PaulRobinson

I'm currently weighing up a sabbatical (3 months), for a bunch of reasons.

A vacation (or holiday, or PTO, or whatever you call it locally), is great, but normally too short to properly reset. I often come back thinking "I could do with another couple of those right now". The main purpose for those for me is to travel and get away from home for a week or two and experience new things.

What I'm more excited about is taking a few months to basically do very little. Read, walk, no alarms in the morning, no Sunday night anxiety, no wishing it was Friday. Just wake up, do what feels right, don't work, don't do anything.

I'm in my late 40s, I have chronic health issues, and have symptoms of burn out, so even this might feel "too short", but I think it's such a valuable experiment to try, it's got to be done.

Gary Neville's mini-retirement hot take was the first thing I thought of when reading headline.

There is a pedagogical theory that states a student only really learns something when they apply it.

Not read it, not highlight it, not copy it out: apply it.

The only way to build personal knowledge, in my experience, is to do things.

And I only want to learn things in the context of doing something.

I read and favourite and bookmark stuff all the time to dig into for inspiration.

But then I choose to build something, and I assemble knowledge on a JIT basis and apply it. LLMs make this easier than ever before.

Everything else is theatre at best, procrastination at worst.

I never again want to work on a team full of people who drink a lot of energy drinks.

I'm not going to pretend I'm a saint and I only ever sip mineral water or kombucha (although I do like both), but that stuff sends people crazy in a long term way I've not seen before. I'd rather people were slightly tipsy after a lunch in the local pub than on that stuff.

I guess you also think the "Hand of God" goal, was also a legitimate and fair goal because something something Maradonna something something Malvinas something something god?

Look, it's football. We're all entitled to have opinions, state them with vigour and passion, but...

Step back. Is it fair to drop the "racist" slur on somebody for having a different perspective?

Referees do their best, and often get it wrong, but it seems to most people - including many neutrals - that refs got it wrong more times than not when Argentina was playing, and in Argentina's favour.

Is that a conspiracy? Well, Infantino definitely wanted Argentina in the final, so did he try and create that somehow? He seems pretty shady, it's hard for people to trust that he _didn't_.

But I'm prepared to accept it was just variance. The problem is, that variance undermined the competition for many people, in the same way hydration breaks did. It made it a worse World Cup for many fans.

I get that for an Argentinian, that seems unfair, but if this had happened to England's benefit, do you honestly think you'd be making the same accusations and would accept it if any criticism that came in their direction was met with slurs?

Peace. Chill. We're good. We're brothers in football. Let's hope we get another World Cup in our lifetimes, and that we get to enjoy it - and that means something different to what we just saw, for everyone's sake.

This is definitely an iterative improvement on what we have today. However, I think from a fundamental approach, you've just highlighted the pain of working with AI that I hadn't quite been able to verbalise until seeing this.

Go and have a look at the history of Microsoft OLE and OpenDoc and you'll see some of these ideas have been around for a while. When these ideas were first being shared in the industry (and yes, I am that old), here's the story that was being told:

In the future, you wouldn't buy "a word processor" or "an image editor", you'd buy components that could do specific tasks, and you'd bring them to your document/work as and when you needed. I'd have DTP components for layout, word processing components for spell checking, image components for resizing and colour balancing, and I'd be heads down in my fancy newsletter or report, and I'd be bringing functionality to the work I needed to do, without having to context switch and import/export things by hand.

The demo you've shown on the homepage is almost the opposite of this: I'm not bringing things into my work (fanning in), I'm starting agents to go and do lots of different bits of work (fanning out), that eventually I'm going to need to think about how to structure and fan in again.

I think coding interfaces kind of force us into a fan-in: here's my code base, now I want to improve test coverage, now I want to refactor, now I want to add this SDK, now I want to change this logic to use the SDK based on configuration params, now I want to improve my integration tests, and so on. I'm flowing through my work in a single asset (the code base), and able to bring agents and other tools to it.

Now, I'd like that, but for everything else. The UI I'd like allows me to build around the assets I am building.

Planning a trip? OK, I need to bring various agents along to help me plan an itinerary (in my calendar), based on location, budget, and preferences. It's not good enough to list the top 10 must-see places in Venice, I need them to fit into my calendar including travel time, and my budget - I want agents to help me plan that out. Ideally I want agents to help me and my partner both contribute to that based on our wants.

Need to do a report for the big boss? OK, I need to start with an empty report doc, and bring various sources of data, and take it through transformations to build the narrative/deck I need to write. Here's some excel data, here's an agent that can do some competitive analysis, here's another that is pulling quotes from real customers on social media, and so on...

