Nobody is coming. There is no course, no corporate rollout, no two-day workshop that will do this for you. I have spent nearly thirty years in civil engineering and transport planning, on several continents, and I can tell you that the single biggest determinant of whether someone my age gets any value out of AI is not intelligence, technical background or budget. It is the willingness to look stupid for about six weeks.

I looked stupid for about six weeks.

Stage one: writing wishes

My first prompts were wishes. “Summarise this report.” “Write me a market entry strategy.” “What are the benchmarks for metro construction costs?”

I got mush. Confident, well-structured, plausible mush — the sort of thing a very keen graduate produces on their first Friday when they have not asked a single question all week. And like most people, I drew exactly the wrong conclusion. I decided the tool was overrated.

It wasn’t overrated. I was under-briefing it.

Think about how you actually delegate. You do not walk up to a new starter and say “do the rates benchmarking.” You give them the contract. You tell them the client, the currency, the seniority tiers, the format the PMO expects, the three things the last submission got wrong. You tell them what good looks like. Then you leave them alone.

I was doing none of that. I was typing one sentence into a box and expecting a deliverable.

Stage two: context is the whole job

The turn came when I stopped writing prompts and started writing context.

Markdown files. Plain text, structured with headings, bullets, tables. Nothing clever. One file describing the project — the site parameters, the institutional landscape, the standards that apply, the terminology. One file describing how I want work produced — Word documents with source citations and confidence ratings, no unsourced assertions dressed as fact, professional judgement flagged as professional judgement. One file holding what we have already established, so I never have to explain it twice.

That is it. That is the whole trick. Markdown files that travel with the work.

The difference is not incremental. It is the difference between a stranger and a colleague of three years. When I open a benchmarking task now, the model already knows that unit cost rates are in SAR, that the deliverable is a colour-coded workbook with one firm per tab, that the naming convention is fixed, that a triangulated estimate must carry an explicit confidence caveat. I do not restate it. It is written down.

Write the context down. Refuse to repeat yourself. Every sentence you type twice is a file you should have written once. If you’re already there, maybe start thinking about creating skills so that you can streamline your processes in the future.

What it actually does

Three examples from the last year, all real, none of them toys.

Benchmarking. Infrastructure CAPEX work using actual outturn costs as the primary comparator, cross-referenced against the NYU Marron Transit Costs Project and the Flyvbjerg cost-overrun literature. Actual detailed comparisons of publicly available, exemplary mobility and transport hubs and pulling out details that are pertinent to my work, and that a human would normally miss from the sheer amount of data. Work that used to be a fortnight of a junior’s life, done properly in an afternoon, with the sourcing hierarchy visible so I can audit every number.

Go-to-market. Pitch narrative and market sizing for NAYRobotics, my warehouse automation venture. A Gulf-focused market entry due diligence pack for a Madrid consultancy — market conditions, competitive landscape, MISA licensing, Saudization, financial estimates, twenty-three slides and a full Word appendix. Drafted with me, not for me.

The virtual PA. Inbox and diary. Email triaged and drafted, meetings scheduled, documents filed to Drive, the CRM kept current. Every outbound draft is flagged DO NOT SEND until I have read it myself. That rule is not negotiable and never will be.

Where it will burn you

It lies. Beautifully.

It will invent a citation with a plausible author, a plausible journal and a plausible year, and it will do it in the same even tone it uses for the things that are true. It will produce a number that is directionally sensible and factually fabricated. If you cannot tell the difference, it will eat you.

This is precisely where thirty years is worth something. I know what a plausible day rate looks like for a Grade 3 engineer in the Gulf. I know when a cost-per-kilometre figure has drifted into fantasy. I know which standards actually apply and which ones the model has hallucinated into relevance. The judgement is the scarce input. It always was.

So verify. Demand sources. Make it separate verified citation from professional judgement, explicitly, in every document. Then check the citations anyway. Create feedback loops like I’ve done, and if you’re really getting competent with your prompt engineering, you can even create independent AI checkers with explicit checking criteria (like a really good QA/QC person).

The part nobody says out loud

The advantage in this is not with the twenty-five-year-olds. It is with us.

The judgement is the scarce input. It always was.

They can prompt faster. Fine. But prompting is not the point — the prompt is the least valuable part of the entire exercise. The valuable part is knowing what question to ask, what a good answer looks like, and when you are being told something that is confidently wrong. That takes decades. You already have the decades.

What you do not have is an excuse.

Start with one real piece of work. Not a toy, not a demo — something you actually have to deliver this month. Write down the context first. Give it the brief you would give a competent new starter, then set the guardrails to ensure that your data isn’t misused or co-opted to train the AI provider’s models. Read the output like a reviewer, not an audience.

Then do it again tomorrow.

Show up, do the work, go home. Same as it ever was. The tools changed. The hard graft didn’t.

It feels free and I think that’s what makes it dangerous. The project award goes to someone else, the flight gets cancelled, the pilot data comes back ugly — and before you have processed a single fact, you have already spent forty seconds narrating why this is unfair. Multiply that by a career. Nearly thirty years in civil engineering and transport planning has taught me that the complaint is never the problem. The complaint is the price you pay before you start solving the problem.

Jocko Willink has the best fix I have found, and it is one word.

Good.

Something goes wrong, and you say it out loud. I didn’t win the framework contract. Good. Now I find out what the winner priced that I didn’t. Robot vendor pushed the delivery back a quarter. Good. More time to nail down the brownfield site survey for the client’s warehouse before anything lands on a warehouse floor. Client killed the scheme. Good. I was building it on an assumption I never tested.

The trick is that you have to finish the sentence. “Good” on its own is stoicism as a bumper sticker. “Good, because —” is a demand. It forces your brain out of the grievance loop and into search mode, and your brain, being lazy and obedient, will actually go and look for an answer. Sometimes the answer is thin. Take it anyway. A thin answer beats forty seconds of narration.

Refuse to complain first. Complain later if you must, on your own time, with the work already moving.

The second habit is harder, because it attacks something men our age quietly treasure: competence.

I moved to Riyadh in July 2021. I now run a mobility practice by day and a warehouse robotics venture, NAYRobotics, alongside it. I am learning Najdi Arabic. I am learning Mandarin. I spend a good part of my week being visibly, measurably worse at things than people twenty years my junior.

The first year of that was awful. Not hard — awful. Because at fifty-something you have built an identity out of being the person in the room who knows. Being a beginner again feels like a demotion so you avoid it. You stay in the sectors and comfort zone where you are already the expert, you delegate the unfamiliar, and you tell yourself it is a use-of-time argument. It isn’t. It is ego management dressed up as prioritisation.

Here is the reframe that worked. Frustration when learning something new is not evidence that you are too old. It is the sensation of the thing working. It is lactate building up in muscles that you haven’t been using. You would not stop a hill climb because your legs hurt — the hurt is the point of the hill. Mental effort is identical, and it is the only signal you get that you are actually building something rather than rehearsing what you already own.

So chase it. Look forward to it. Go and find the meeting where you are the least knowledgeable person present and sit in it deliberately.

One caveat, because I dislike self-help delivered as gospel: the academic literature on “growth mindset” is far weaker than the TED-talk version suggests, and large meta-analyses have found the effects small and inconsistently replicated. I am not making a scientific claim. I am making a practical one, and the practical one is unarguable — the man who is willing to be a beginner has more options at sixty than the man who isn’t.

