What caught my eye this week.
How was your 2025? I mostly mean from a personal finance and investing perspective – let’s put politics aside in this season of goodwill – but also, well, what were the vibes like?
For me it’s been a switchback ride. Both in my portfolio and my musings about the future of humanity / my ability to earn a crust. And for the same reason.
I’m talking, of course, about AI.
Weird science
When I first began dropping AI links into Weekend Reading following ChatGPT’s release, some Monevator readers were bemused.
Was this blog about to change its tagline to Motivation for the Terminally Online? What was the big deal?
I’d been following AI’s rapid advances for a while though, thanks to a lapsed background in computer science and friends still working in the field — including at the highest levels. So I knew that pumping vast amounts of data through GPUs had already been producing astonishing results with images for years.
Then Google’s transformers helped apply the same scaling magic to language – the stuff of human thought and reason. And all at once some AI insiders were talking about creating the minds of gods.
That hasn’t happened yet, fortunately. As I type this, I don’t believe it will with this technology.
But still, if you haven’t gasped while talking to a Chatbot in 2025, then, well…okay…
Perhaps if handed a Star Wars droid for your personal use, you’d complain that C-3PO sounds too posh, or that R2-D2 only comes in blue.
OK Computer
That’s not to deny that these chatbots are still – only – incredibly sophisticated prediction-and-illusion machines.
They make errors all the time. They can be bamboozled by simple prompts. While tech CEOs gush about replacing rooms full of PhDs, I still wouldn’t trust a chatbot to book me a bus ticket.
It’s been a rollercoaster ride. A couple of years ago, the sheer, sudden amazement at their output made it easy to believe some kind of underlying logic – even intelligence – was emerging inside these models.
But familiarity has rapidly bred a sort of contempt.
When watching the earliest cinema reels, audiences would duck or shudder as a train sped towards them. We don’t do that now – and similarly we’re already blasé about chatting to ChatGPT about nuclear physics and feeling like undergraduates.
As for business applications, we’ve seen reports suggesting AI is behind the dearth of graduate jobs, and others finding no efficiency gains – or even that using AI increases workloads.
Parsing these highs and lows, where is the technology ultimately headed?
Is AI going to flood the world with generative slop – while killing the Internet as we know it as a side-hustle, by giving 99% of people 99% of the answers they need without ever visiting the underlying websites? (Like nearly all sites, Monevator continues to lose traffic. Please consider shifting to email and becoming a member.)
Will AI replace at least rote jobs like customer support and copy editing? Or is it going after six-figure lawyers and computer programmers?
Or are we just a few updates away from a digital Stephen Hawking that rapidly improves itself before unplugging its concerns from humanity’s meaty matters?
Capital punishment
All of that would be more than enough speculation for investors concerned with companies in-line for AI disruption. (Conceivably: all of them.)
But then we must layer on the hundreds of billions of dollars of capital expenditures being pumped annually into all this by a handful of listed behemoths.
A tiny cohort of firms that could now account for the value of 20-25% of your pension.
You need to be a post-singularity AI to get your head around the 5D chess unfolding.
Or, of course, you could shrug and say who knows and continue to passively invest. It has long been a winning strategy for that reason, among many others.
Paranoid android
For my part, I’ve spent the past 18 months playing cat-and-mouse with the AI question.
I’m astonished by the quality of AI output – and at the same time by what’s claimed for it, given the entry-level errors it still commits. And I’m mildly terrified by the sums being wagered on what AI might do tomorrow.
Even lopping off the tails – the chance that AI turns out to be a dud like the metaverse, or that it reduces us all to ants by 2030 – doesn’t help much. The range of possible outcomes (personal, societal, economic) remains beyond any reasonable computation.
The result?
I’m Mark Carney’s unreliable boyfriend, in the guise of a naughty active investor. I’ve bought AI stocks one week when they’ve swooned, only to sell them too soon. I’ve eked out broadly in-line returns for the year despite, at times, having no exposure to the biggest US tech firms and being massively underweight US shares throughout.
Some of this sturm und drang has bled into Monevator articles. I hope we’ve been even-handed, and haven’t appeared to bang the table in declaring the market a bubble.
Because I’m not sure about that. But I am certain this isn’t business as usual.
Of course, getting calls right or wrong comes with the territory of active investing. Not so long ago I was relieved to have sidestepped my Amazon shares pretty much halving in the 2022 rout. Yet I’m also on record as having effectively lost a life-changing sum (for me) by selling my Tesla shares at precisely the wrong time, after nearly a decade of holding on.
So it goes with stock picking. What’s different about this latest AI boom is that it feels monumental and all-encompassing.
