Nat Voisey's blog

Outlook newsletter; For my colleagues.

2026-04-22T12:28:00Z

I have this small blog with the byline, “For my friends.” However, my personal website isn’t the place to be posting about specific Spark work I’ve been doing. There are posts I’ve made on Yammer/Engage, but it’s not the best place for blog-style posts.¹ Sharepoint sites are another alternative, but they seem impermanent, sometimes disappearing when an admin leans on their keyboard. Then I found the Newsletter feature in Outlook; what a concept!

It’s fascinating to see Microsoft implement Twitter- and Substack/newsletter-like features on an enterprise level, but I can definitely see the benefits for my company.

The next few posts will be about:

Speeding up a credit process using Excel, then Snowflake
An inspiration, encouragement, or feelings-based post
My recent projects, questions, or leading

The newsletter is a way to get my ‘work-only’ thoughts out, processing through writing, and sharing them in the most corporate way possible.

It’s also a way to have fun. 🕺

Ngā mihi,
Nat

The first line ends up bolded to act like a title. ↩︎

Smaller ideas

2026-04-18T11:55:00Z

I keep writing about social platforms while drafting ideas on paper. These ideas are often existential and philosophical, requiring analogies and exposition. In a word; big.

One day I may write about these ideas. Maybe they’ll be individual posts that turn into a book, or hopefully something comprehensible. But right now, I need small ideas.

Small ideas are palatable and easy to process. Small ideas are accessible to everyone, and simple to understand. Small ideas can also be a consumable, but even small bites can be rich in flavour if you cook them just right.

Refine the bits that sound good, trim the bits that don’t fit the bill.

Trim your darlings.

Grit and tenacity

2026-04-14T13:21:00Z

Projects are an expedition through ‘The Jungle of Unknown’. Clearing a path with a machete, hacking through vines of problems, setbacks, issues, and trees with bushes, and people with negativity. The toil is rewarded by reaching a clearing.

Today, I have cleared a path to a decision clearing. Our team has discussed what options we have, and now I need to take a sharp left turn to start hacking away again. Pushing without relent, slicing through ideas, barriers, and the mind.

This is where ‘grit’ is needed. The project can reach a clearing by the end of the day, but only if I put the effort in to reach a clearing right now. I can even see the clearing ahead, but to reaching it I the stamina to keep going. Grit is carrying on; tenacious unrelenting. If I keep going, we’ll be able to rest, assured that someone will pick us up, or maybe they’ll encourage us on our next leg of our journey.

This is grit; the will to carry on. There’s always a clearing ahead.

’Grit and tenacity’ was inspired a Hacker News comment.

Following, reading print

2026-04-01T13:25:00Z

Surprisingly, I’ve done moderately well in ‘reading print’ since my previous post:

Formatted The New Yorker for print using GoodLinks and Marked 2
Printed an edited PDF version in a Duraclip binder¹
Realised how much simpler the magazine version is to 80 gsm paper, so I got it out from the library
Hope to to subscribe to The New Yorker again in the future, but I want to be up to date before I commit to a lifetime torture

Reading print at night is a success, whether it be books, The New Yorker, printed bibles and commentaries; it’s all print.

In my quest for print, the library kindly asked me to store a small collection of their catalogue at my house. I send these books back whenever I’ve read them, or when someone requests a hold on one.

I don’t mind that the library needs to store books at my house, because the books in question are expertly selected for the tastes of someone like me.

The way I receive non-print media has also changed. I now favour² newsletters over RSS, algorithms, and point systems, because:

Emails can’t be revoked
Most newsletters are weekly
Everything’s under one roof

I receive; the Hacker News digest (30), and some other good newsletters through email, then some RSS and YouTube using FeedMail.

Although my habit of reading print may be solidified, a habit of publishing has not. This is my next goal, and like most goals, it started when my neurons fired.

Duraclip is the best! ↩︎
I also find myself favouring published media, over news-print or self-printing. ↩︎

Reading print

2026-01-07T17:00:00Z

Since returning from camping, I’ve been trying to continue the routine of reading at night. My final crutch was removed by using our HomePod as an alarm instead of my phone.¹ My hope is that this is now a freestanding habit.

While researching Ezra Klein from something I was reading, I found an article from 15 years ago describing a similar approach:

At night—if all this goes perfectly—I want to read print.

I also—if all goes well—want to read print in the evenings.

Most of my friction to achieving this has been removed. Although there’s just so many things to do in the evenings; so much to research and find, so much fun to be had. If I really feel the need to read an article, it is printed out. The biggest step is always my first; getting off my computer and going to bed.

My mornings have a similar tune, although I’m not like Ezra, who gets up at 6am to read news clippings. I’d like my mornings to be standardised, relying on a run for energy and mediae I’ve already collected to read; not new media that has come into my inbox.