The third point is the payoff, and it is the one only a middle-aged man gets.

Frustration when learning something new is not evidence that you are too old. It is the sensation of the thing working.

Thirty years of experience is not thirty years of knowing more transport planning. Its real value is the pattern library and joining the dots across other disciplines and sectors.

When I look at brownfield warehouse automation, I am not seeing robotics. I am seeing a network problem I have solved before — flow, conflict points, throughput at the constrained node, the difference between designed capacity and operated capacity. At its essence, a warehouse shuttle system is a metro system with a smaller loading gauge. When I read a robotics-as-a-service contract, I am reading an availability-payment concession with different nouns. Rate benchmarking, technical due diligence, dialect learning — every one of them is pattern-matching under uncertainty.

Young people learn faster. They cannot do this. They haven’t got the second sector to connect it to.

That is the whole argument. The complaint reflex costs you the first minute. The competence reflex costs you the new sector. Kill both, and thirty years of scar tissue stops being a record of what you have done and starts being an instrument for reading things you have never seen before.

Good.

It’s for the above reasons that I personally think that the sardine fast is the workaround. In case you’ve never heard of it, the concept behind it is pretty straightforward: three to five days, a few tins of the oily fish, water, black coffee. You stay in a deep calorie deficit, but you’re not staring at the ceiling at 3am wondering whether you’re catabolising the deltoids you spent years building so that you look like the Ultimate Warrior (yes, I’m from that era). It’s cheap. It’s simple. It takes no willpower to prepare. And it has become the fasting protocol of choice for people who don’t actually enjoy fasting.

Having said all that, there is also almost no direct evidence behind it — but I know how a sardine fast makes me feel personally. Let’s go through what the science actually says.

The protocol

Dominic D’Agostino — the USF neuroscientist who has spent his career on ketone metabolism — described his version on The Tim Ferriss Show in September 2025: roughly one to two cans a day, for up to a week, repeated monthly, with magnesium and vitamin C added.^1, 2 He reaches for it when he feels something coming on — illness, brain fog, a flare-up. Other practitioners run it heavier, four or five cans a day, eating to satiety rather than counting.

Know what you’re actually consuming. One 3.75 oz tin of Atlantic sardines in oil, drained, bones in, gives you roughly 191 calories, 22.6 g of protein, 0.89 g of EPA and DHA, 351 mg of calcium, 8.2 mcg of B12 and 48.5 mcg of selenium. Two tins is under 400 calories. Five tins is about 955 calories and 113 g of protein.³

That spread matters more than anything else in this article. Hold that thought.

Where it came from

Fred Hatfield — “Dr. Squat,” the first man to squat over 1,000 lb in competition — was diagnosed with metastatic skeletal cancer in 2012 and told by three doctors he had three months. He went ketogenic, ran repeated sardine fasts, and the cancer went into remission. He died in May 2017. Of sudden heart failure, at 74. Not of cancer.^4, 5

It’s a hell of a story. It is also a single uncontrolled self-reported case, with concurrent treatment, no imaging trail in the literature, and every survivorship bias you can name. Anecdote generates hypotheses. It does not test them. Anyone who sells you the Hatfield story as proof is selling you something.

You stay in a deep calorie deficit, but you’re not staring at the ceiling at 3am wondering whether you’re catabolising the deltoids you spent years building so that you look like the Ultimate Warrior

The strongest argument against calling this a fast

Here’s the objection nobody in the sardine-fast community wants to engage with.

The mechanistic pitch is that this mimics fasting — low insulin, low mTOR, AMPK up, autophagy on. The closest properly tested cousin is Valter Longo’s fasting-mimicking diet. In the randomised trial, 100 participants ran 750–1,100 calories a day for five days a month, three cycles. Results: about 6 lb lost, 1–2 inches off the waist, systolic pressure down 4.5 mmHg, diastolic down 3.1, IGF-1 down into a lower-risk range, fasting glucose normalised in the at-risk group, lean mass preserved.⁶

But Longo’s diet is deliberately low protein. That’s the active ingredient in the IGF-1 and mTOR story. Eat five tins of sardines and you’ve taken in 113 g of protein — more than most people eat on a normal day.³ You have not mimicked a fast. You’ve run a protein-sparing modified fast, which is a legitimate, well-characterised intervention in its own right, but it is not the same lever and it will not do the same things.

So pick your objective. Want the IGF-1 and autophagy signal? Stay at one or two tins and accept that you’ll be hungry. Want the deficit with your muscle intact and your training tolerable? Eat four or five and stop claiming autophagy benefits you haven’t earned. You cannot have both at once. Anyone telling you otherwise is hand-waving. I’ve personally gone with the second route in order to preserve my muscle mass but it’s entirely up to you, depending on your own objectives.

What sardines earn on their own merits

Strip the fasting narrative away and the fish still stands up.

The best trial is Spanish: 152 adults over 65 with prediabetes, randomised, twelve months. Both groups got a diabetes-prevention nutrition programme; one group added 200 g of sardines a week — two tins. The proportion in the high-risk-of-diabetes category fell from 37 per cent to 8 per cent in the sardine group, against 27 per cent to 22 per cent in the controls. HOMA-IR down, HDL up, adiponectin up, triglycerides down, blood pressure down. Twelve months, real endpoints, control group.⁷ That’s the good stuff.

Then there’s the micronutrient load. Bones in, you’re getting calcium and vitamin D from a whole food. B12 at several times the daily requirement per tin. Selenium. And an EPA/DHA dose that most people never reach from supplements.³

Mercury is the objection people raise, and it’s the wrong one. Sardines are small, short-lived and low on the food chain. FDA sampling puts them at 0.013 ppm — against 0.126 for canned light tuna, 0.350 for albacore, and 0.995 for swordfish. Roughly seventy-five times cleaner than swordfish.⁸ Eat the sardines like me.

Who should not do this

Gout. This is the real risk and it is not theoretical. Sardines are among the most purine-dense foods available, and in 47,150 men followed for 12 years, top-quintile seafood intake carried a relative risk of gout of 1.51. Concentrating five days of purines into a body that already struggles with uric acid is a bad experiment with a painful endpoint.⁹

Chronic kidney disease. Same logic, worse consequences.

Anyone on insulin or sulfonylureas. A sudden deficit of a thousand calories or more, combined with hypoglycaemic medication, is a situation that needs a doctor involved, not a blog.

Anyone on blood-pressure medication or diuretics. Watch the sodium, which varies wildly by brand — drained tins run around 280 mg, some brands considerably more.³

Anyone with a history of disordered eating. Restrictive protocols with rules and a countdown are not a neutral tool for everyone.

If you’re going to run it

Buy tins with bones in, in olive oil or water. Drain them. Salt your food and take magnesium and potassium, because ketosis dumps electrolytes and most people who feel terrible on day two are simply low on sodium. Drink more water than feels necessary. If you like data, measure glucose and ketones and track your GKI — glucose in mg/dL divided by 18, divided by ketones in mmol/L; under 3 is deep territory.¹⁰ Train light. Walk instead of lifting heavy.

And break it with real food, or a bone broth as a transition back into heavier foods. A five-day protocol undone by a Friday night blowout is a five-day protocol wasted.

The verdict

The sardine fast has never been tested as the sardine fast. Not once. What it has is two well-supported limbs — short-term energy restriction, and sardines themselves — bolted together by a plausible mechanism and one dramatic anecdote.