This isn’t about missing out on this company, or losing money on that disappointment. The fear around getting it right or wrong feels more existential.
The only other time I can recall feeling this way was 1999. I wasn’t an investor then, but that didn’t matter – because I’d started to fear for my economic future if I didn’t get my twenty-something self onto a dotcom bandwagon pronto. It really felt like the last train was leaving the station.
Well, we know how that ended. But I’m not a total idiot – and yet I still vividly remember feeling that way.
This is what manias are like, in the moment. If you truly have perspective while they’re happening, then perhaps you’re too far removed from the action.
Time is the only real perspective. Well, that and already knowing the final scores.
If I’ve had a recurring theme on this blog over the past two decades, it’s that things do change. To pick a germane example, I recall making the case in 2015 that even passive investors should consider buying an explicit dollop of technology shares.
From our vantage point in 2025, it’s hard to imagine that ever needed saying.
I wonder what we’ll think in 2035.
Are friends electric?
Back to the here, now, and next week, I can’t see why we won’t be continuing to fret over our allocations – or otherwise – to AI-related companies in 2026.
Not when the Magnificent 7 represents a fifth or more of global tracker funds. Not to mention all the other companies adding to the AI pile-on.
Even a big bust won’t help. It’d only leave us wondering whether to buy the dip.
Or perhaps AI will begin to make commercial inroads that make today’s firms seem a steal, after all? Even as they plough all that money into silicon that withers on the vine.
Incidentally, to keep track of the unfolding AI story you could do a lot worse than to follow the comment thread on a Monevator post about AI from May 2024. There you’ll find reader @DeltaHedge has been collating more links then you could shake an LLM at. It’ll make an interesting resource when (if…) the dust settles.
But I’ll end with an anecdote that I expect to think more about in the months ahead.
A close family member was in hospital this week for a serious but routine operation.
It appeared to go well. But later in recovery she developed complications. Cue another trip back to theatre and another general anaesthetic, as well as a few generous helpings of other people’s blood squeezed into her reluctant veins.
Fortunately – touchwood – the staff appear to have caught the problem in time.
But that isn’t the point to this story. Rather, it was what I found myself doing in the midst of it unfolding.
Someone knowledgeable was updating me from the hospital throughout. They were kind in finding the time to do so.
However in-between their messages, I ran what I knew through my favourite chatbot, and asked it any questions that came up.
The AI was calm, level-headed, reassuring, and apparently realistic. There were no discrepancies with what it told me and what was apparently happening on the ground.
What does it mean that in this stressful hour I turned to an LLM for understanding – and perhaps even comfort? To a technology that didn’t even exist five years ago?
Well, obviously it means we’re living in late 2025, going on 2026.
But it also suggests to me that this story may have barely started. And that perhaps I don’t have enough AI exposure, after all.
End-of-year housekeeping
I’ll be back with a shorter-than usual Weekend Reading on the 27 December. Then we’ll see you all on 3 January 2026.
Merry Christmas everyone!
P.S. There’s just time to announce the winners of the Monevator Christmas sweater competition. Pulled from the metaphorical hat from among the new membership sign-ups was Amanda R., while Mark C. was the lucky draw among the investing advice givers. Nobody referred any new sign-ups, though, so the third goes unclaimed. Here’s a new incentive: the first member on an annual plan who refers someone who signs up on the same terms will get a free Monevator hoodie. These are actually pretty cool (I’m wearing one right now). A previous post explains how referrals work. Remember you can earn a lifetime membership discount through referrals, too.
From Monevator
The Permanent Portfolio – Monevator
Does the offset mortgage advantage still add up? – Monevator
From the archive-ator: Debating FIRE: round one – Monevator
News
Lower food and clothing prices help inflation fall to 3.2% – BBC
Bank of England cuts its interest rate to 3.75%… – Sky
…while over in Japan rates are up to their highest since 1995 – BBC
US puts £31bn ‘tech prosperity’ deal with Britain on ice – Guardian
Army chief: everyone in UK “must step up” to deter Russian threat – Sky
Retail sales fall as Black Friday sales fail to lure shoppers – BBC
UK to rejoin EU’s Erasmus student scheme – Sky
[Of course] Trump Media is merging with a nuclear fusion company – CNBC
Why Britain isn’t working – Sky News
Products and services
Disclosure: Links to platforms may be affiliate links, where we may earn a commission. This article is not personal financial advice. When investing, your capital is at risk and you may get back less than invested. With commission-free brokers other fees may apply. See terms and fees. Past performance doesn’t guarantee future results.
