I had something like this:

{% for pebble in pebbles | sort: "updatedAt" | reverse %}

Which gives a build warning:

Liquid Warning: Liquid syntax error (line 18): Expected end_of_string but found pipe in "pebble in pebbles | sort: "name" | reverse" in pebbles.md

Instead you have to do a few reassignments to put the array in the right shape:

{% assign pebbles = pebbles | sort: "updatedAt" %}
{% assign pebbles = pebbles | reverse %}
{% for pebble in pebbles %}

So, now you know! 🛋

However, if you load a file that has an ANSI encoding, it will show as a completely white glyph. This is different to a regular space, which shows a tiny orange dot between words.

How to make the NBSPs appear

You can make the NBSP show up in an ANSI file by selecting the Character Set that’s in use, from the Encoding | Character sets menu. Usually on an “English” computer, this will be Encoding | Character sets | Western European | Windows-1252. This info isn’t saved with the file, so you’ll have to reselect it after each load.

Or, instead, you could convert the file to UTF-8. On the encoding menu, click “Convert to UTF-8” and then save the file.

Bye bye now 💽🧑‍🚀

I was shlepping along in my Teams instance the other week and I thought of some cool features I want (additionally to multi-way screen sharing and ambient presence).

Cool features for collab virtual offices

Acronym/initialism hints

Say I’m writing “lmk” and some poor soul doesn’t know what it means. I could proactively highlight the word in my message, right click, and add a definition, “let me know”, which would go into my profile glossary.

Next time I write lmk, it would have little dotted lines underneath which will show you the definition when you hover.

You could use it to spread business knowledge by defining internal terms this way too. So you know what Steve means when he says you need to “Raise an INC in SN re RRA changes for LL+RP”… darn it, Steve.

Comms preferences in profile

I read a piece ages ago about the Obama Whitehouse and how all staffers circulated their “communications style memo” which detailed their comms likes/dislikes, down to whether they were pro-or-against emoji, bullets, platitudes, getting to the point succinctly, etc. Haven’t been able to find the article in a long time.

There are lots of popular views held by information workers nowadays, many of which have handy (if somewhat blunt) “websites as answers”. My favourite is No Hello.

We could have preferences attached to our profiles that list this stuff out using a mix of selectable items and free text.

Imagine you’re about to send someone “Hello” and their preferences include NO_HELLO - then you’d get a little warning + advice in a status bar next to the message input - just like when someone has an active status in Teams today.

🤔✅ Gary prefers messages that include the topic, like “Hello, could I schedule the cert change for tonight?”

Likewise with emoji, GIFs*, very short messages, very long messages, AI slop, etc.

How to detect these situations is left as an exercise for the implementer.

*Preserving Joy

Ok, the further I get through this post the more I sound like a totally unacommodating nark, so you should know that when someone sends me a “Hello!” I don’t actually just reply with “https://nohello.net”!

All human interactions in the virtual office are kind of precious, because we’re often engineered out of each other’s lives.

Really what we want to control is whether we’re getting and enjoying messages from the right people at the right times to preserve focus and complement team unity.

I think it’s worth screening some of this stuff at the recipient side, so if the sender is just trying to inject some joy (like a GIF), they don’t get slapped down, but someone who wants to reduce motion and distraction can filter that out automatically.

There are ways that these papercuts could be solved at recipient side in all cases instead of attempting to modify sender behaviour. An autoresponder could know my acronyms and initialisms and answer “lmk?” with “Let me know :)” and answer “Hello” with “Hi! You now have my undivided attention and I’m ready for your question! What’s up??” (kidding).

But, knowing that you were being screened by someone’s robot secretary would be its own black-mirror-ish problem.

We go forth! 🛃⛹️‍♂️

Back in 2023 I thought this could be a big idea, something I could build and market, that could earn a following and expand over time to be a business, grow features, and even challenge incumbent language learning software (!).