That makes it a reasonable, cheap, low-risk thing for a healthy person to try for three to five days. It does not make it medicine. Don’t let anyone sell it to you as cancer treatment, and don’t let the story of one very strong man carry more weight than it can hold.

Confidence in the sardine evidence: high. Confidence in short-term energy restriction: moderate to high. Confidence in the sardine fast specifically doing anything beyond those two things: low. Eat the fish anyway — your skin and bones will also thank you.

Sources

Evidence tier is flagged on each entry. Nothing below is a trial of the sardine fast itself, because no such trial exists.

1. Ferriss, T. (host) and D’Agostino, D. “Dr. Dominic D’Agostino — All Things Ketones, How to Protect the Brain and Boost Cognition, Sardine Fasting, Diet Rules, Revisiting Metformin and Melatonin, and More.” The Tim Ferriss Show, episode 825, 3 September 2025. — Tier: expert opinion, unreviewed.
2. Podcast Notes. “All Things Ketones, How to Boost Cognition, Sardine Fasting, Diet Rules, & More — Dr. Dom D’Agostino on The Tim Ferriss Show Ep. 825.” 2025. — Tier: third-party summary of [1]. Protocol figures cross-checked against the episode notes.
3. U.S. Department of Agriculture, FoodData Central. “Fish, sardine, Atlantic, canned in oil, drained solids with bone,” FDC ID 175139. Values per 92 g (3.75 oz) tin. — Tier: reference database.
4. English, N. “Fred ‘Dr. Squat’ Hatfield Passes Away at 74.” BarBend, 15 May 2017. — Tier: journalism reporting a self-reported case. Records the 2012 diagnosis and the three-month prognosis.
5. “Frederick Hatfield.” Wikipedia. — Tier: tertiary, weakest source cited here. Reported causes of death differ between [4] and [5]; both agree the cause was not the cancer.
6. Wei, M., Brandhorst, S., Shelehchi, M., et al. “Fasting-mimicking diet and markers/risk factors for aging, diabetes, cancer, and cardiovascular disease.” Science Translational Medicine, 2017;9(377):eaai8700. doi:10.1126/scitranslmed.aai8700. Trial summary. — Tier: randomised controlled trial (n=100 randomised, 71 completers).
7. Díaz-Rizzolo, D.A., Serra, A., Colungo, C., Sala-Vila, A., Sisó-Almirall, A., Gomis, R. “Type 2 diabetes preventive effects with a 12-months sardine-enriched diet in elderly population with prediabetes: an interventional, randomized and controlled trial.” Clinical Nutrition, 2021;40(5):2587–2598. doi:10.1016/j.clnu.2021.03.014. PMID 33932804. Summary. — Tier: randomised controlled trial, 12 months, n=152. The strongest evidence in this article.
8. U.S. Food and Drug Administration. “Mercury Levels in Commercial Fish and Shellfish (1990–2012).” — Tier: regulatory sampling data.
9. Choi, H.K., Atkinson, K., Karlson, E.W., Willett, W., Curhan, G. “Purine-rich foods, dairy and protein intake, and the risk of gout in men.” New England Journal of Medicine, 2004;350(11):1093–1103. doi:10.1056/NEJMoa035700. PMID 15014182. — Tier: prospective cohort, 47,150 men, 12 years, 730 incident cases.
10. Meidenbauer, J.J., Mukherjee, P., Seyfried, T.N. “The glucose ketone index calculator: a simple tool to monitor therapeutic efficacy for metabolic management of brain cancer.” Nutrition & Metabolism, 2015;12:12. doi:10.1186/s12986-015-0009-2. — Tier: methodology paper. The GKI range quoted was derived in a brain-cancer context, not validated for healthy people running short fasts.

That is a mistake. Creatine monohydrate is not a bodybuilding supplement but a naturally occurring chemical compound. It is an energy-buffering system, and your brain works it as hard as your quadriceps do.

The mechanism is simple. Every cell runs on ATP. When demand spikes — a heavy set, a hard problem, a night without sleep — ATP gets spent faster than it can be rebuilt. Phosphocreatine is the reserve tank that regenerates it on the spot. Load more creatine into the tissue and you extend the window before performance falls off a cliff. That is the whole story. Muscle and brain, same currency.

The brain case is now stronger than the muscle case

This is the part that should change your thinking. It’s certainly changed mine, especially with my inter-continental trips across time zones. At my age, I’ve definitely started feeling the effect of doing so, not just physically but also mentally (screw you brain fog!!).

It boosts ATP production to fuel your muscles, and emerging research shows it also sharpens cognitive function, especially under stress and fatigue.

In 2024, researchers at Forschungszentrum Jülich gave fifteen volunteers a single high dose of creatine — 0.35 g/kg, roughly 26 g for a 75 kg man — during 21 hours of sub-total sleep deprivation. Cognitive performance improved from three and a half hours after the dose and the effect held or grew out to nine hours. Word memory rose about 10 per cent. Processing speed on language, logic and numeric tasks improved by 16 to 29 per cent. Measured fatigue fell. They watched the brain’s high-energy phosphates shift on magnetic resonance spectroscopy while it happened.

Read the caveats before you get excited. Fifteen people. Average age 23. A single dose, never replicated in men in their fifties. And a dose that size in one hit will send a fair number of people straight to the bathroom. This is a proof of mechanism, not a protocol.

The wider literature is where the age pattern gets interesting. A 2023 meta-analysis in Nutrition Reviews pooled memory trials in healthy adults and found a modest overall effect (SMD 0.29). Split by age, it was 0.88 in adults aged 66 to 76 and essentially zero — 0.03 — in the young. Creatine does very little for a well-rested 25-year-old’s memory. It does something measurable for an older brain under load. Which is exactly the brain you and I are now relying on.

Then 2025. A pilot trial at the University of Kansas gave 20 g a day for eight weeks to twenty patients with Alzheimer’s disease. Brain creatine rose 11 per cent on spectroscopy — hard proof the stuff crosses into brain tissue at that dose — and scores improved on the global cognition and fluid reasoning composites, plus the individual working-memory, reading and attention tests. I’m not quite ready to up my own dosage from the 5 g a day to 20 g but nevertheless, I’m keeping an eye on the research because this is a signal that deserves a proper randomised trial.

Muscle, and what the numbers actually say

The physical case is settled. It is also smaller than the marketing implies.

A 2025 meta-analysis of eight randomised trials in adults over fifty found creatine plus resistance training beat training alone for lean tissue mass (SMD 0.27) and lower-limb strength (SMD 0.29). Real, statistically significant, modest. Upper-body strength did not reach significance in the pooled analysis.

Note the structure of that sentence. Creatine plus resistance training. It is a multiplier on work you are already doing, not a substitute for it. Take creatine and sit still and you will gain water weight and nothing else. Train hard and take creatine and you get maybe 10 to 15 per cent more out of the same programme.

That sounds unglamorous. It compounds. Sarcopenia — the slow default loss of muscle that begins in your fourth decade and accelerates in your sixth — is a twenty-year problem, and a 10 per cent edge applied every year for twenty years is not a rounding error. It is the difference between carrying your own bags at seventy and not. Just one thing worth pointing out: my weight has hovered between 62 and 64 kilograms for the last two decades but taking creatine has pushed my steady-state weight to 67 to 69 kilograms. I can’t quite separate creatine ingestion from my starting to do more resistance training but reliable research all point to water weight accumulation so that’s the story I’m going with.