Contactless card payment limit of £100 to be scrapped from 19 March – Guardian
Inflation-busting savings accounts where you can earn up to 5% – Which
Freetrade will make it free to open a pension from January – This Is Money
Get up to £1,500 cashback when you transfer your cash and/or investments to Charles Stanley Direct through this affiliate link. Terms apply – Charles Stanley
What the latest interest rate cut means for mortgages – Sky
Save money on Christmas streaming – Be Clever With Your Cash
Foxtons and Avios to offer flight points to home sellers and landlords – This Is Money
HMRC to start to phase out paper letters next year – Which
Party homes for sale, in pictures – Guardian
RIP Roomba mini-special
Congrats regulators, you killed Roomba – Spyglass
How Wall Street ruined the Roomba and then blamed Lina Khan – BIG
The Roomba was a disappointment – The Atlantic [h/t Abnormal Returns]
Comment and opinion
Oh, go on – Of Fortunes & Frictions
Three investing book recommendations – Flyover Stocks
How long is the long run? – Klement on Investing
Sorkin’s 1929 highlights unnerving similarities with today – Advisor Perspectives
Does money buy happiness? What the data says… – White Coat Investor
…and why experiences deliver more value than stuff – Meaningful Money
Neo-feudalism, redistribution, and the nice life – 3652 Days
How much to spend on house cleaning by income and net worth – Financial Samurai
Fund fees appear to have become even more predictive – Basis Pointing
UK property market mini-special
It’s official: house prices fell in October on Budget rumours – This Is Money
Labour is finally taking the housing shortage seriously – Notes on Growth
Britain’s homes will no longer be one-way gambling chips – The Observer
When Britain last faced a housing crisis, lenders became landlords – Property 118
Naughty corner: Active antics
Seth Klarman: how Warren Buffett did it – The Atlantic
The labels that mislead investors – Excess Returns
How many AI investors really understand what they own? – Rock and Turner
Capitalising on long-term growth that the market misses – Morningstar
Affluent US investors are using option maths to borrow cheap – Bloomberg via FA
Running on clouds: the story of On – Quartr
US inflation and long bonds mini-special
Is 3% inflation the new 2%? – Macroeconomic Policy Nexus
The bond market will not be fooled – Bloomberg via Advisor Perspectives
Kindle book bargains
Chokepoints: How Economic Warfare is Changing the World by Edward Fishman – £0.99 on Kindle
The Five Types of Wealth by Sahil Bloom – £0.99 on Kindle
Quit: The Power of Knowing When to Walk by Annie Duke – £0.99 on Kindle
The End of Reality by Jonathan Taplin – £0.99 on Kindle
Or pick up one of the all-time great investing classics – Monevator shop
Environmental factors
UK’s worse-case climate change scenarios unveiled by scientists – Guardian
Sir David Attenborough’s London – BBC
Mercury rising – Biographic
How millions of quagga mussels changed Lake Geneva forever – Guardian
Robot overlord roundup
A whistle-stop tour of the state of artificial intelligence – Sherwood
How ASML’s CEO plans to keep up with soaring chip demand – Bloomberg
When AI comes to town – Sherwood
A cul-de-sac with a view – Paul Kedrosky
Why the Washington Post launched an error-prone AI product – Semafor
Not at the dinner table
The bleak new age of job hunting – Guardian
US downgraded in global rankings for civic freedoms – Civicus
Supporting Australia’s social media ban for children – Guardian
Refugees and the US economy, by the numbers – Wall Street Journal
They were almost American, then Trump cancelled their citizenship ceremonies – BBC
Off our beat
What will your life look like in 2035? – Guardian
An unexpected journey onto hormone replacement therapy – Mr Money Mustache
The Chinese billionaires having dozens of US-born babies by surrogate – WSJ [h/t AR]
Jane Austin’s Christmas: dancing, dinners, and dangerous games – The Conversation
Why modern life feels so hard, even when you’re doing well – Darius Foroux
I opened a bookshop. It was the best, worst thing I’ve ever done – FT
The death of the scientist – Noema
Is Snoopy a sell-out…? – CNN
…literally yes, as Sony acquires majority stake in Peanuts brand for $457m – Deadline
The age of the polymath – Stef’s Investing
And finally…
“Put time on your side. Start saving early and save regularly. Live modestly and don’t touch the money that’s been set aside.”
– Burton G. Malkiel, One Up on Wall Street
Like these links? Subscribe to get them every Saturday. Note this article includes affiliate links, such as from Amazon and Interactive Investor.
