WILL AMAZON MOVE AWAY FROM ANDROID?

I keep wondering how much further Google will need to go in order to trigger Amazon going all-in on their Vega OS. Given the recent Android changes, maybe we’re already there. Android still powers a number of Amazons devices; Most notably their Fire range of devices (TV streaming sticks and tablets). Google is now doing more work behind closed doors, and seems to have started a big move to ensuring that they’re in control of Android, and that the open source community, non-GMS devices, and smaller developers, are a secondary concern.

HYBRID WORKING

When talking to companies about my next adventure, and the idea of Hybrid working occasionally comes up, and one of the most illuminating questions I try to ask is “Why?” Some companies come up with good, solid, answers. For example, the folk at one company talked about my role potentially involving talking at conferences, and understanding how customers currently work. Those things are far easier to do on-site (you’ll be amazed at how open folk can become when you’re chatting over coffee compared to a video chat), and as a result these are hybrid companies that I’m interested in, along with those that are remote first.

POLARISED AI EXPERIENCES

After a month of digging into AI I’m developing some more solid views, and it’s a very nuanced situation. As things stand, using machine-local hardware isn’t going to give a great experience for most folk, and I think this explains why there are some pretty polar views. There are two main variables when it comes to LLMs; Parameter count, and Quantization (i.e. down-scaling). Both of these numbers are proportional to the quality of the output, with parameter count being the largest influencer. This means if you’re using a model that has 120 billion parameters, in 32 bit floating point representation, you’ll probably get a good experience…. but… you’ll also be looking at around 500GB of memory, and that would be GPU memory for the best experience, which is a ridiculously expensive setup to own, which is why cloud providers have a place in this world.

TESTING GPT OSS ON A MAC MINI

I’m pretty impressed by the 20 billion parameter gpt-oss model as a code reviewer. The comments generated, which are posted on this pull request, seems to indicate the model inputs allow for a reasonably sensible output. In terms of performance, a couple of minutes on a 24GB m4 Mac mini seems reasonable given the turn-around speed (i.e. no waiting on other devs for a review), and the relevance of the comments.

REDUCING AI POWER USAGE

This interview gives me some hope that AI will not be the climate disaster that folk think it could be, if we extrapolated the current power and usage trends. There’s a growing understanding, by folks like Fractile, that two things are at play; Datacenter hardware refreshes happen every 3 (sometimes 4) years. GPUs are power hungry beasts. So there’s definitely opportunities for companies to make AI calculation focused hardware, without the GPU components, that perform at much lower power consumption levels.

LEAVING SNAPP AUTOMOTIVE

Today is my last day at Snapp Automotive. It’s been a difficult decision to make, but, for me, leaving is the right thing to do for many reasons. Firstly, I’m a believer in adding value or getting out of the way. The current state of the automotive industry, and where Snapp Automotive is right now, means I don’t believe I’m adding enough value to stay around. I could keep taking a salary and spend my time on hobbies, but I believe that course of action will only cause friction in the company, and put things on a bad path. I’ve been in the position where I’ve put in long hours to earn money for a company, and watch other folk seem to deliver little and take home a good salary. It’s not a nice feeling, it’s not something I consider an honourable thing to do, and so it’s not a path for me.

ORGANISING TEAMS

It always makes me smile when I see the moment that a project manager/scrum master/team lead realises that you can run a project using software algorithms, and not just follow a book. Imagine having groups of your servers stopping at a set time every day, some exchanging information with others (most of which gets discarded), others not really sharing any useful information (but they have to share some information because that’s what’s expected), and then all going back to serving customers 30 minutes later…. sounds like a really bad design? Well, that’s what a meeting is, but with people instead of servers.

GOOGLE'S EVER INCREASING POWER USAGE

If you thought Google was achieving anything in relation to reducing carbon emissions, read pages 80-82 of their 2025 environmental report. They try hard to make it look like they’ve achieved something by calling out “avoided emissions” on a graph showing a multi-year upward trend in actual emissions. Even when, as a planet, we’ve seen multiple years of unprecedented destructive weather anomalies, the “Don’t be Evil” company is failing to tame its emissions.

FUNGIBLE DEVELOPERS

I’m always puzzled as to why many purchasing departments don’t understand that software engineers are not fungible assets. The term “fungible asset” is something I first came across when Simon Stewart was giving a talk during my boot-camp induction when we were both at Facebook. A fungible asset is something you can swap for an equivalent with no detrimental effect, and software engineers definitely aren’t trivially swappable. You can swap one 10 euro note for another because they’re fungible assets, but few other things are. Would most folk accept someone giving you a Polo instead of a Porsche?, No?, Why not? They’re both cars and they’ll both get you from A to B, what’s the problem?

LOCAL AI FOR CODING

I’ve recently been playing with Artificial Intelligence and large language models, and there’s a trajectory that I’ve not seen widely discussed; Local AIs. There are a lot of folk and companies who don’t want to use a cloud based service. They want to run things locally, and they want their data to stay behind a firewall to give them some sense of privacy when it comes to their work. At the moment there’s a push around cloud based AIs that doesn’t align well with this privacy focused thinking; Do you really want to use an IDE plugin that can push your source code, documents, etc. off to somewhere you don’t know, outside of your control?