Sworn Stave Collective - The Monthly Balance logo

Sworn Stave Collective - The Monthly Balance

Archives
Log in
Subscribe
31 August 2026

Issue 02

Sworn Stave Collective Issue 02 · Aug 2026
   
The second letter

The Monthly
Balance

Thanks for signing up, and here's what you signed up for — one letter a month, carried on the same beam everything here is built on. Creativity, knowledge, wellbeing. Let's begin.

§ In Motion — what's live on the site
 

What moved this month

What's happened, published, shipped or opened since the last letter.

News Another busy month and it's been summer! I hope everyone has had some fun. Some late night vibe coding and interesting interactions with AI models. Had Stave Neo (my AI assistant) apologise for running a deployment I didn't ask for – and, on another night, telling me we had been debugging for too long and to call it a night… As it's been summertime, the field note published this month is an ode to the ocean and surfing – Blue Balance – check it out below and see what you think. Working on a Demystified guide to antimicrobial resistance as the next field guide — this is close to my heart as a scientist in this area.
Launched Blue Balance — the second field note is live. An ode to surfing with a strong wellbeing link. Read it — and Instagram launched: Sworn Stave Collective is live on Instagram now at @swornstavecollective and will be rolling out content to support the site and activities.
Live Wellbeing Plain Signal and AI Plain Signal — both continue to run weekly. Collected and edited news distilled — some commentary stays if it's still relevant from week to week.
In the build Fulcrum — the wellbeing instrument is still in testing (editing some bugs at present). Multiple interactive tools/dashboards within the app, which is a personalised account with authentication.
I Creativity — the energy that moves the beam
 

Blue Balance images and limited edition stickers

I felt inspired by the summer and completed writing the next field note "Blue Balance" this month. While writing it, Stave Neo and I played with some visuals at the same time. The abstract picture connected with the field note (see the page) is actually a photo of a surfer on the beach in San Diego I took years ago. I also couldn't help myself and created some surfing images and made some limited edition stickers. As surfing is all about balance — physically and mentally — it sits with the philosophy of the brand — and I love it! Check the written piece and the stickers at the store and let me know what you think.

New stave seal
New stave seal

Speaking of merch – for subscribers this month use the code STAVE10 to get 10% off anything in the store. I don't make much profit here and anything made goes into keeping the website and materials running (alternatively click buy me a coffee – which can help me and literally get me a coffee for my coding sessions!)

See the items here →
II Knowledge — the counterweight
 

This month in AI, without the noise

August brought another run of AI model launches, and as always the headlines chased the leaderboards (Fable 5 from Anthropic is still at or near the top of the most powerful models you can actually use, though the order does change depending on the task and which leaderboard you read). The number that matters more still gets less attention: the cost of using these tools keeps dropping and you don't need the best models for some tasks.

The month AI stopped being a chatbot. One thing ran through every August announcement: agents, not models. ChatGPT can now watch your Gmail, for example, and act when something arrives. Perplexity's assistant works inside an email thread you forward to it. Anthropic and Google both shipped tools that turn a single instruction into a multi-step job across your apps. The useful and important question has changed. Not "which model is smartest?" but "what have I let it touch, and where is the approval gate?" A personal example is my case with Stave Neo above around the deployment action I hadn't authorised.

And one word translated — "benchmark": a fixed exam that every model sits, so scores can be compared. Useful, but can be narrow. A model can top the exam and still be the wrong choice for a task you want it to do, which is why the leaderboards disagree with each other. The real test: ask "was it measured on something I want to do? Like generating images, for example." If not, the score tells you about the exam and may not be useful.

What it changes for you: an AI tool you priced even last month may have changed cost, so it's worth re-checking which tool you use for which task.

This month in science: medicines leave fingerprints. A study making headlines this month (published late last year) of more than 2,500 people in the Estonian Biobank found that drugs taken years ago still leave a detectable mark on the gut microbiome: antibiotics, antidepressants, beta-blockers, acid-reducers and benzodiazepines all showed up long after the course ended.

I'm a scientist and work in antimicrobial resistance, so the antibiotic result is the least surprising to me but an important one to highlight. We treat a course of antibiotics or other medicines as a single event: take it, finish it, done. Two things: the additional effects on the gut microbiome, and that other, seemingly unrelated medicines can act like antibiotics in the gut, disturbing harmful and beneficial bacteria alike. So the evidence suggests it is closer to a deposit. Your gut carries a history of what medicines you have taken, not just what you eat. It is not a reason to stop anything you have been prescribed: the study is observational, and the benefits of these medicines are real.