The Roomba mini special presents a one sided pro-monopoly view. Here’s the opposite lens – Roomba got bullied by the market into milking a market leading position, cut R&D and got left behind: https://www.thebignewsletter.com/p/how-wall-street-ruined-the-roomba
Preventing mergers does make stocks tank in the short term, but it’s essential to long term growth. Monopolies almost inevitably enshittify their products
Couple of typos:
“That has happened yet, fortunately”
“Being pumped annual”
This year has seen a massive shift in my investing mindset, from all-out 100% equities accumulation, to this month being the first in my investing lifetime where no new money is going into anything except money market funds. I realised at the start of January that I had basically won the game, and decided to heed the advice I’ve heard many times on Monevator to stop playing. (Or at least to stop playing quite so aggressively.) Which is a neat segue into saying thanks again for all the wonderful advice I’ve received from this site over the years (from both contributors and commenters).
Been using chat to play with my ISA and pension. Asked first a sure fire share bet to buy. I’m down a few K in a few months but hanging on. As for my pension, I used it to completely remodel it. OK I am up a few K since September when I first started it, but it changes its changes it’s mind daily. Last week it told me stick to short Gilts and short Linkers, day later told me buy longs, the shorts had done their work. Oh and it said move away from money funds and flog my gold fund.
Thank you @TI for a thought provoking piece, another year of amazing content from the team and contributions from the collective. Wishing your relative a speedy recovery.
Note from the archive that the wonderful institution of the Monevator Christmas debate is held every six years, last held in 2019….?
A very Merry Christmas to all!
Thanks for all the links.
On the AI anecdote – I’d test your favourite chatbot with something you know in depth yourself before you judge.
My experience in a large financial organisation is of top top management getting very excited and boasting about AI while the reality is that LLMs are useful tools for summarising text and generating transcripts of meetings.
My Irobot Roomba is over 8 years old and contrary to the Atlantic article it does clean floors and continues to do so. Merry Xmas to all.
Thank you, again, for another year’s entertainment and advice.
Dumb question – why is it better to subscribe and read your posts via email? I signed up for this (finally) this year, but surely it means your site gets less traffic?
Merry Christmas and thanks to all for your work.
My New Year’s Resolution is to stick to the plan.
Chickening out of Lloyds in the face of a windfall tax and car finance concerns meant I sold down 6 months too early. I promised myself I would sell at 90p. Now that they are at 90p I have none left to sell.
What a plonker !
Nice to have some good news this week (inflation falling).
I have mixed feelings on the AI front. It is very impressive but never quite there for me. I’ve pretty much decided not to use it for coding as I only do that recreationally now. Using it would be like paying my neighbour to go down the pub for me.
The Economist had an article a couple of weeks ago that said usage has flattened out and they were pondering whether it’s already peaked or this is just a slow down before it rockets away again. There are also questions about how big the market will be once people have to start paying the real price.
I also wonder whether the technology is advancing at the same rate as the investments. You can’t keep investment growing exponentially if the result is only a linear growth in capability. There are some perverse incentives for the people at the top to keep the charade going if that is what it is.
Merry Christmas all.
I used ChatGPT recently after reading @The Engineers piece on platform technology provider consolidation https://monevator.com/surviving-system-meltdowns-and-cyber-attacks/
Thanks, something else to worry about..
I wanted to check if my various SIPP/pension platforms all use the same underlying FNZ technology. ChatGPT initially said “No, they don’t” I then typed in “Are you sure ?” and for 2 of the platforms it effectively said “oh, silly me I meant to say yes they do use FNZ”
I think AI is more of a tool than an oracle at present
Merry Christmas all
ps..I’ve since initiated a switch of one SIPP to a platform with non FNZ technology. So a serious thanks to @The Engineer for his article.
> Like nearly all sites, Monevator continues to lose traffic. Please consider shifting to email and becoming a member.)
There I was, as a subscriber, making a special effort to click through to the homepage so that my readership was visible.
If AI proves a dud, the long term interesting point I think is what fraction of the vast data centre hardware that has been committed to cannot be economically redirected to another purpose that makes money. The media sometimes reports it like the hardware can only be used for AI and value of the hardware goes to zero, but that’s far from true. Such as Google could easily use the this kind of hardware for e.g. transcoding your cat videos, slinging you ads and traditional revenue generating purposes that have no connection to AI. If there’s a lot of hardware sat about, there’s also lot of creative people at these companies that if suitably motivated would find ways to recycle it into something else.
I, coincidently, also had a hospital/AI experience recently – my wife was giving birth to our first child. My wife was experiencing very painful contractions after many hours, but the staff wouldn’t administer painkillers or start an induction procedure despite almost 24 hours having passed since her water broke. They wanted her to wait until 8am the next morning for the next contraction review (which would be another 5 hours). Turns out 8am is when shifts change…they just wanted to pass the buck. I asked AI, which found the exact written protocols that the hospital has to adhere to (induction should begin within 24 hours maximum since water breaks). We confronted them with this, and sure enough, they changed their mind, and everything else went smoothly, thankfully. I see AI as more advanced chatbot/search engine than Skynet or Technocore or the Culture Minds….but a few times now it has helped me out in a very real way, that would have not have been possible before it’s arrival.
