–of that achievement painfully clear.
But Cook was never a product person, in my view, and that showed in the decline of overall software quality over the years. Contrast that with the tremendous growth in services and the sheer volume of the App Store–run by Phil Schiller, who is also changing roles today–and the lasting impression is one of relentless efficiency, perhaps a tad lacking in soul.
Everyone has high hopes for Ternus as his successor, particularly around QA and actual product focus. Cook will stick around as executive chairman to deal with politics (his understated main focus these days), but I am going to reserve judgement for a year or so…
Aug 30th 2026 · 3 min read
·
#eink #productivity #readers #supernote #xteink
I’m now back to work and already embroiled in far too much, so I’m going through my infinite set of mental checklists and trying to relax by offloading some of the stuff I never got around to writing down.
While I was on vacation, one of the things I did was tune out (as much as possible), which meant relying more on my e-ink devices.
I came back to work this week, started catching up on everything, and decided to go out for groceries without an umbrella (it is still August, right?) and got drenched.
This is extremely impressive, not just from a design and kinematics perspective, but because the entire 50Hz control loop runs locally on an RK3566 (not the RK3588 I first assumed), with the ONNX movement policies running on the robot itself.
The repository includes the MuJoCo/PPO training code and Rust runtime, so every shipped behaviour can be retrained–which is quite something for a 25cm, 800g biped with fifteen motors that can walk, skate, grab things and get back up by itself.
It is a completely superfluous thing to spend north of $500 on, and I have no idea what I would do with one or where I would find the time, but it presses all the “shut up and take my money” buttons–and if NVIDIA’s acquisition of Hugging Face goes through, I suppose we may have to call it the Jetson Donald or something…
Aug 28th 2026 · 1 min read
·
#ai #hardware #ram #semiconductors #tariffs #trump
Trump’s timing is impeccable, as usual: after the AI boom helped turn RAMageddon into a sustained memory price surge, he has apparently decided that what data centres, PC builders and everyone buying electronics need next is a tax on the chips themselves.
And this lands just as the AI industry’s “teaser period” is supposed to end–when vast take-or-pay compute commitments become actual bills and utilisation starts to matter. Nothing says “winning the AI race” quite like making the entire stack more expensive just before the economics get real…
Aug 27th 2026 · 1 min read
·
#acquisitions #ai #hugging face #nvidia #open source
Well, this is unexpected. There were rumours while I was away–and Hugging Face reportedly turned down a $500 million NVIDIA investment earlier this year–but buying the entire thing for $12.9 billion doesn’t fit my mental model of NVIDIA, given its… spotty Open Source report card. They certainly have the money, given their current valuation and a quarterly revenue run rate closing in on $100 billion.
This feels like the same sort of stack consolidation that happened as public cloud took off and GitHub was acquired: owning the place where models live is a lot more ecosystem presence than pushing Nemotron models out in a corner and hoping people notice. I guess it makes sense. I am just not sure Hugging Face under NVIDIA is quite the same proposition…
Aug 25th 2026 · 1 min read
·
#ai #apple #apple silicon #mac #mac mini
The big difference Apple isn’t really drawing attention to is memory bandwidth. The M6 tops out at 170GB/s and 32GB of unified memory, whereas the M5 Pro offers 307GB/s and up to 64GB–nearly twice the bandwidth, as well as twice the memory ceiling. That is the number to watch for AI inference, not just CPU and GPU benchmark deltas.
The less amusing part is pricing: a halfway decent M6 configuration with 32GB RAM and 1TB storage lands at roughly €2,000, and an M5 Pro with 64GB RAM and 2TB storage costs about another €2,000 on top. And the Mac Studio can be yours for… a kidney, or two.
Aug 25th 2026 · 1 min read
·
#ai #legal #llm #open weights #privacy
The bit I like most about Thomson Reuters’ new model is that it starts from open weights–specifically Qwen3.6-35B-A3B, via Snowdon–and then uses continual training on Thomson Reuters’ own legal, tax and news material. There is likely an opportunity here for AI labs to provide tailored industry models with much better provenance, rather than aiming for AGI.
I am curious about how hyperscalers will tackle this–fine-tuning and training have always been available in Azure, for instance–but what I would really like to see is a definite move towards entirely local, entirely private models for regulated industries. That would also imply some decentralisation, and might give Europe a chance in this madness.
Rui Carmo