Now that there is no “software moat” and anyone could build Lingomio in a day, the goal is much more humble. This is simply a home-cooked app for me to enjoy, that I can invite friends into if it would be useful to them.

That goal suits its current state which is basic. I’m spending some courage/embarrassment to announce and release even though I’m at perhaps a v0.1:

https://lingomio.sign.me.uk

Anyone can register and log in today. All basic accounts come with 20 LLM explanations per month.

Because “Debí Tirar Más Screenshots”:

Features

• Add glosses/headings/comparisons/mnemonics/usages to organise your language learning.
• Store and sync multiple books across devices.
• Use the parrot 🦜 to break down word-by-word explanations of a Usage block, then save these as Glosses if you want.
• LLM calls are cached in my DB so you’re only debited once for each unique parrot call, even if you don’t save the glosses.
• 20 parrot calls for free per user, and email me to get more… lingomio at this domain would be good. I could probably expand this a lot cos each call is about 200 tokens including the cached prompt, and the billing is cents per million. I have $2 in credit with DSeek so we’ll see how long that lasts 😊
• Import/Export to/from Langpad.
• Export all your Vocab as plain text to keep in a notes file.

Planned

• Get the parrot to return a full sentence translation as well as word-by-word.
• Drag and Drop

Hoped For

• Better support for more languages (Ruby markup for CJK)
• More imports/exports - Anki/Duolingo/Mass text import/?
• Allow conversion of entire Book to another learning language
• Better onboarding, but also, who cares?
• Quizzing you on book content
• Sharing

Really the goal is to be useful to me. I’ve wanted a sync’d langpad for 3+ years and this visit from foreign relatives was enough of a kick to make it happen. I’ll fix my own papercuts and add my own wants/needs first. No need to be super-product-minded anymore.

Btw, quickly, by way of privacy policy: I will never sell or share your data. Entries you press the Parrot on are sent to DSeek or OpenAI for translation. Your email is not currently used for anything and I will not send any automated marketing mails. I may reach out by hand to email addresses that sign up and show usage. App feature usage is tracked without use of cookies. Your book content is not encrypted at rest. Your password is securely encrypted!

Thanks for reading this, and if you like learning languages, I hope you find Lingomio interesting 💡👨‍🏫

I took off the last full week of June because we have family from Argentina here. This was a great idea and I’ve enjoyed it a lot. But, I got ill on the Wednesday with some kind of Ear/Nose/Throat/Fever thing which has made heatwave holiday week less perfect. And I got prickly heat, too! So fun.

Both the kids have had lurgy and la niña even had antibiotics. I guess they infected me.

Having Argentinos holidaying in the neighbourhood means lots of time sitting around with mate 🧉 and snacks 🥯 and the weather means every afternoon is a paddling pool party for the kids. We’ve been for walks in the pine woods, and the parks, and it does make you appreciate how many cool things there are in the local area. So in those terms it’s been a great staycation.

Time off, speaking foreign languages, and generally changing pace all do wonders for my creative energy. Oh hey, you know what, I did play some OpenTTD after threatening to get back into it and this also has some mysterious effect on my ability to use a computer outside of work hours to create stuff. Somehow fake gaming productivity parlays into real productivity.

I also put OpenTTD on 5yo monkey boy’s Linux machine at the weekend so he could have a go painting some railtrack with no pesky competitors to stop him. He found the bomb tool and decided that blowing up farms and replacing them with forest trees was the best task. Maybe a future in regnerative agriculture?

The Big Work News

I finish at Markerstudy Group on the 28th of July (or slightly earlier if baby #3 comes before that).

I’ve wanted for some time to get out. Working at Paymentshield was one of the priveleges of my life. Working at Paymentshield under Atlanta Group was also pretty fine and I had some really great times, particularly working with the NashTech ODC in Hanoi and going through their Take The Lead training. Since our acquisition by Markerstudy Group, although I was 100% ready to give it my all, let’s say the company has not given its all back in return. It’s not for me.