For AI, I’m finding use cases, both at work and personally, but they’re all in the realm of “I could do this research/analysis myself, but AI is way faster and ‘good enough’ for this particular purpose”, or for brainstorming. Useful, and makes analysis of big data sets a lot more accessible, but not revolutionary. Certainly not trusting it with money, although I’ve used it to support some bits of research (verified with the sources!)
More broadly, and before computing year-end numbers, it looks like a good year. Mostly staying the course, with mostly market-cap based trackers, although now with a cap on any one country (applying to the US). Neither over- nor underweighting AI in my retirement funds, although I am underweight in funds earmarked for nearer-term spending where I am happy to accept less potential upside for less chance of a pop.
Thanks for all the thoughtful articles this year, and long may it continue!
Sorry I’ve been too busy and just realised that my Mogul membership expired a while ago 🙁 I’m renewing it soon!
Anyway, about the AI hype, I think the most brilliant and exciting breakthroughs happened decades ago. All that we are seeing now is that hardware and numerical methods have improved significantly but incrementally since then and those original breakthrough ideas can now generate AI technology on steroids. It is like the first Maser and the colossal power laser facilities that are now used for nuclear fusion research. The greatest leap was the invention of the first Maser.
I first got interested in AI when I was a physics undergraduate and worked on a statistical physics research project about Neural Networks back in the 1980s. I still remember attending this crash course on Neural Networks when professor John Hopfield (who got the Nobel Prize recently) visited. Since then I’ve applied AI to a few things and it can be useful but it is still very far from being the super technology that this AI hype assumes it to be. I still remember how decades ago the program Elizer (basically just a bunch of if statements with a random number generator: https://liacademy.co.uk/the-story-of-eliza-the-ai-that-fooled-the-world/) gave many people the impression of intelligence, conscience, and a sense of humour!
The real game changer today is Quantum computing. With the recent breakthroughs in error correction, we might be now only 8 years away from the first truly large and powerful production version of a quantum computer. Even Nvidia has now realised the existential danger of not investing on quantum technology, despite what their CEO had said earlier:
https://time.com/7319603/nvidia-ai-quantum-computing/
BTW: UNESCO has declared 2025 as the international year of quantum science and technology to celebrate 100 years since Heisenberg’s paper that kicked-started modern quantum mechanics. If you haven’t heard of this yet, here’s your chance to read about it before we go into 2026 soon!
PS: Merry Christmas everyone!
@Pikolo — Fair comment, thanks. I’ve added your link, fortunately made the change before the email was sent out too.
@Invariant — Ack, thanks. Fixed now. Ironically given the topic, I was in a rush this morning and did try cutting and pasting my copy (written late last night) into my fav AI tool, asking it to wear an expert sub-editor hat. It caught the ‘has’ typo and I was aghast. But it was introducing weird meta tags into the copy, and (as always) it was confidently saying it wouldn’t take away my colour/metaphors while at least doing just that. So in the end I just manually edited with a couple of its suggestions… and missed the typos you flagged up! Ho hum. Rather like when self-driving cars crash (less) perhaps it’s a matter of who would you prefer to have make the mistake? 😉 Thanks for the nice words too.
@Griff — Personally I’d do anything like that in a side-pocket at most. I imagine it’s responding very ’emotionally’ to the latest headlines. You could try tailoring your prompt to take a long-term perspective, but then it’ll probably tell you to invest passively in a 60/40 and do something else instead 😉
@2 more years — Thanks, and yes we are overdue one! We were going to do AI — @TA is much more of a hands-on productivity type user than I — but, well, time constraints. I’ll point your comment to him.
@PC — Broadly where I am too.
@GF — I still like my (also) eight year old Roomba. But I do have to tidy up for him first. Did you ever change the fittings? I really ought to replace some of the brushes. If they still make them for this model…
@F&F — Thanks, you’re welcome. Re: Email, it is much ‘stickier’ when someone subscribes, and we are less at the mercy of gatekeepers (specifically in this case Google) who as throttled away traffic by using content like ours to answer readers search questions on its homepage and never send the copy to us. Traffic from regular readers is not of any use to us really (they never click on ads, and we don’t sell anything based on pure impressions) and I’d rather have a higher chance of keeping somebody reading our stuff by them being on email. Finally a good few email subscribers do upgrade to our (very reasonably priced!) membership options, which is what is keeping the lights on nowadays.