Where I would push back on the standard pitch

Two claims you will hear repeated on every podcast do not survive contact with the evidence. I am including them because you will hear them anyway, and you should know which parts of your own argument are load-bearing.

Bone density has been oversold. Chilibeck’s 2015 trial showed creatine slowing femoral neck bone loss in postmenopausal women, and it launched a thousand blog posts. The two-year follow-up randomised trial published in 2023 found no benefit to bone mineral density. Forbes’ meta-analysis found none either. In 2025, Osteoporosis International ran a letter titled, without ambiguity, “Insufficient evidence to recommend creatine supplementation as a therapeutic option for osteoporosis.” Bone geometry and bone quality signals are more encouraging than the density numbers, but if you are taking creatine for your skeleton you are running on hope, not data. Take it for muscle. Accept that stronger muscle pulling on bone is the plausible route to a stronger skeleton — an indirect one, and a slow one.

Depression is thinner still. The 2025 meta-analysis in the British Journal of Nutrition pooled eleven trials and 1,093 participants. The effect was SMD −0.34, about 2.2 points on the Hamilton Depression Rating Scale, against a clinically meaningful threshold of 3.0. The evidence was graded very low quality. Heterogeneity ran at 71 per cent. Correcting for publication bias shrank the effect further. Something may be there. It is not a mood treatment, and nobody should be swapping care for a scoop of powder.

Cardiovascular markers are early but worth watching. A 2025 pilot in twelve sedentary adults aged 50 to 64 found flow-mediated dilation improved from 7.7 to 8.9 per cent, microvascular reperfusion improved, and fasting glucose and triglycerides fell over four weeks. Twelve people. No control group. File it under “promising” and check back in three years.

What to actually do

Buy creatine monohydrate. Not HCl, not buffered, not the liquid, not the version with a stack of amino acids and a marketing budget. Monohydrate is the form used in essentially every trial cited above, and it is the cheapest thing on the shelf. I personally consume the unflavoured kind because I find the artificial berry flavours completely vile, and they’re much easier to add to smoothies and protein shakes.

Five grams a day. Every day, training days and rest days alike. Four years in, that is still my dose, and nothing I have read since has given me a reason to change it. Timing is irrelevant — you are filling a reservoir over weeks, not spiking a hormone, so take it whenever you will actually remember to take it. Skip the loading phase unless you want saturation in a week rather than a month; if you do load, 20 g split into four doses for five days, then drop to five.

Expect a kilo or two (five in my case!) on the scales in the first fortnight. Like I mentioned above, that is intramuscular water. It is not fat, and it is arguably the point.

The safety file is about as clean as nutrition science gets: the International Society of Sports Nutrition position stand covers doses up to 30 g a day for five years in healthy people with no evidence of renal harm. One practical warning — creatine raises serum creatinine, which is the marker your doctor uses to estimate kidney function. Tell them you are taking it before somebody panics at a lab result and sends you for a scan you do not need.

Roughly a quarter of people are non-responders, usually those already near saturation from a heavy red-meat diet. Give it eight weeks and judge honestly.

That is the whole protocol. Five grams, daily, indefinitely, alongside the training that makes it work. The evidence for the brain is thinner than the evidence for the muscle, and the evidence for bone barely exists at all. Take it anyway. When the downside is twenty quid a month and a heavier bathroom scale, and the upside is a hedge against the two things that genuinely take men apart after fifty — losing muscle and losing sharpness — the arithmetic is not close.

Sources

1. Gordji-Nejad A, et al. “Single dose creatine improves cognitive performance and induces changes in cerebral high energy phosphates during sleep deprivation.” Scientific Reports, 2024. doi:10.1038/s41598-024-54249-9
2. Prokopidis K, et al. “Effects of creatine supplementation on memory in healthy individuals: a systematic review and meta-analysis of randomized controlled trials.” Nutrition Reviews, 2023;81(4):416–427.
3. Smith AN, et al. “Creatine monohydrate pilot in Alzheimer’s: feasibility, brain creatine, and cognition.” Alzheimer’s & Dementia: TRCI, 2025. doi:10.1002/trc2.70101 (trial NCT05383833)
4. Systematic review and meta-analysis: creatine with resistance training on muscular strength and lean tissue mass in the aged. European Review of Aging and Physical Activity, 2025. doi:10.1186/s11556-025-00392-9
5. Chilibeck PD, et al. “Effects of creatine and resistance training on bone health in postmenopausal women.” Med Sci Sports Exerc, 2015;47:1587–1595. — and the 2-year follow-up RCT, Med Sci Sports Exerc, 2023 (PMID 37144634).
6. Machado M. “Insufficient evidence to recommend creatine supplementation as a therapeutic option for osteoporosis.” Osteoporosis International, 2025. doi:10.1007/s00198-025-07713-9
7. “Creatine supplementation for treating symptoms of depression: a systematic review and meta-analysis.” British Journal of Nutrition, 2025 (PMID 41189312).
8. “Effect of creatine monohydrate supplementation on macro- and microvascular endothelial function in older adults: a pilot study.” Nutrients, 2025;17(1):58.
9. Kreider RB, et al. “ISSN position stand: safety and efficacy of creatine supplementation in exercise, sport, and medicine.” J Int Soc Sports Nutr, 2017;14:18.

Angel investing is the opposite of that. You meet the founder. You read the science. You decide.

I want to be precise about why I do it, because the reasons that get trotted out in most articles on this subject — returns, excitement, “giving back” — are either wrong or lazy. There are three reasons that hold up. Here they are.

One: Somebody has to go first

Every company you admire was, at one point, two or three brilliant (and usually young) people with an idea and no money.

That gap — between the idea working on a bench and the idea working in the world — is where good technology goes to die. Institutional capital will not touch it. Venture funds have minimum cheque sizes and mandates that make pre-seed uneconomic for them; grant bodies move at the speed of committees with a million bureaucratic hoops to jump through; banks want assets to secure against, and a founding team has none. So the ideas that deserve to exist often sit in a drawer, not because they were wrong, but because nobody was willing to be first.

The angel is the person willing to be first.

Now don’t get me wrong — this is not charity. It is the single highest-leverage moment in the entire life of a company, and it is the only moment at which a private individual with a modest cheque can materially alter what happens next. A hundred million pounds arriving at Series C changes a company’s trajectory. A five-figure cheque arriving at pre-seed changes whether there is a company at all — therefore the matter is of an existential nature.

And the capital is the smaller half of it. Founders at this stage are not short of enthusiasm. They are short of people who will tell them the truth. Being the first outside investor buys you a seat at the table where you can say that assumption is wrong, that customers will not pay for a particular product or feature, or that you are about to sign something you will regret. Do that once, properly, and you have earned your entry price before the company has sold anything.

Two: You get to pick who you back — and what they stand for

This is the part that most institutional investing cannot give you at any price.

When you buy a fund, you buy everything in it. You buy the companies you admire and the ones you would cross the street to avoid, bundled together and sold to you as diversification. You have no vote on ethos. You have no vote on anything.

As an angel, you have exactly one vote, and it counts for everything: yes or no.

I use mine on companies whose work sits close to something I actually care about. Food security is the clearest example. It is easy to treat that as an abstraction from a comfortable seat in Riyadh or London. It is not an abstraction. It is arithmetic about protein, land and water, and the arithmetic is not going the right way.

