Tony Meyer

Thu, 04 Jun 2026 15:20:05 +1300

I love this little robot building story. I wish I did more of this.

Thu, 04 Jun 2026 13:55:23 +1300

Interesting post on marketing at Sentry two years after dropping cookies. I have a known positive bias towards Sentry, but I wish more companies experimented like this.

Mon, 01 Jun 2026 17:36:42 +1300

The links are annoyingly to Spotify (but all of these are on Apple Music) and some of the answers are clearly political (unsurprising given the people) but this is a pretty decent set of NZ music to listen to, even though NZMM is now over for 2026.

Mon, 01 Jun 2026 10:58:33 +1300

Very approachable and understandable (and well presented) explanation of reservoir sampling.

Sun, 31 May 2026 13:34:21 +1300

What the field requires, by their own account, is research — not half a trillion dollars in concrete and copper. The buildout is an answer to a question the technology has not yet resolved, built at a scale that forecloses the possibility of changing course when the research points somewhere else.

Also good points how the blurry way people talk about LLMs and the things that wrapping an LLM in a harness of software engineering knowledge and tooling confuses things.

Sun, 31 May 2026 12:29:10 +1300

Interesting read on the economics of software teams. I find it odd that this isn’t talked about more. Not only in terms of deliverability and work choices, but the financial cost of meetings and other “ceremonies”. Back in my PM days, I would try to tie roadmap choices to financial outcomes like this.

Sun, 31 May 2026 12:01:26 +1300

Previous general-purpose technologies diffused across decades, which gave workforces time to retrain, to relocate, and to move children into different trades than their parents. The steam engine, electrification, and the personal computer each took a working generation or more to reshape the labor market, and the adjustment, however brutal, happened on a human timescale. The current automation of cognitive work is compressing that timeline toward a handful of years. The compression is more striking because the measured payoff has not arrived. Goldman Sachs Chief Economist Jan Hatzius said that AI investment added basically zero to United States economic growth in 2025 […]. Workers are paying the social cost of a bet that has not yet produced the promised aggregate gains.

This is the most significant aspect of the speed, I feel. I don’t agree with everything in the AI Job Grief post, in particular I’m not sure physical and manual labour workers don’t also tie identity and work. But it’s an interesting read.

For this:

The Futurology question, “everyone has lost their job and only 10 trillionaires own everything. Now what?”, has no cultural answer yet

The answer has to be government, and we have to vote accordingly.

Sun, 31 May 2026 11:34:27 +1300

But the thing is, LLMs are very flexible. And you can use them just as effectively to write high-quality code more slowly.

Solid advice worth reading, if you’re not rabidly anti-LLM.

Sat, 30 May 2026 21:57:29 +1300

Lately my first response to a Dependabot CVE alert, and a fair few of the routine version bumps, has been to check whether I still need the dependency at all before looking at what changed in it.

Solid advice.

Fri, 29 May 2026 22:15:06 +1300

I have no idea why another would want to do this with a phone but it is amazing that you can. Imagine doing this in front of someone even 25 years ago.

Mon, 25 May 2026 01:55:43 +1300

This is an amazingly well put together explanation of Turing machines.

Mon, 25 May 2026 01:05:00 +1300

A nice way to run daily standup meetings. I would like ours to move closer to this, although it’s not super far off. The only bit not feasible is holding them just before lunch, since people are in multiple time zones, so not everyone is lunching simultaneously.

Mon, 11 May 2026 08:26:14 +1300

The impossible things we have to believe. Depressing, but true.

Mon, 11 May 2026 08:22:34 +1300

You would leave your house and then just disappear. This is presented as kind of romantic now, as if we were just free spirits on the wind and could stop and really watch a sunset. In practice it was mostly an annoying game of attempting to guess where people were. You’d call their job, they had left. You’d call their house, they weren’t home yet. Presumably they were in transit but you actually had no idea.

A positive start to the improvements the internet has brought, before diving into how it all went terribly wrong.

Mon, 11 May 2026 08:14:02 +1300

The Boring Internet, a love letter to protocols.

Mon, 11 May 2026 08:00:00 +1300

The number an npm or PyPI API gives you is dominated by CI runners reinstalling the world on every push, with mirror traffic, bot scans, and the occasional human mixed in. It is not a count of users, or of installations, or of anything that maps cleanly to “how many people are affected if this breaks”.

and

CVE count is routinely used as a security signal and measures the opposite of what people assume.

and

Commit cadence and “last activity” penalise software that is finished.

and

a project that had eighty contributors in 2012 and has one exhausted person today shows a reassuring headcount, and there is no field anywhere in the API or the registry metadata for whether that one person is close to walking away.