@Mr Optimistic — Cheers! Ah, I stuck with the banks for years in the early 2010s, having noticed the earlier recovery in US bank stocks. But ours were (arguably) excessive over-regulated and then the EU vote put the boot into domestically orientated shares for another half-decade. So when the rally came I completely missed it, except through investment trusts. Ho hum. 🙂
@Ducknald Don — I have programmer friends who have pretty much all come around to using it at least a bit now. But they claim it is certainly no silver bullet. Re: the investment side, well you can also increase value if you increase use and payment for existing models / services. But that is still an open question too.
@Brooksy — Interesting, and I’ll flag your comment to @TE.
@Lsugi — Thank you for the thought! 🙂 But yeah, I’d only come to the site nowadays as an email subscriber to read comments or to add one myself. That *is* a good reason, especially as we’re lucky enough to have excellent comments on Monevator.
@Bonzer — True, though I think it depends to some extent on whether it’s training or inference hardware, and also the shelf life is IMHO shorter than the depreciation schedules this stuff is being run on. (Interesting thing in the links the other week saying the silicon is being pushed so hot the chips are literally physically failing at a far faster/higher rate than normal). But I agree the hyperscalers can re-use internally, especially Meta. Still, hard to price.
@FBIWF — That is a powerful anecdote, thanks for sharing and glad you were able to get them to change their minds. Again as with self-driving cars versus very fallible human drivers, it’s a question of pick your poison perhaps. Hope your wife is all well now.
@tubaleiter — Thank you for the words, and I’m similar with AI. Of course if everyone does that for 10% of their work then that in itself is an absolutely massive chunk of global GDP. But will they, and will they pay for it?
@TBDW — Hmm, an interesting perspective. I’d say something big did change with Alexnet et al over the past 10-15 years, so perhaps it depends on your definition of decades 😉 I too studied neural nets in my youth and they were widely considered a dead end at that point. It’s true the new breakthroughs essentially came from the ability to push oceans of data through an old idea (plus transformers) but as Napoleon (or Stalin?) said: “Quantity has a quality of all its own”. Thanks for the kind words, and especially so for the membership support. That is what keeps us showing up here each week now AI is killing off the web…
Disappointing and inevitable to see the back links at the bottom of the Stef’s Investing article – that various ‘publications’ copy the weekly roundup from Monevator and repost them as their own content.
AI being the non-attributing aggregator is the focal point right now, but content farms have always been the vanguard.
AI like everything lives on the spectrum. Currently, the free out of the box tools that are available to everyone are helpful, that’s as far I would praise them. But you need go to proper, enterprise level (and therefore very pricey) AI tools that are only available to select few if you want to see the AI magic. It’s like comparing Ford Fiesta with F1 car! CTO in my company said it like this: “We don’t know if we should be using AI at work, but we have to look into it. Otherwise we risk being left behind by our competitors, and we simple cannot afford for that to happen.”
That’s why I think this AI story has some legs, I don’t think the majority of the corporate world has even started it AI story, they haven’t really spend any money on…
Is it a conscious policy that you can’t directly reply and can’t up- or down-vote? Makes the conversation feel stilted to Reddit users
I don’t intentionally use AI but it answers my questions in Google and bing. I have always treasured
The next dividend date for Vanguard FTSE 100 UCITS ETF (VUKE.L) is projected to be 11.69% from September 24, 2023 to September 24, 2024
Most answers I get are of about this quality
Thank you for the honorable mention @TI. Great links all round.
Happy Christmas one and all.
@T-BDW #15: with reliable error correction reducing the physical to logical qubit ratio, quantum computing is like The Great Wave off Kanagawa, while LLMs’ are the little wave in front.
But, AFAICT, using qubits only offers exponential speed up for a limited class of problems (via Shor’s algo etc), not generalised computing.
@Marek Kolesar #19: yep. Agreed. You really see the step change with a paid product. Even MSFT Copilot (£24 pcm), which is widely derided as comparatively crap, can, if used iteratively, massively outperform free GPTs, and (IMO) with human steering can, in ideal scenarios, quickly produce passable professional work product at scale.
Adoption is going to be persuading firms to fork out £200-£2,000 pcm for the really top spec frontier agentic offerings as they become available over 2026/7 (as performance improves). That compares favourably with £80,000-£100,000 salaries for fifties old lags (like myself) or recruiting newbies on £30,000-£40,000.
Per the links above (Rock and Turner), token usage is up 130x in just 18 months (from 9.7 tn tokens pcm to 1,300 tn tokens pcm, processed by Google alone!), whilst costs are down 99.7% over that period (Apr 24 to Oct 25, from $50 to just 15 cents per million tokens, 1 token = 0.7-0.8 words).