I’ve been so jaded lately with Big Enterprise Software Development that my first choice would’ve been a sideways step into DevRel if possible. However, almost all DevRel roles now seem to be for companies making AI tooling that erodes quality and purpose. For the last few good ones, it turns out to be a tough sector to break into.

I felt such a need for a total refresh that I looked at Kotlin jobs! I had in mind that I wanted to change two out of the following three things: 1. .Net, 2. Teams, 3. Jira.

In the end, I happened upon applying at DayShape. While they’re another MS ship, they’ve concretely demonstrated before, during, and after the interview process that their Values are real, and it shows promise to be a tremendous place to work.

For one thing, they were very understanding about my growing family, saying I should start when I feel most ready and comfortable, so I’m not working at all in August. It’ll be the summer holidays, the baby will be here… and this one gesture had Helen almost crying with happiness.

In September I start at DayShape as a Tech/Team Lead.

I’ll go from managing 7 people across 3 scrum teams working on 5 platforms… to managing about 5 people in 1 scrum team working on 1 product. A much cleaner, clearer job role. I’m looking forward to that.

Still working from home, on the same family friendly 9-5 hours, with occasional trips to Edinburgh (a beautiful city).

Lot of change this year and God has been so good.

I leave you with some fanart that monkey boy did of the GCompris loading screen.

Bye now 🎎🛣

The breakthrough turned out to be setting the ‘trust proxy’ setting in Express to 1. Does that mean “yes”, you might wonder? Not quite! This value can be set to a bool, a number, an IP address string, or a function! Welcome to JAvaSCritp!

// Get NODE_ENV, which defaults to development
// and trust first proxy (pass session cookie) in prod
const isDev = app.get("env") === "development";
if (!isDev) app.set('trust proxy', 1);

I have to thank Bobby Iliev for his helpful answer on the Digital Ocean forum that led me to the solution.

Admittedly, I have confused myself a little bit and I’m still not sure I 100% undertand the solution.

When I was testing this fix in Lingomio, it was by sending Username and Password for login, and there was a separate face-palm-worthy bug extant where I was just not setting the Username/Password values in the web request at all. They were always blank. 🤦‍♂️🐛

I tested using sg-login as a simpler example and this started working when I changed from true to 1.

Now, true should accept all proxies, so why does 1 work?

Seemingly there is a difference in what Express sees as the Client IP at the end of the request chain.

From the docs:

If true, the client’s IP address is understood as the left-most entry in the X-Forwarded-For header.

(If number) Use the address that is at most n number of hops away from the Express application. req.socket.remoteAddress is the first hop

…and so - while I haven’t tested this - I feel like the remoteAddress would be the address of the reverse proxy like Caddy.

Is this why it worked? Dunno. But I feel satisfied that 1 is both a more restrictive solution (than trusting all proxies in the chain) and more effective. And I’m happy with that.

Stay cool 🌇🍃

• sg-sste - Ste’s simple templating engine. This had very little documentation and was way behind the version I’d lashed up for sg-login before. So now it has parity and some more (correct) docs.
• sg-login - Basic login apparatus boilerplate for Node/Express.

Moved from Netlify to sign.me.uk

• pigskin
• idioma

Fixed

• numberland - fixed background image

Code takes long… life is short… 🎭

His Mum looked for and found the egg, returned it to the shop, made boy apologise to staff, etc.

Consequently “no eggs in my pockets” has become code for “I’m obviously hiding or negating something which I’ve definitely done”.

Claude has this habit of mentioning design decisions, in public-facing documentation, about things that were explicitly removed during the development process.

I’ve ended up with Swagger docs that say things like:

 - There is no <c>IsIndicative</c> flag — quick vs full is determined by the route, not the body.
 - The Something ID (SID) is NOT carried in the body; it is derived from the API key (decision 3).

…they… don’t need to know this, man. And in the API docs it generated:

The host is API-only (e.g. rest.myservice.co.uk), so routes do not carry an /api prefix.

In general it has this habit of putting work-in-progress notes in final work outputs.