Which brings me to Algiplex.

Algiplex is a UK company working on the productivity and resilience of commercial algae farms. Their approach is unusual and, I think, correct: rather than engineering the algae itself, they engineer the microbial ecosystem around it — evolving symbiotic microbial communities, selecting the best variants, and deploying them as plug-and-play modules that keep physical separation while still delivering the benefit. In lab trials they have reported productivity gains of up to 8.5 times, alongside protection against the contamination events that wipe out crops.

Now understand why that matters, because the connection to food security is one or two steps removed and that is precisely the point.

Spirulina is between 55 and 70 per cent protein by dry weight. Per unit of land it yields roughly twenty times the protein of soybeans, forty times that of corn, and over two hundred times that of beef. It uses about a quarter of the water that soy protein needs and two per cent of what beef protein needs. It does not require fertile land — it will tolerate saline conditions that would kill a conventional crop. The FAO has documented cases in Mexico, Togo and China where roughly ten grams a day produced rapid recovery in malnourished children. Those are the FAO’s figures, not a marketing brochure’s. Did I also mention that it can become a cash crop for farmers in these countries too?

So the technology exists. The nutrition works. Why has it not eaten the world?

Because algae farms are biologically unstable. Crops crash. Contamination arrives, the pond turns, and months of production disappear in a week. That instability is what keeps yields down, costs up, and the product priced as a wellness supplement for the affluent rather than a protein source for the people who need protein most. Fix the resilience and you fix the unit economics. Fix the unit economics and the protein becomes cheap. Cheap protein grown on land nobody wants, using water nobody is fighting over, is not a lifestyle product. It is infrastructure.

That is the chain. Microbial modules, then stable farms, then affordable biomass, then feed and food, then people who could not previously afford the protein. Two steps, maybe three. Unglamorous every one of them.

The unglamorous step is usually the one that decides the outcome.

Three: Thirty years has to be worth something

I have spent nearly three decades building and assessing physical systems — planning them, financing them, tearing apart other people’s business cases in due diligence, and watching what happens when a beautiful design meets an operating environment that does not care.

That experience has one recurring lesson. Things do not fail in the laboratory. They fail at scale, in the field, in the rain, when the operator is tired, assets are not maintained, and the maintenance budget has been cut. The gap between “it works” and “it works reliably, repeatably, at cost, for years” is where almost all the value and almost all the risk actually live.

The unglamorous step is usually the one that decides the outcome.

That is not a transferable insight. That is the insight, and it applies to an algae farm exactly as it applies to a metro system.

So when I look at an early-stage company, I am not trying to out-guess a specialist on the science. I am asking the questions I have spent thirty years learning to ask. What breaks first? What does this cost when you have built a hundred of them rather than one? Who operates it, and what happens when they get it wrong? What is the failure mode nobody has modelled? Founders rarely resent those questions. They have usually been waiting for someone to ask them.

Angel investing is the only vehicle I know of that converts a career’s worth of pattern recognition into an actual position. Deploy it into sectors you understand, at the stage where judgement is worth more than capital, and you are no longer a passive allocator. You are useful.

The part nobody puts in the headline

Be clear-eyed. The most credible UK dataset we have — Robert Wiltbank’s study for NESTA and the British Business Angels Association, covering 1,080 investments by 158 angels — found an overall return of 2.2 times capital over an average of 3.6 years. That is the good news. The bad news sits in the same report: fifty-six per cent of exits failed to return the capital invested, and nine per cent of investments generated nearly eighty per cent of all the positive cash flow.

Read that again. More than half of everything you back will lose money, and your entire result will be decided by one or two positions. This is not an asset class. It is a portfolio of lottery tickets where diligence improves your odds and nothing guarantees them. Invest what you can lose, and mean it. There are however some safeguards and tax relief, especially for EIS-registered startups.

I invest anyway in spite of the risks. Not despite that distribution — because of what the winners do when they land. For me, this is an asymmetric bet that has a high chance of failure, but the upside and positive impacts if successful makes it worth my time, effort and money.

You cannot buy that exposure through a platform. You have to choose it, one founder at a time, and be willing to be wrong in public.

Choose well. Then be useful.

Sources

1. Algiplex — company website, algiplex.co.uk (technology description and reported lab productivity figures).
2. Habib, M.A.B., Parvin, M., Huntington, T.C. and Hasan, M.R. (2008), A Review on Culture, Production and Use of Spirulina as Food for Humans and Feeds for Domestic Animals and Fish, FAO Fisheries and Aquaculture Circular No. 1034 (protein content, land and water comparisons, malnutrition case studies).
3. Wiltbank, R.E. (2009), Siding with the Angels: Business angel investing — promising outcomes and effective strategies, NESTA / British Business Angels Association (returns, holding period, loss rate).

It took me a while to learn the difference. For about five years, I ran my diet close to zero carbs, and it did exactly what I asked of it. I got lean and even managed to get my body fat percentage to about 11 per cent. No bread, no rice, no potatoes. Protein, fat, green vegetables (usually broccoli, kale and cavolo nero when I can get it). Repeat.

It worked. I won’t pretend otherwise.

Then the goal changed, and the diet had to change with it.

From lean to long-lived

Getting lean is a body composition goal. Living longer (and staying useful while I do it) is a different job, and it needs a different toolbox.

The number that changed my thinking was VO2max: the maximum amount of oxygen your body can take in and use during hard exercise. It is about as close as we have to a single measure of the engine you’re running on. And the data linking it to how long you live are brutal.

In 2018, a team at the Cleveland Clinic followed 122,007 patients who had done a treadmill fitness test, for a median of 8.4 years. People in the elite fitness bracket had roughly 80 per cent lower risk of death than those in the lowest bracket. Being in the bottom group carried a bigger mortality risk than smoking, diabetes or coronary artery disease. And there was no ceiling. Elite performers still beat the merely “high” fit group. More fitness, more benefit, all the way up.

(It’s an observational study, so it shows association, not proof of cause. But the effect is huge, it’s consistent with decades of other data, and I’m not waiting for a perfect trial before I act on it.)

So I rebuilt my training around cardio, following Iñigo San Millán’s polarised model. Most of the week is Zone 2: easy, conversational running on Monday, Friday and a long run on Sunday. That builds the base. Once a week, on Wednesday, I go hard: Zone 5 intervals. Sometimes a fartlek session replaces one of the Zone 2 runs. What I don’t do is spend time grinding away in the grey middle of Zone 3, which is too hard to build the base and too easy to raise the ceiling.

The hard day is where the ceiling goes up. In a classic 2007 study, 40 moderately trained men did eight weeks of training with the total workload matched across groups. Long, slow running and threshold running didn’t move VO2max significantly. Four-minute intervals at 90–95 per cent of max heart rate improved it by 7.2 per cent.

Here’s the catch though. High intensity runs on carbohydrate and I found out the hard way when my body just basically refused to do more work as it was running on fumes.

Fat is a fantastic fuel for easy work. It’s a lousy fuel when you’re going flat out. San Millán’s own research with George Brooks measured this directly: as intensity climbs, everyone shifts from burning mostly fat to burning mostly carbohydrate. Professional cyclists held on to fat until around 75 per cent of their maximum power. Moderately active people crossed over at around 55 per cent. By the time you’re doing VO2max intervals, you’re running almost entirely on glycogen, the carbohydrate stored in your muscles and liver.