In the time since that article in The Atlantic, it’s crazy how little the world has changed over 10 years. Articles from the past 10 years still describe my present moment, whereas articles from the past 20 years feel nostalgic and utopian, almost foreign to a young adult.

My hope is also, that writing and publishing this will solidify this habit even further.

How I got here:

Hacker News Algolia search, because I’m waiting for my Hacker News Weekly to come through (was using Digest, but it doesn’t always send; before that was Newsletter, but it’s editorial and contains too many links)
Hacker News post
The suck is why we’re here by Niklas Göke
Nik mentioned an interview with Ezra Klein (which I saved the podcast version of). I’ve heard of his podcast, but didn’t know he started Vox, or used to blog.
Wikipedia
Trying to find Ezra’s old blogs, Reddit post
Ezra’s original Typepad with original and newer designs²
Ezra Klein: What I Read on The Atlantic

Turns out ‘Forest’ sounds are very similar to the ‘Bird’ sounds alarm. ↩︎
The fact that Ezra used Typepad, then went to The American Prospect and WaPo, is a reminder that the author of the blog is the thing that matters, not the technical aspects or design. Ezra is his own person; platform agnostic. However, John Gruber is Movable Type, Manuel Morale is Kirby, Nicolas Magand is Blot (along with me), Jason Fried and DDH are Hey World, Sam Altman is Posthaven (investor), Todd Vaziri is still on Blogger. ↩︎

Stand Up, Teams bot

2025-11-28T12:45:00Z

Every week at around 8:30am our Squad asks, who’s on Stand Up? How could we automate an answer for this, every week?

Workflows in Power Automate

I’ve seen workflows used in company wide messages, but I’ve never used one myself, or one that runs on a weekly schedule.

The flow needs to:

Run every Monday at 8am
Post a message in our Squad chat
And @mention that Squad member, so they get a notification

Using the ‘post message in chat or channel’ is the first test. I linked the action to my Squad chat with a test message. Mentioning a user seemed too hard in the moment, so I gave up to come back later. Unfortunately ‘saving’ a flow can also test it, so it sent a test message to my Squad chat:

Workflows: @Name is on Stand Up this week.

To which my Squad responded:

Thank you Name!

Name can be on stand up every week imo

Well, now the secret’s out, I should probably make a working concept.

To mention a user you need their User ID and a ‘mention token’, which is really just their User ID with <at> tags, but you have to use the ‘mention token’ action, you can’t just use tags in the message text.

Find the User ID by searching their full name in the ‘search for users’ action, then pull that result through to ‘get an @mention token for a user’, then you can pull the token through to the ‘post message’ action.

But how do we know who’s on each week?

OneNote to Excel, and serial dates

I moved our roster from OneNote to Excel, and even shifted the roster to account for the summer break.

I asked Copilot to suggest a way to filter tables, and it lead me astray immediately… Suggesting a Data Operation to get the row and column, which may work, but the Excel actions are so much simpler!

Instead of using 3 actions to get the table, current date, and name column; 1 Excel action can do everything!

You’d expect this to be the end, however, you failed to recognise that Microsoft product family members do not play ball! Excel obviously stores dates as serial numbers, and Power Automate stores dates as ‘dates’, whatever dates are.

Copilot was actually helpful in writing an expression for this:

Convert now in UTC, to NZST, format date, covert to ticks (100 nanosecond intervals since 0001-01-01), subtract Excel’s epoch date (1899-12-30), divide the ticks by 864,000,000,000 (ticks in a day) to get the date serial number.

div(sub(ticks(formatDateTime(convertTimeZone(utcNow(), 'UTC', 'New Zealand Standard Time'), 'yyyy-MM-dd')),ticks('1899-12-30')),864000000000)

Great.

Works a charm! I updated the ‘post message’ action to be our Squad chat, and scheduled the flow for next week.

Here’s hoping it works on Monday!

Notebooks and platforms

2025-11-03T14:00:00Z

Perfection; notebook, writing app, blog

Field Notes for notes, Writer for writing, Blot for hosting (with a Writer inspired design).

But I’m always looking for perfection—optimisation of my notebooks and platforms.

Field Notes is expensive and embraces seasonal design; but I bought some a while ago, meaning I can’t buy the Moleskine alternative which is somehow hard to find anyway, and Leuchtturm’s simplest version is an expensive pocket Bujo.

Writer feels bloated and not as simple as it used to be; but it still looks good, and the alternatives are worse. I also use it for my blog design, which is hard to depart from.

I worry that I haven’t found the best version of my tools, when I should really just be using them.

Sometimes the ‘use’ of the tool is hardest to grasp. If I found the perfect notebook, writing app, blog host, or computer, what would I actually do with it? If a dog caught a car, what would it do?