Holding it wrong

The trouble here is that the Generalist Claude has tackled all the tasks in the epic.

On a one-and-done agentic development effort, this is quite normal. I put a lot of time into the spec and supporting context, and clarifying the plan that it generated (21 sub-decisions), but I am not going to write skills for this codebase at this time.

Ideally, writing swagger docs would be delegated to a /swagger-docs skill which the generalist calls upon when it notices it’s writing swagger docs. Likewise with writing API docs to a markdown file.

The lighter, more immediate fix might be to add to AGENTS/INSTRUCTIONS/CLAUDE/WHATEVER.md and give it some stern warnings about these things.

I personally feel that’s a losing battle:

Negation doesn’t work as well as affirmative requests (with LLMs as with humans - say “remember your keys” rather than “don’t forget your keys”) and you can never cover all of the breadth of the specific things you want it to avoid through rules like this. Instead you risk an expensive, inflated context.

So What

No “So What”! I just wanted to joke about it really.

Sometimes this will happen and it’s a laugh.

If you’re working long-term on a single project with a big future, you can invest in writing the skills so that each area is carried out in a special way.

But for me? At this time of year? In this part of the country? Localized entirely within my kitchen? Nope! 🍔🥚

🛶

Welcome to this months’ STEream of consciousness.

I have some really nice friends that encourage me, and love me, and I appreciate them a lot.

If/when I log into a personal computer at the end of the day I really don’t feel like coding or building. I don’t feel like much of anything. I might get back into OpenTTD… which are dangerous words to think or write.

Partly I’m like, what’s the point of writing guides to doing anything any more? You can ask gippity or claude. And if my solution to the problem is totally original, it’s just going to get stolen and regurgitated at some time. Maybe that’s a little nihilistic… but I suppose it means the only thing to write that gives me a worthwhile feeling is something unique, personal, and unuseful to the MACHINE.

I have a laptop here with an EndeavourOS build from 2022 on it. I tried to revive it and found I had forgotten the sudo password. It took me 2-3 days to remember it. Then I found out that Arch has moved on in a few interesting and very breaking-changes kinds of ways since 2022 that required me to get real familiar real fast. This was an enjoyable intellectual challenge… doing linux sysadmin makes me feel smart.

I did this because the Windows laptop that my kids use sometimes for MSPaint takes forever to start up - just hanging on a black screen - due to a driver bug. And I thought it would be good for them to enjoy watching Linux boot up from time to time. I found that KolourPaint is a faithful recreation of 90s MSPaint which is better than the one you get on Windows 10 now. So I took a few hours over a few days to yay and pacman my way into a working Arch system to put some stuff on it for them.

It worked! And let me tell you about this really cool thing, GCompris. I mega-recommend this. Turns out you can get it on Windows too, so now it’s on both computers, but wow. This is some really cool free software - I’m thinking of wiring £10 to the maintainers or KDE, just bc this is the kind of package my parents would’ve paid for on DOS when I was a wee tyke.

My kids do not get tablets, personal devices, no unsupervised screen time of any sort. So ‘computer labs’ is a special time set apart, generally at the weekend, where they can work on mouse and keyboard skills. That’s how we do in the Griffhaus. And GCompris is a perfect fit.

Unrelated to anything: I cleaned up some old files and found this single entry in a file called bible-tweets.txt which I now share:

…For ALL are problematic and have fallen short of the glory of God …but this is how we know what love is; while we deserved to be cancelled, Christ died for us

I think this is actually funny and good and appropriate, and the more times you read it over, the more it makes sense. Like, if you’re just picturing one segment of the population that feels it fits, read it a few more times.