Show up with an empty tank and the intervals go flat. You can’t hit the pace. You can’t hold the effort. The session that was supposed to raise the ceiling becomes another mediocre Zone 3 slog.

The best evidence on this comes from Louise Burke’s work at the Australian Institute of Sport. She put elite race walkers on either a high-carb diet or a ketogenic, low-carb high-fat diet during three weeks of hard training. The low-carb group more than doubled their fat burning. Sounds great. It wasn’t. Their oxygen cost at race pace went up (they got less economical) and they got none of the performance benefit from the training block. The high-carb group raced 10km 6.6 per cent faster. The low-carb group didn’t improve at all. Burke’s team repeated the study in 2020 with 26 walkers. Same result: high-carb 4.8 per cent faster, low-carb 2.3 per cent slower.

I’m not an elite runner or athlete and neither are most of the people reading this. The physiology is the same though. Hard efforts burn carbohydrate, and if you want quality hard efforts, you have to put some carbohydrate in the tank.

So the carbs came back. Some, not lots. Enough to fuel the work.

Choose your carbs like you’d choose your tools

Once you accept that you need some carbohydrate, the question becomes which carbohydrate. That’s where most people go wrong. They hear “carbs are OK now” and go straight back to white bread and pastries.

Don’t.

Fat is a fantastic fuel for easy work. It’s a lousy fuel when you’re going flat out.

My base is vegetables. Lots of them. They bring carbohydrate, but they also bring fibre, and fibre is the part that does the heavy lifting for long-term health. The biggest review on this, published in The Lancet in 2019, pooled nearly 250 studies and trials. People eating the most fibre had 15–30 per cent lower rates of death from all causes and from cardiovascular disease, plus less coronary heart disease, stroke, type 2 diabetes and bowel cancer, compared with people eating the least. The sweetspot was 25–29 grams a day. Most people fall well short.

The vegetables also keep the glycaemic index (GI) of my meals low. GI measures how fast a food raises your blood sugar. Low GI means steadier energy and no crash an hour later. Worth knowing though: the same Lancet review found the evidence for fibre much stronger than the evidence for GI on its own. Fibre is the main tool. GI is a useful secondary gauge.

On top of the vegetables, I add the occasional portion of three things:

Sweet potato. Brilliant, but only if you cook it right. This is the one most people get wrong. A 2011 study tested sweet potato cooked different ways. Boiled, it scored a GI of 41–50: low. Roasted, 79–93. Baked, 82–94. Same potato, and the oven takes it from low GI to higher than a lot of white bread. Boil it.

Brown rice. Better than white, mostly because it keeps the bran, so you get more fibre and more minerals. Don’t assume it’s automatically low GI, though. In the University of Sydney GI database, brown rice has tested anywhere from 48 to 75 depending on the variety and how it’s cooked. Keep the portion modest and eat it with vegetables and protein, which slow the whole meal down.

Flatbread. Occasionally. Pick wholemeal over white when you have the choice. And treat it as a side, not the main event. I’m lucky because most of the year, I’m based in Riyadh (Saudi Arabia) so flatbread or khubz al-jamri (flatbread baked in charcoal oven) is easily accessible.

The principle is simple. Vegetables are the foundation. The starchy carbs are the top-up, sized to the training I’m actually doing.

Strictly no empty carbs

This is where I refuse to budge.

No sugary drinks. No sweets. No pastries. No fruit juice pretending to be healthy. None of it.

“Empty” is the right word. Sugar gives you calories and nothing else: no fibre, no micronutrients, nothing that keeps you full. It spikes your blood sugar, drops it again, and leaves you hungrier than before.

The World Health Organization recommends keeping “free sugars” (added sugar, plus the sugar in honey, syrups and fruit juice) below 10 per cent of your daily calories, and ideally below 5 per cent. On a 2,000-calorie day, 5 per cent is about 25 grams. A single 330ml can of regular cola has around 35 grams. One can, and you’ve blown the whole day’s budget before you’ve eaten anything.

The drinks are the worst offenders. You don’t feel full after a sugary drink, so those calories get stacked on top of everything else you eat. They sneak in all day long: the fizzy drink with lunch, the sweetened coffee in the afternoon, the juice with breakfast. Cut them out completely. Not “cut down”. Cut out.

If I’m eating carbohydrate, it has to earn its place. Vegetables earn it with fibre. Sweet potato, brown rice and the occasional flatbread earn it by fuelling my hard sessions. Sugar earns nothing.

The blueprint

Here’s where I’ve landed, and I suspect it’s where a lot of people chasing longevity will end up:

• Low carb is a good tool for getting lean. Use it if that’s your goal.
• Once the goal becomes fitness and longevity, the hard sessions need fuel. High-intensity work runs on carbohydrate. Starve it and you undercut the very training that builds VO2max.
• Build every meal on vegetables. Fibre is the most underrated tool in the box but just remember that not all vegetables are equal.
• Add starchy carbs in moderation, and choose them carefully. Boiled not baked. Brown over white. Portions sized to the work.
• No empty carbs. Ever. No sugary food, no sugary drinks.

None of this is complicated. The hard part is sticking to it week after week when there’s a can of something sweet within arm’s reach.

Carbs are fuel. Buy good fuel. Burn what you buy.

Sources

1. Mandsager K, et al. “Association of cardiorespiratory fitness with long-term mortality among adults undergoing exercise treadmill testing.” JAMA Network Open, 2018;1(6):e183605.
2. Helgerud J, et al. “Aerobic high-intensity intervals improve VO2max more than moderate training.” Medicine & Science in Sports & Exercise, 2007;39(4):665–671.
3. San Millán I, Brooks GA. “Assessment of metabolic flexibility by means of measuring blood lactate, fat, and carbohydrate oxidation responses to exercise in professional endurance athletes and less-fit individuals.” Sports Medicine, 2018;48(2):467–479.
4. Burke LM, et al. “Low carbohydrate, high fat diet impairs exercise economy and negates the performance benefit from intensified training in elite race walkers.” Journal of Physiology, 2017;595(9):2785–2807.
5. Burke LM, et al. “Crisis of confidence averted: impairment of exercise economy and performance in elite race walkers by ketogenic LCHF diet is reproducible.” PLOS ONE, 2020;15(6):e0234027.
6. Reynolds A, et al. “Carbohydrate quality and human health: a series of systematic reviews and meta-analyses.” The Lancet, 2019;393(10170):434–445.
7. Bahado-Singh PS, et al. “Relationship between processing method and the glycemic indices of ten sweet potato cultivars commonly consumed in Jamaica.” Journal of Nutrition and Metabolism, 2011;2011:584832.
8. University of Sydney Glycemic Index Research Service database, rice values. Summary.
9. World Health Organization. Guideline: Sugars intake for adults and children. Geneva: WHO, 2015.

That is older than agriculture. Older than writing, older than the wheel, older than the first walled town. Researchers from Griffith University and Indonesia’s BRIN dated the Leang Karampuang panel using laser-ablation uranium-series analysis and published the result in Nature in July 2024. It is currently the oldest known narrative art on earth.

Think about what that means. Before we could farm, before we could write anything down, we could already make a mark on a wall that meant something to another person. Art is not a luxury that civilisation produced once it had spare capacity. Art came first. It is the oldest surviving evidence that there is a mind behind the eyes.