Want to learn a new skill? We're going to need to have a framework for assessing where I am in terms of that skill (the "single asset" we're working on is "what I know"), and then agents to figure out the best way for me to fill gaps and reinforce what I already know.

Your demo is nicer than a chat box, but I think you need to think about the job that needs to be done, and how to bring agents to that job (perhaps even suggest those jobs - that could be another agent), not build a UI to just allow me to fire off dozens of agents that then have output sat in multiple windows across my screen.

- The number of teams expanding weakened the group stage to a point of being ignorable

- The extraordinary pricing of the tickets

- "Hydration breaks" to cater for TV advertising

- The US red card being rescinded after a phone call from an interested party

- Argentina's free pass throughout

FIFA has been corrupt since the days of João Havelange, although his stance was broadly to bring football to more countries and just needed a way to finance that (enter stage left, Coca-Cola, and stage right, Adidas and ISL), but it started getting really wild after him.

It's gone unchecked. There was talk of the English FA pulling out of UEFA and FIFA some years ago, as there was a perception the soul of the game had been tarnished by both orgs.

This is a good first step, and I hope FIFA gets itself sorted. I hope UEFA also gets its house in order.

    The moment external investors acquire ownership interests in FIFA competitions, football changes forever. Commercial return becomes a permanent obligation. Investor expectations become a daily pressure. From that moment onwards, every decision on the international calendar, every decision on competition formats and every decision shaping the future of football is no longer driven by what best serves the game, but by what best serves shareholders.

These words apply to any investment. It's worth considering as part of any business, including many of the businesses we're here to watch grow - the moment you have others involved as equity partners, the entire culture has to change. Do it consciously, or not at all.

This is cool, but do you know what was really cool? 2 years earlier, they made a TV watch: https://by.seiko-design.com/140th/en/topic/51.html

The only thing that annoyed me about HS2 more than "Oh, and it'll stop at Oak Common, and won't go to Manchester", was by that point they'd already flattened the Bree Louise.

OK, short version: no.

And logically this makes sense, because you can see the pub problem as a subset of locations within all locations, and all of these pubs are clustered around an equidistant point at ~72 minutes from everywhere in London. If such a location existed, it would have to be within that cluster, but somehow an average of > 12 minutes travel to all those pubs in that grouping. An eye ball test on a map tells you that isn't possible.

You might be asking "is there another location without pubs that is more equidistant", and mathematically that could yield a coffee shop, except in London it won't because the equidistant point for all of London and the equidistant pub basically are at the same point to within hundreds of yards, because of how public transport and public houses are distributed across all boroughs.

I have a readers card for the BL, and agree its a marvellous thing - I regularly head over there despite my earlier comment about getting depressed on Euston Road.

My preferred way to deal with this part of town: slip out of Kings X and into BL as fast as possible, do what research I need to be do in BL, then walk at speed to the Waterstones on Gower Street (which is not like any other Waterstones in the country, I love it), then if you have time, maybe head down to the British Museum and check out one or two of the rooms. Then, get out of there. If you want a pint, the pubs nearer BL are not as depressing, and you're now on your way to Soho anyway...

To contextualise my earlier comment though: I grew up in the North of England, and for many years I hated London. Every time I visited it felt oppressive and horrible. In 2005 I ended up needing to spend a week doing some client visits, stayed in a place near Warren Street, but had a couple of days off to go exploring in between things. Went to Hyde Park. Went to Greenwich. Went to South Kensington.

I realised I didn't hate London, (how can you hate being lost in the middle of Hyde Park of Kensington Gardens on a warm sunlit day?), it's just that my experience of London until that point was getting a train into Euston and then getting a tube to somewhere like Covent Garden, or Bank, or staying near Euston on Euston Road.

Long story short: my partner and I moved to Barons Court in 2012, and then Twickenham in 2017. Couldn't be happier.

But I still feel sad when I'm on Euston Road.

Logically, no, as they all neighbour non-pub locations. There's a coffee shop or restaurant within a minute or two of each of the top 5, from memory.

There used to be some UK celebs (David Baddiel, Clive Anderson, others...), who used to regularly eat in a restaurant near the Crown & Anchor after their weekly 5-a-side football games, which is why it stands out in my mind.

The problem - and I can't understate this enough - is that these are all relatively close to each other (naturally), in one of the most miserable parts of London you could spend time in. More than 4 hours in that part of town and I start to feel more noticeably depressed.

I get why equidistant is attractive, but you'll have a much nicer time if somebody is going to have a longer journey. So do yourself a favour: go out to Greenwich, or Richmond, or Hampstead. Go somewhere that is greener. Somewhere that's not on/adjacent to Euston Road.