My utopia of perfect tools frees me from thinking about the ‘perfect tool’. But then what would I do? This has been my only constant consideration for years. As the world produces tools, I am the one to consider them. This is my job.

Therefore, I am a fool.

The tools were always perfect. I am the one who is imperfect. The tools were already as optimised as they could be. If the goal was consumer optimisation, I would never get there. There’s always a new product for the consumer to try. I am a fool, once again.

If the goal of simplicity is enabling other things, try doing those things first. Focus on the thing, not the tools for it, or the means. If we need more time, what do we need more time for? Why not just do that instead?

Optimisation is a singularity. You can only project optimisation, you cannot wear it.

Fascinations from Anthropic’s Prompt Engineering

2025-10-13T10:15:00Z

Everything is easy, and simple.

Sometimes these features are the ‘product’ the AI companies are trying to sell you. Some document it, some don’t. Sometimes it’s just their prompt telling the LLM to ‘think through it’, sometimes it’s more than that.

This document is for Anthropic’s ‘Claude’ from 2024. The most interesting section is the answer key, but the introduction and chapters are helpful context as well.

A ‘role’ is helpful. It’s good to define, but not always necessary.¹

XML tags (like this or this this is an email) are very helpful for define where things start and end. You can name your tag anything you like.²

Every interaction is separated by ‘User:’ and ‘Assistant:’. System prompts don’t get this treatment.

The Assistant section is used to help Claude start an answer.³

You can almost use this like a formula⁴ and if you already have examples.⁵

Strange that getting it to think (first give me the pros and cons) can give you a better answer.⁶

The XML and Assistant really came together in the email classification example. “Do not include an extra words except the category. (A)…” And using the Assistant to start the prompt with a ‘(‘.⁷

The magic of LLMs is how they’re able to consider what they’ve written. ‘Thinking’ through step by step is still relevant, but may be implemented differently depending on your model.

You can ask it to brainstorm or make a scratchpad of quotes to consider to avoid hallucination.⁸

You can also ask for reference quotes.⁹

And it’s okay to be comfortable with not knowing something.^9:1

We could all learn from this.

A comment acknowledged how similar this was to managing and communicating well.

Formula for a complex prompt:¹⁰

Task context
Tone context (optional)
Information (in any order?):

Detailed task description and rules
Examples
Input data to process

Immediate task description or request
Optional:

Precognition (thinking step by step)
Output formatting
Prefilling Claude’s response (if any), ‘Assistant:’

The whole course was designed to help you send multiple queries to Claude using Python with multiple variables.

via Hacker News, previous

Writing something that’s ‘good enough’

2025-10-09T09:10:00Z

What is, good enough? Is the point of good enough to publish as quickly as possible; or should I be making the most magnificent piece of writing anyone has ever written that no ever has to write again, because I have finished writing?

Clearly, it’s the first option.

The mountain of drafts I have is ridiculous. Sometimes they’re offhand comments, sometimes they’re small ideas that need extra context, and sometimes they’re good in the moment, then bad when I read them later.

Publish quickly. Don’t overshare. Focus on creating, writing, and story. 3 acts. Publish to let people you’re here and you’re thinking.

If I finish a bad piece of writing, it’s finished. If I only draft a good piece of writing, it isn’t actually ‘good’. I’ll never know if it was ‘good’ at all, because it doesn’t exist. Good writing is finished, bad writing is drafted. Separating the wheat from the chaff takes a harvest. In order to harvest, you need to grow your writing, harvest your stories, then you can reflect on your output.

Sharing and publishing something will always be good enough, it’s just you who doesn’t think you’re good enough.

Apps I actually use

2025-05-19T11:50:00Z

For a long time I’ve used the default iPhone Home Screen as my own. I like to use the defaults for products so I best understand how a company wants to showcase their user experience.

However, I thought it could be simpler. So I moved to solely using the dock, and replacing Safari with the app that was most in fashion at the time.¹

Recently I started listening to audiobooks again, and suddenly had 5 apps for only 4 slots. My first thought was to move the non-standard apps to the page, forming a pyramid, but I didn’t like seeing the app names, and additionally, large icons are for toy phones.

Some of these apps I don’t even use that often, especially Phone and Messages. Could I just do without them?

Yes, and these are the apps I actually use on my phone:

Infuse for movies, TV, and YouTube videos (auto downloaded). Bound for DRM-free audiobooks. And Music, which is used least of all.

These apps are the reality of my everyday. This is what my phone really is. You’d think it would be a calendar, notes, communication, or productivity device, but it’s actually a media machine.

Treat it that way; you’ll see the true purpose of your phone. It doesn’t help you, it consumes you.

Reeder (Classic), Play, nothing, Camera, Infuse ↩︎