So when people ask why I put money into paintings, the honest first answer has nothing to do with returns. Owning a piece of a Cecily Brown or a Yoshitomo Nara connects me, in a small and very literal way, to the same impulse that put a pig on a cave wall in Sulawesi. That is not a financial argument so I am not going to pretend it is. But it is the reason I looked at the asset class at all, and I would rather say so than dress it up.

Now the financial part.

The gate, and how it opened

A serious Nara canvas trades in the millions. A Soulages, a Frankenthaler, a Joan Mitchell — millions. That market has always been open to the same people: those who can write a seven-figure cheque, absorb the buyer’s premium, insure the thing, store it, and wait a decade.

I cannot do that since I’m not an ultra-high net worth investor. Neither can you, probably.

Fractional platforms — Masterworks is the one I use — buy the canvas, put it into a single-purpose LLC, register the offering with the SEC, and sell you shares at $20 a piece. You own an economic interest in one specific painting. You do not get to hang it but you do get exposure to a market that was, for my entire working life, sealed.

That is the real attraction for me and not just the potential returns. Access. I now hold positions in thirteen works by Banksy, Nara, Soulages, Cecily Brown, Helen Frankenthaler, Robert Colescott, Avery Singer, Liu Ye and Christine Ay Tjoe. Total outlay: $17,000. Not one of those artists would have taken my call.

Three of them, if you want to see what I actually own:

• Yoshitomo Nara, Under the Hazy Sky — my largest single position, Sotheby’s Contemporary Art Evening Sale, Hong Kong, 2021
• Christine Ay Tjoe, First Type of Stairs — same Sotheby’s sale
• Yoshitomo Nara, Sprout in Hands — Sotheby’s 50th Anniversary Contemporary Evening Auction, 2023

The correlation argument, told straight

Here is the pitch you will hear everywhere: art does not move with the stock market, so it diversifies your portfolio.

From my own research, the measured correlation genuinely is low but I do think that it’s important for you to understand why before you lean on it.

Paintings trade rarely and are valued by appraisal in between. Appraisals smooth. A smoothed series cannot fall in step with a market that reprices every second — not because the underlying value is stable, but because nobody is looking. That is a measurement artefact dressed up as a risk property, and it is the same trick that made unlisted property look bulletproof for thirty years.

Art is not a luxury that civilisation produced once it had spare capacity. Art came first.

Worse, when researchers correct for the fact that owners choose when to sell — you sell the winner, you sit on the dog — the numbers deflate badly. Korteweg, Kräussl and Verwijmeren studied 32,928 repeat-sale paintings from 1960 to 2013 and found that selection correction cut average annual index returns from 8.7 per cent to 6.3 per cent, and the Sharpe ratio from 0.27 to 0.11 (Review of Financial Studies, 2016).

Nor is art independent of equities. Goetzmann, Renneboog and Spaenjers tracked art prices against markets from 1830 to 2007 and found equity capital gains — same-year and lagged — significantly drive art prices. Which makes sense. The people bidding at Sotheby’s got rich somewhere, and mostly they got rich in financial assets.

So: low measured correlation, yes. Real, structural independence from the wealth cycle, no. The diversification is thinner than the brochure suggests. It is not zero, and I still think a small allocation earns its place — but hold it for the right reason, and size it accordingly.

My own book, warts and all

I have had one exit. In August 2024 I put $1,000 into a Joan Mitchell. It exited in November 2025 at $23.40 a share — $1,170 back, a 17.0 per cent gain, roughly 12.7 per cent annualised. Clean, and exactly what the model promises when it works.

Across the twelve positions still open, the spread is wide. Cecily Brown’s This is Funfun is marked at $24.26 against my $20 cost — up 21.3 per cent, and it is my second-largest holding. Christine Ay Tjoe’s First Type of Stairs is up 16.9 per cent. Frankenthaler’s Seaspace, up 14.1 per cent.

The other side of the ledger is uglier. A Soulages is marked down 37.9 per cent. An Avery Singer, down 22.4 per cent. My biggest position of all, the Nara, is down 10.5 per cent.

Net of everything — one realised exit plus twelve open positions at Masterworks’ mid-point valuations — I am at roughly $16,773 against $17,000 invested, so I’m down about 1.3 per cent over three and a half years.

Over the same stretch the S&P 500 returned north of 100 per cent on a total-return basis. I am not going to pretend that away.

Two more things you should know before you copy me. Those valuations are Masterworks’ own appraisals, not transaction prices — two of my holdings carry no appraisal at all and sit at cost by default. And the platform takes 1.5 per cent a year plus 20 per cent of profits, which is a private-equity fee load on an asset that may take a decade to clear.

So why am I still in?

Because I did not buy this to beat the index. I bought it for a small, uncorrelated-ish sleeve of a portfolio that is otherwise entirely exposed to listed markets, and I bought it because I wanted to own a piece of the thing that makes us human. On the first count the jury is out. On the second, I am already paid.

So my advice to you, if you’re keen on art, is to keep it small. Be patient. And be honest with yourself about which of those two reasons is doing the work.

Sources

1. Oktaviana et al., “Narrative cave art in Indonesia by 51,200 years ago”, Nature, July 2024.
2. Korteweg, Kräussl & Verwijmeren, “Does it Pay to Invest in Art? A Selection-Corrected Returns Perspective”, Review of Financial Studies 29(4), 2016.
3. Goetzmann, Renneboog & Spaenjers, “Art and Money”, NBER Working Paper 15502.
4. Sotheby’s lot records.
5. S&P 500 annual total returns, Slickcharts.
6. Portfolio and transaction data: Masterworks account exports, 9 September 2026.

It pays no dividend. It doesn’t grow, hire, invent or compound. You buy it, lock it away, and then pay someone to guard it. Stocks, by contrast, are claims on real businesses full of people who get up every morning trying to make things better. Over long stretches of time during my lifetime, that engine has won. Comfortably.

I still believe most of that. But I’ve changed my mind about where precious metals belong in my own portfolio, and I’ve moved part of my money out of stocks and into gold and silver. Three things did it.

1. The printing press never really stopped

Here’s the number that got under my skin. Before the 2008 financial crisis, the US Federal Reserve’s balance sheet was under $900 billion. Three rounds of quantitative easing (QE) later, it was around $4.5 trillion. Then Covid arrived, and by 2022 it was nearly $9 trillion.

Roughly tenfold, in fourteen years.

Closer to home, the Bank of England bought £895 billion of bonds through its QE programme. In the US, the broad money supply grew by about a quarter in a single year during the pandemic, the fastest pace since the 1940s.

Then came the clean-up. The Fed spent three and a half years shrinking its balance sheet, and stopped at the start of December 2025 at about $6.6 trillion. That’s still more than seven times its pre-2008 size. Within days it was buying Treasury bills again. These were “reserve management purchases”, not QE, as it was careful to point out. (Fair enough. It walks a lot like a duck, though.)

That was the moment it clicked for me. The emergency measure had quietly become the operating model. And if the supply of money can be expanded at the stroke of a pen, I want some of my wealth parked in something that can’t be.

I’ll be straight about one thing: QE doesn’t push gold up in a neat straight line. Gold peaked at around $1,920 an ounce in 2011 and then slid to roughly $1,050 by the end of 2015, even as the Fed pushed its balance sheet to a then-record $4.5 trillion. Anyone who bought in 2011 purely on the money-printing argument sat through four very uncomfortable years. The logic is long term. The timing is anyone’s guess.