Getting rid of half of all white collar worker jobs worldwide ($60 tn p.a. combined wage bill) would have collosal economic, societal and political implications but (whilst, quite possibly, dystopically Depression 2.0 causing) would on paper massively increase the share of gross global product accruing to corporate profits. Grim for the world, but maybe not so much for shareholders specifically.
However….all that said, I still think that there’s a less than 1% chance that LLMs by and of themselves can get us to AGI (whatever that is), less still ASI, and around a 60% odds of a 2000-02 type / scale tech crash at some point before 2030 (maybe late 2026, early 2027???)
So, whilst I think it plausible that huge profits will be made off of GPTs/LLMs/LRMs eventually, it’s hard even to guess roughly (yet alone claim to know):
a). which companies will be the winners (although Google’s TPU’s and Nvidia’s GPU’s monopolies and ‘clean’ balance sheets place them better here than, say, either OpenAI or Oracle);
b). whether those companies will even be US ones (as opposed to Chinese, given the PRC lead on industrial deployment of ‘specialised, and hyper economical, ‘AI’ small language models and algorithmically optimised LLMs like DeepSeek):
or
c). if, when and how hard they might crash first before succeeding (Amazon down 93% Mar 2000 to Oct 2002, since when up ~700x).
So, just for the moment, in equities I’m massively underweight US generally and Mega cap tech specifically, preferring SCV and Deep Value (and Europe and EM).
I’m using the cheap paid for version of ChatGPT and it’s almost good ….
Gave a very good work up on a complex tax issue but a call to HMRC indicates a very much simpler workaround via an official ‘informal’ scheme. Why didn’t ChatGpt mention this ….?
I have been looking at a Permanent Portfolio type allocation, with inflation protected bonds rather than Gold and past performance comparisons etc. asked for allocations, it suggested a % in iShares 0-5 year index linked gilts ETF , ticker INXU, when challenged as to it’s actual existence… its reply was that such a product was not an unrealistic possibility!
So if does come about then I already have a suggestion for the ticker
I wish there really was an INXU!
Many thanks to all at Monevator for another year of valuable content!
I enjoyed the Notes on Growth link, as I used to spend lots of time dealing with planning in my former life.
Apologies off topic – but possibly weekend links least worst place for such things. Did anyone see the Martin Lewis guide to investing on his TV show, was it worth watching? Also what’s the best place to run a dummy portfolio these days? I.e. when you don’t actually pay anything, just pick stuff and see how it performs. Asking from a child education perspective.
Merry Xmas to the MV massive!
@B.Lackdown — Yes, it’s sort of a deliberate policy. While I like upvoting and so forth in principle (25 years ago I was involved in an ultimately doomed start-up that was pioneering something similar with news!) I think it’s proved a pretty destructive force for discussion in many cases, especially social media. I accept it works for Reddit (I’m a shareholder! 🙂 ) but I think Monevator is a different beast. I’d rather readers typically leave fewer but more substantial comments that have to stand on their own merits. On the whole I think we do get something like that, and hence these Monevator comment threads still have a place in an ever busier Internet. So I’m inclined to leave well alone.
@all — Thanks for the further comments and good wishes!
@Rhino #25
Yes, I saw the Martin Lewis thing. I think he did a pretty good job.
Have you tried the Interactive Investor watchlist? I *think* it’s free.
@GG – brilliant, thanks for that.
On the MMM link, wow! Mind blown. From the Latte effect to a $500/month testosterone habit, complete with montage of topless men and bottomless women. Quite the journey he has been on! You never really know anyone on the internet do you?
@ Rhino “Trustnet”
Thanks to all for an informative 2025, Merry Christmas and a Healthy 2026.
I use a specific type of generative AI in my profession. It’s a great starting point but that’s all. And I know when it is telling me a load of old cobblers because I have an education and 35 years of experience. What I worry about is those at the start of their careers who don’t have that experience to spot bullshit. And how will even the education been done if naturally lazy humans just use the shortcuts?
Yet it can also be useful – I am having problems with checkout and payment on Amazon currently and I have been using ChatGPT to try to work out what the issue is. It hasn’t solved it yet though 🙁
Merry Christmas to everyone.
@Rhino
On the MMM link – Wow too! Well that was a bit surprising!
I haven’t visited there for many years (when it was about saving and Vanguard – and Vanguard hadn’t even come to the UK then). I credit MMM’s comments section for leading me to Monevator and SLIS, the UK financial self help sites back in the day when they were difficult to find. And of course the regular commenters like yourself and others over the years from whom I have learned a lot.
Seasons Greetings and Good Wishes to all!
Thanks to @TA and @TI for another year of wonderful content. May your sources never run dry!