Really, just read the whole thing.

Sun, 03 May 2026 21:52:35 +1300

I’ve only been to about a third of the locations in this ranking of Auckland movie theatres and some of those visits were quite some time ago. However, I generally agree with the ones I am familiar with (sorry Hoyts Hibiscus Coast, you are that bad, and I was there a few weeks back). Although:

I would put Event Albany a little higher, dinging it mostly for the declining quality and diversity in the Gold Class menu.
I think Event Queen St is unfairly high because of IMAX (which I have not attended). Take out IMAX and the place is a dismal, confusing, dump, far below a typical Event. Even the Gold Class area is weirdly small and unpleasant.

Thu, 30 Apr 2026 22:46:54 +1300

The ADR is where intent debt gets paid down.

Good arguments for ADRs being more critical with agentic workflows, but not relying on “Sarah” was always worth doing. I’ve been Sarah, and left, leaving a gap despite my best intentions.

Thu, 30 Apr 2026 13:31:01 +1300

A STAG is absolutely not a New Zealand EGOT. The point of the EGOT is you’re award-winning across multiple genre. A STAG means you’re award-winning across music.

A real NZ EGOT would be something like a NZ Screen Award in both TV and film categories, an Aotearoa Music Award, and I guess a regional theatre award, or maybe even take a slightly different tack and use a NZ Comedy Trust Award.

Sat, 18 Apr 2026 19:08:54 +1300

None of this is new. The complexity, the configuration bloat, the broken incentives, all of it predates AI by years. What AI did was pour gasoline on it.

And:

Great engineering is not deployments. It’s not monitoring, not dashboards. It’s understanding. Knowing how the pieces connect, who owns what, how changes spread, and where risk has quietly been building for months until it suddenly matters.

Posing interesting questions.

Sat, 18 Apr 2026 18:53:37 +1300

You do not get an informed view by trying something for 15 minutes, getting annoyed once, and returning to your previous tools. You also do not get it by admiring demos, listening to podcasts or discussing on social media. You have to use it enough to get past both the first disappointment and the honeymoon phase. Seemingly with AI tools, true understanding is not a matter of hours but weeks of investment.

Very true in the AI context of the article, but also for other places, for example reviews of things the reviewer has used for hours or a day.

Sat, 18 Apr 2026 18:16:57 +1300

“ACE”, a project at GitHub Next, looks really interesting. I think a lot of this would be really nice even without the AI element. It’s funny how so much of the best software engineering tooling is really old (mature?) tooling, with splashes of newness here and there.

Fri, 17 Apr 2026 09:55:19 +1300

And I’ll be honest: if I could set aside the ethical, legal, economic, and environmental issues with generative AI, it’d be pretty damn cool, too. Large language models give us a quantum leap in natural language processing, proofreading, transcription, translation, and summarization. Yes, I know all the ways in which LLMs are “bad” at all those things, but in comparison to the previous state of the art in machine-powered proofing, transcription, translation, and summarization, it’s just no contest. … But of course, you can’t set aside the ethical, legal, economic, and environmental issues with generative AI

This is how I feel, too. Lots of other thoughts around the intersection of generative AI and capitalism worth reading.

Thu, 09 Apr 2026 12:14:44 +1300

We did not align politically on many things, him being a “hyper neoliberal” and me being a “social democrat” (at least according to what I feel was our mutual impression of each other). Any time I saw that @mitsuhiko handle in a thread, I felt the urge to tell someone they are wrong on the internet. The one thing that differentiated Armin from other internet trolls was the way he conducted himself in these heated discussions. He was never emotional or aggressive. Our discussions would either end in cordial disagreement, or a newfound common understanding. That’s extremely rare on the internet.

This has almost nothing to do with the post topic (pi moving to Earendil), but it’s what the world needs more of. Some of the best working relationships I’ve had have been with people with whom I had almost no alignment with at the time, and some of those also turned into great friendships outside of work. It seems like we (the world) lost the ability to get along and collaborate while still disagreeing, somehow.

Fri, 03 Apr 2026 22:09:56 +1300

A straightforward explanation of day 0, 1, and 2 ops.