2. Watch what central banks do, not what they say

The second thing that changed my thinking was seeing who else was buying.

Between 2010 and 2021, central banks bought an average of about 473 tonnes of gold a year. Then something shifted. They bought more than 1,000 tonnes a year for three years running, from 2022 to 2024. 2025 “slowed” to 863 tonnes, which is still close to double the old average.

Poland alone added 102 tonnes last year. Gold is now 28 per cent of its reserves, and it has set itself a 30 per cent target. Kazakhstan, Brazil, Turkey, China and the Czech Republic were all buyers too, and the list isn’t confined to any one political bloc.

Then there’s this. According to the European Central Bank, gold made up 27 per cent of global central bank reserves at the end of 2025. That put it ahead of US Treasuries (22 per cent) and the euro (15 per cent). Part of that jump is simply the gold price going up, and I don’t want to oversell it. Still, the institutions that print money are choosing to hold more of the one reserve asset that nobody can print, and based on what I’ve read in the last few weeks, central banks are starting to repatriate a lot of that gold as well.

Central bankers are not gold bugs. They are cautious, committee-driven people. When they quietly back up the truck, I pay attention.

Key figures referenced in this piece. Prices approximate, early September 2026. Gold record set 28 January 2026; silver, 29 January 2026.

3. Silver: part money, part raw material

Gold is a monetary metal, pure and simple. Silver does two jobs.

It has been money for thousands of years. In French the word for money, argent, literally means silver. It is also one of the most useful industrial materials we have: silver conducts electricity and heat better than any other metal. It goes into solar panels, electric vehicles, phones, missiles, and the data centres behind the AI boom.

Solar has been the big story. In 2014, photovoltaics accounted for 11 per cent of silver’s industrial demand. By 2024 it was 29 per cent. Industrial use now takes close to 60 per cent of all the silver consumed each year. And the market has been in deficit, with more silver used than produced, every year since 2021. The Silver Institute is forecasting a sixth straight deficit for 2026.

The supply side is what really caught my eye. Roughly three-quarters of the world’s silver comes out of the ground as a by-product of mining lead, zinc, copper and gold. So when the silver price rises, miners can’t simply dig more silver. Supply is stubborn, and stubborn supply meeting steady demand is an interesting set-up.

Two caveats, in fairness. First, industrial demand isn’t growing in a straight line either. It peaked at around 680 million ounces in 2024 and dipped about 3 per cent in 2025, because solar manufacturers are working hard to use less silver per panel (high prices tend to do that). Second, silver is volatile. Properly volatile (and some say, prone to market manipulation). It hit a record of over $121 an ounce at the end of January 2026 and is trading around $66 as I write. That’s a drop of nearly half in about seven months.

Silver is gold’s wilder younger brother. I own some, but I keep him on a shorter lead.

How I actually hold it

I’ve split my holdings across three routes, because each one does a slightly different job.

Vaulted bullion in Singapore, through BullionVault. This is professional-grade bullion held in a vault in Singapore, one of Asia’s main bullion hubs. The gold sits in large London Good Delivery bars, the wholesale industry standard, and my share is legally mine rather than an IOU on someone’s balance sheet. Dealing costs are low, and I can sell online whenever I need to.

Gold coins: the Queen’s Beasts and the Tudor Beasts. I’ll admit it, I bought these partly because they’re beautiful. The Royal Mint’s Queen’s Beasts series (2016–2021) depicts the ten heraldic beasts that stood guard outside Westminster Abbey at the 1953 coronation. The Tudor Beasts that followed are inspired by the King’s Beasts on the moat bridge at Hampton Court Palace. They are small works of art you can hold in your hand. They are also bullion coins, so the premium over the spot gold price is modest and you aren’t paying through the nose for the artistry. As a bonus, Royal Mint bullion coins are UK legal tender, which makes them exempt from Capital Gains Tax for UK residents. A handy bonus for anyone who is, or later becomes, UK tax resident.

Physically backed gold and silver ETFs. This is the practical route. They trade like shares in a normal brokerage account, they’re easy to top up or trim, and they’re backed by metal sitting in vaults. The trade-off is that you own units in a fund rather than the metal itself, and you pay a small annual fee for the privilege.

The vault is my long-term store. The coins are the part I can hold, and the part I enjoy most. The ETFs are the easy-to-adjust part.

Why I stop at 15 per cent

This is where I part company with the gold bugs.

Like I mentioned previously, gold doesn’t pay anything. No dividend, no coupon, no rent. It will never compound the way a good business does. If you had bought gold at its 1980 peak of $850 an ounce, you would have waited 28 years just to get your money back in nominal terms, and roughly 45 years (until 2025) once you allow for inflation. Stocks did vastly better over that stretch.

So I don’t think of gold as something that makes me richer. I think of it as insurance that stops me getting poorer. Think wealth preservation, not wealth creation. Different tools, different jobs. You wouldn’t use a spirit level to drive in a nail.

That’s why I cap precious metals at 15 per cent of my total portfolio and my net worth. It’s enough to matter if the monetary system takes a nasty knock, and not so much that I give up the long-term compounding that equities and income-producing assets provide. For what it’s worth, it sits at the top of the 5–15 per cent range that Ray Dalio has argued for. It is not a small bet.

A cap also brings a bit of discipline. When metals run hard, as they did through 2025, they can grow past the limit, and the cap says it’s time to trim. When they fall, as silver has this year, there’s room to add. The rule does the thinking, so emotion doesn’t have to.

Where I could be wrong

I’d be kidding myself if I pretended this was a one-way bet. Here are the things that keep me grounded:

• Interest rates. Gold tends to struggle when cash and bonds pay a decent real return relative to inflation. With oil back near $100 and markets openly pricing a possible Fed rate rise this month, that headwind isn’t theoretical.
• Central bank appetite. Buying fell 21 per cent in 2025. If it keeps easing, one pillar of the story gets weaker.
• Thrifting. Solar makers are getting very good at using less silver. Industrial demand could keep slipping.
• Price. Gold more than doubled from the start of 2025 to its peak in January 2026, and it has since given back about a fifth. Buying after a run like that carries its own risk.

The bottom line

Gold isn’t going to make me rich. That was never its job. Its job is to still be there, holding its value, when other things aren’t. After watching the printing presses run for the best part of two decades, I’m happy to pay a small price for that peace of mind.

That’s my approach, for my circumstances. Yours may well look different, and that’s fine. If you’ve been through a similar change of heart, or think I’ve got this completely wrong, I’d love to hear from you.

Sources

Prices are approximate as of early September 2026.

1. World Gold Council — Gold Demand Trends: Full Year 2025, Central Banks.
2. MINING.COM — “Gold overtakes US Treasuries in global reserve shift: ECB”, June 2026.
3. Silver Institute — Silver News, February 2026.
4. Mexico Business News — World Silver Survey 2026 summary.
5. Statista — global silver production by source, 2025.
6. International Banker — “Why the Fed began a new phase of balance-sheet expansion”.
7. Bloomberg — “Fed begins $40 billion monthly bill purchases”, December 2025.
8. CBS News — highest gold price in history.
9. Yahoo Finance — silver prices, 8 September 2026.
10. Yahoo Finance — Ray Dalio says gold should be 5–15 per cent of your portfolio.
11. BullionVault — the Singapore vault.
12. The Royal Mint — gold bullion coins and Capital Gains Tax.