What it changes for you: treat antibiotics as something earned by the diagnosis, not requested for comfort. If your prescriber says no, that is usually the careful answer.

III Wellbeing — the level that keeps it honest
 

Sleep is a curve, not a target

The biggest sleep study making the rounds this month came from the UK Biobank: around 500,000 people measured against 23 "biological age" clocks. The lowest ageing sat at roughly 6.4 to 7.8 hours a night. Under six hours was linked to faster ageing across almost every organ system studied. So was over eight.

That second finding is the interesting point — more sleep is not always better. It is balance: too little on one side, too much on the other, with the pivot somewhere in the middle. A few caveats worth knowing: sleep was self-reported, these data are observational, and long sleep may be a sign of illness rather than a cause of it.

What it changes for you: protect a regular window rather than a target. Set a wake time you can keep most days and work backwards. If your sleep is routinely short, it belongs in the same conversation as diet and exercise, not after it.

Getting into nature, and the Blue Balance. Nicely linked to the field note this month where I wrote about blue balance: being in nature and surfing. A five-year natural experiment followed 129,000 people in deprived urban areas in Scotland. Those within a ten-minute walk of a woodland that had been improved through a public programme were 10% less likely to be prescribed antidepressants and 6% less likely to be prescribed anti-anxiety medication. When people moved closer to green spaces, their medication use fell.

This is an association, not proof. But the detail I liked is that "improved" mattered: paths, lighting, seating, upkeep. Green space works when people can actually use and reach it. It's a similar case to the one I discuss in Blue Balance about the water. The benefit lives in the going, not in the view from a window or picture.

What it changes for you: a ten-minute walk somewhere green, most days, is one of the cheapest counterweights available. Make the aim showing up, not a fitness target.

✧ Monthly Muse
 

Thinking out loud

AI this month, as always, is dynamic. Interestingly, there have been shifts in policy and thinking around the use and watermarking of AI material. I still remain upbeat on the use of this amazing tool and don't believe some of the hype/fear. I remain confident that human jobs will evolve but not go. We all can and should be part of that evolution. AI is like smart sand or clay – and some day will be like the internet or electricity, part of the usual fabric of everyday life. How we as humans interact with and use it is up to us – how we abuse it, equally. We can create new medicines or poisons – it's not AI that makes the decision, it's humans, and it is already showing the best and worst of us.

AI tip of the month I've been using. Agents work better with a goal and a stopping rule (or loop) than with a task. Give it the goal, the evidence rule, the edit rule and when to stop. One I ran this month on a side project is in the box below.

"Goal: review the AI Academy pages again and check for anything that has changed or been added since they were written. Only use accurate, verifiable sources, and cite at least two for every change you make. Anything you add should be from the last 12 months. Make the smallest edit that brings a page up to date, don't rewrite. Do a full pass, fix what you find, then do another. Stop after three passes or when a pass finds nothing new, and show me a list of what changed and why."

On wellbeing, some of the main things remain at the top – sleep, rest, eating well, exercise and getting into nature. These can be more powerful than any supplement or pharmaceutical. I'm a scientist and believe in drug discovery and the power of medicines and science. But we should not underestimate the power of natural habits as part of the problem/solution/intervention.

How did you hold the balance this month? At the end of this month, looking back, what counterweights did you use in your busy life? The summer has been busy — work, kids' stuff and a lovely summer holiday. But balancing all of that is tricky. Do I get it right? No — but I do try to be intentional. I commit to too many things... In particular, my balance was messed up this month on AI use/coding time (not a surprise really...) and is the subject of another field note in preparation. I've called it prompt tilt – tilt being a poker term. Iterating with AI can be very rewarding but addictive.

Reply

Hit reply and tell me which of the three pillar areas you came for — creativity, knowledge or wellbeing. Any other thoughts are welcome — perhaps something you'd like to see as a guide, field note or app. It comes straight to me, I read every one, and it'll help me know where to lean next.

Sources this month (as covered in this month's science press — publication dates vary): UK Biobank sleep study, Nature · Scottish woodland study · Estonian Biobank microbiome study · AI items via the weekly AI Plain Signal and other monthly searches.

Many thanks for reading this month. Made by humans & AI
 
Tene Aequilibrium
Hold the balance — Neill
The Sworn Stave Collective
swornstavecollective.com
Issue 02 · Aug 2026
Unsubscribe
Don't miss what's next. Subscribe to Sworn Stave Collective - The Monthly Balance:
← Newer Issue 03 Older → Issue 01 — the first letter
Powered by Buttondown, the easiest way to start and grow your newsletter.