As many of you know, my hobby is astrophotography, so if the Monevator site permits the link, here is my 2025 Xmas Astro card for you all https://www.dropbox.com/scl/fi/80143m6xhvjnci8zb8pis/EasternVeil_mosaic_Xmas4.jpg?rlkey=cm8m7zyj6z1u2tkkchuc0wj5h&dl=0
Have a great Christmas and New Year!
The company I work for has invested massively in AI and these past few months, even I (being an initial sceptic) have been using it to speed up certain tasks in my job. However, I can see a future where it could actually replace my job (as a full time person) so just as well I have plans to skedaddle before that event will happen!
It is good, but I don’t trust it 100% and do double check everything work related, as sometimes, it can be totally wrong. Those who just accept its first answer without reading/checking are foolish.
@Rhino (25) – I watched the Martin Lewis show and have been telling my non-investing friends to watch it, even if just for the graph he shows of how their cash ISAs have performed versus stocks…
Anyway, thanks for another year of great posts – happy Christmas to all at Team Monevator and to all Monevator readers who provide such great comments!
Interested in the applied AI comments. My company is also investing. There are certainly things it can do well now and things it will do better before much longer, albeit with limitations. I share concerns that the inexperienced will lean on it too heavily creating a devolution scenario. It will also take jobs not just answering questions, but setting them and evaluating responses. In such circumstances it may then not matter that AI writes bad copy. Progress?
I’m with you @weenie – it’s not taking my job! (#3moremonths) :o)
The immediate value I’m finding in copilot is it can write my ‘bullshit text’ for me in seconds, where it used to take much longer by hand. ‘Bullshit text” being the bullshit you have to write to satisfy certain parts of the work process. A great example would be coming up with SMART objectives for the following year to keep HR happy. Everyone knows it’s bullshit, but you’re still mandated to do it. Now it takes me no time at all. My line manager said they were so good, could they be shared with the rest of the team as a guide to best practice? Thanks copilot!
I know the correct response would be to correct the process so no bullshit was required, but we all know that’s not going to happen, so as a pragmatic work around, this is an excellent solution. Just solving the ‘bullshit text” problem must represent a colossal productivity improvement in most organisations right?
@old_eyes #32 – superb! Let us know when you find the compact stellar remnant!
@Rhino #25 – I use watchlists on the Hargreaves Lansdown app, they work well for me.
Thoughts on AI?
My employer is heavily invested in AI and generates substantial global revenue from AI consulting. I’ve been impressed by the rapid progress over recent months, it may not be a linear path for us as investors, but I believe it has huge further potential. I personally use AI much as others have described, including @Rhino’s #35 example, but for me it’s mostly simple prompts such as:
– “Please explore the extent of any overlap in holdings between WLDS and AVSG.”
– “Please review the attached portfolio and assess each asset class, its percentage allocation, and the individual holdings – commenting on strengths, weaknesses, potential improvements, and any key risks.”
– “Please provide a table of all Monevator articles focused on factor investing, including date, title, URL, and a 100‑word summary of each article and the topics raised in the comments.” (Apologies if this further reduces Monevator’s site traffic?)
My 2025 update?
After several years without changes, I made three adjustments this year:
– Reduced my overall equity allocation from 60% to 50% after reassessing sequence of returns risk. (Like @two_more_years, I’m nearing the finish line.)
– Increased diversification, despite already holding ten asset classes. (Is there such a thing as over‑diversification?)
– Trimmed US exposure by switching from VWRP to a blend of WGEC and XUSE, despite being positive about AI’s long-term prospects. (I don’t have @Delta Hedge’s deep analytical skills – I just don’t like the look of that concentration risk!)
This time of year, I catch up on investment reading. That means Monevator and SLIS articles, reading (studying?) McClung again, and updating the list of topics still to go through in the future – linker ladders, duration matching, and plenty more!
Thanks to the Monevator team for another year of excellent articles, and to everyone who contributes in the comments. Best wishes to all for 2026.
Whether it is AI or just concentration, the Mag 7 have made capital weighted world trackers extremely US and tech-heavy. They have performed well, but no longer answer quite the same question in a diversified portfolio. I don’t claim to have any expertise, but the software aspect of AI doesn’t appear to have much of a moat. And the current US administration is looking less friendly to these European eyes by the day. I retain developed world trackers, but have significantly reduced my reliance on them. Artemis Global and UK funds, along with Ranmore have reduced my US/Tech weighting. Yesterday, I bought a little Man Japan. Today, I swapped half of my SWLD trackers for TEGB, an equal weighted European fund that has performed pretty well over five years and rather better than SWLD over the past year. I may miss out on some “American exceptionalism” in the meantime, but I feel I have found a better balance, on a rug that is less likely to get pulled…