Quick-fire round:

• YouTube videos are funny but YouTube comments are funnier. Man, people out there are so clever, it just sends me.
• We watched a nice K-Drama I’m calling “Sleepy Korean Girl QVC” which was actually called “Sold Out on You” and you can find it on Netflix. It was on gradual release?! Which felt very oldschool - waiting til next week to watch episodes of stuff!!
• One of the drawers in the new chest of drawers for baby does not yet contain baby things; it contains unbuilt Gunpla and a panoply of Asian snack foods. You see, we already had the unbuilt Gunpla, and Helen said, “why don’t we make an event of it? We’ll order some snacks and have a few nice evenings building these with music and crisps and drinks” - what a great idea. So now we have £55 worth of snacks in a drawer to encourage us to finish building £27 worth of plastic robots.
• Oh, this blog is now off Netlify and onto my VPS. I was a bit scared of doing that bc Netlify observability was saying I get ~4500 hits/week, and that was more than I thought. Mayhap a lot of that is bot traffic and scrapers? But the site seems to be humming along just fine on this OVH box, and actually loads faster in the UK - winner.

Lately if I find a blog post I like, I just email the writer, and that’s my new way of building community. If you’re a real human and you like my things, email ste at this domain! Just do it! Doesn’t matter how you write - no pressure no problem!

Oh dang I still didn’t articulate my big life/professional news. It’ll have to wait, I’m tired of writing.

Til’ next month, or maybe before, but probably not 🎑🎠

(old man yelling at cloud energy) They don’t put model names on the back of cars anymore. I’ll look at a cool car and it just says “Cupra” with no model name. They’re too cool to write it down. Also some makes have the embossed text, in black, on a black car, so it has zero contrast. It’s like skeumorphism is taking over the real world! Where’s my WCAG contrast guideline?!

Well, we bought a very uncool car. I’m a minivan person now. I have a minivan. (slaps roof) this baby can fit so many babies

For you see, in the summer there will be another Griff Baby. So we’ve had a massive chuck-out at home and replaced my wardrobe with some drawers for baby. Did you know Ikea are discontinuing the MALM? Where do the years go??

🧃🧆

It was my birthday recently. The kids and I ripped down one of our small wood sheds and built a makeshift asado in its place, and I’m really happy with it. We got some sausages from the grumpy butcher and cooked them over real wood charcoal, and they were delicious.

Also for my birthday I got Mike’n’Ikes. I love these for the massive flavour hit, but I always have to brush my teeth straight afterwards bc you can feel the instant plaque growth unlike that from any other sweet. Maybe it’s all the corn.

I asked for and got a new USB charge bank (so exciting) but in my rush to optimise for capacity and price I accidentally got one that is 3x heavier than what I’m used to. I have this beautiful, dinky little “POWERGEN” charge brick from about 2010 that’s 8000mAh and only 150g, whereas my new 20000mAh one is 450g :(

I am doing “agentic development” at work, and, as Seuss said: “Don’t ask me what ‘Voom’ is, I never will know, but boy let me tell you, it does clean up snow.”

There is too much to learn and write on the topic to ever begin. And most of it is wrong (or will be next week) and no-one is an expert, but some people have learned a bit more about it without frying their brains - listen to such people.

When Geordi or B’Elanna are using the Voice UI of the ship’s computer (that’s right, we’re having a conversation slightly Star Trek here) they’re still engineering. They prompt like “Computer, what if we moved the plasma confabulator to outside of the nacell conduit – would that reduce the phase variance?” and the computer basically does the dragging-and-dropping for them, and re-runs a simulation, and reports the result. They’re still the one driving/thinking/engineering. Makes you think.

⛵🥽

So, the Hantavirus, huh? We happen to be related to two frontline clinical staff who handled the 2019 Epuyén outbreak of Hantavirus that killed 11. When Covid came along, they said “thank goodness it’s not more Hantavirus”. But I get the impression that while Hanta is deadlier, it doesn’t propogate so easily as Covid.

Every time I publish life notes, I worry, “what about opsec?!” like am I publishing too much identifying detail? And then I decide to relax. I don’t post exact dates, names, or anything on here.

So that’s all. Maybe a technical post will land before next month, maybe not. Next month hopefully I can reveal more interesting life changes.

Until then… annyeonghi gaseyoooo 👋🚞