When I was a teenager, I was a big fan of the show (and books), Red Dwarf. One part of the story describes one of the characters, Rimmer, preparing for an exam. He spends so much time preparing his revision timetable and improving it iteratively, over and over again, that he completely forgets to revise and the night before the exam realises he has a beautiful revision timetable, but no time left to revise. "Rimmer's Revision Timetable" was a shorthand in several places I worked for "we're doing too much preparing, not enough doing, let's get on and just do".

I once spent 9 months trying to figure out how a piece of software should be built - I wanted it to be perfect - while my bosses agonised over the delay. One night, in a bout of anxiety and just trying to get something to show my bosses, I sat down at around 7pm and just started riffing away in the editor. I ended up working through the night. By 5am it was done. By 9am it was in production. By 11am the first customers were giving us money. I didn't need to edit it again for 3 months. That was the moment I decided to ignore an old quote I'd been handed and had until the point believed had real truth in it:

"Give me six hours to chop down a tree, I'll spend the first three sharpening the axe" - attributed widely to Abraham Lincoln, but as the other saying goes "90% of the quotes you read on the Internet are misattributed to the wrong person" - Mark Twain.

There are some situations where the preparation is necessary. If your axe is blunt, it'll never cut down the tree, even in 6 hours. Getting it sharp in the first 3 hours means you might finish the task early, for less physical effort.

But more often than not, you're just procrastinating. You could go find another axe. You could find that the axe you have is just fine and gets your tree down in an hour rather than three hours of sharpening followed by half an hour of chopping. But by spending your time sharpening, and not chopping, you have the illusion of telling yourself you're working towards your goal, but you're not. And guess what, sharpening is a lot easier and a lot less sweaty than chopping, so it's more comfortable too - what's not to like?

YAGNI, Yak Shaving, axe sharpening, Rimmer's Revision Timetable, they all bring into focus different perspectives of the same problem:

Doing a thing that isn't the thing, isn't doing the thing.

Yes, if your editor config is distracting and you want to fix something, go find a quick thing and move on. Installing a tiling window manager and a global keyboard shortcut app of some sort might make things flow a little better, but spending days optimising is likely poor ROI.

Just try and do the thing. Build something. Stop planning to build something. If you aren't sure what to build, spend 5 minutes trying to build something and figure out what's actually in your way (hint: it's not your editor, it's not your lack of gdb knowledge, it's not that you're not using the hottest new programming language or that new agent/skill/model people are talking about), and find the possible quickest way to get back to making progress.

Yes, you need to think about prod. The more senior you are, the easier this comes naturally, but writing tests is doing the thing, not procrastination. Putting in operational and business metrics is doing the thing, not procrastination. Why? Because those are part of the thing. Your prompts are not part of the thing, and neither is your tmux config.

As an aside, LLMs can maybe help here (but don't over-engineer the harness): I have a side project where I feel I need to pick up a lot of theory on reinforcement learning, I could do with a refresher on partial differential equations, and I also want to learn some more advanced socket programming in C (trust me, all these things combine into a single thing).

I could have spent the next three months reading books and doing courses and then started on my project, but what I've done instead is get Fable to build out a curriculum where I build the relevant parts as I'm learning.

The first spec it came back with told me I had about 7-10 months of work ahead of me, but I could see (at least in the early stages which I understand better than the later stages), I'd be in a virtuous cycle of learning → applying → learning → applying → ... and within the first couple of weeks I'd have some basic harnesses that showed progress and have some of the theory down too.

This, I think, is a valuable way to structure the learning without procrastination. A personalised tutor who can help you fill in the gaps in your knowledge as you build, so you build the right thing and learn, but you're not just procrastinating up-front, feels good. And, if I get 3 months in and have enough knowledge to realise the next three months are not quite right, I will have the knowledge to re-prompt, re-draft and move forward.

I've thought about this a few times over the years, and the protocol is broken in ways that lead to poor UX - the whole thing became over-ran with crap. I do think it's possible to fix this with better client software, but nobody wanted to do that work, because everyone interested in building communities could make a lot more money taking VC funding and creating it in a walled garden that's now getting ID verification walls around it.

Tony Sale saved Bletchley Park. He was the instigator of the foundation and basically ran it single handed for a while. He believed deeply in it being preserved for the nation.

His main claim to fame before that was being Peter Wright’s right-hand man. Wright’s “Spycatcher”, talks of their joint exploits over the years. I got to ask Tony about it once, he was not too impressed Wright had done what he had done by writing that book.

Sure, but what you're not taking into account is that $1500/month is subsidised by VC money.

If Uber were asked to pay full price, they'd either get the same value for significantly more outlay, or keep their outlay capped but get significantly fewer tokens.

This would, I'm sure you'd agree, change the ROI outcome.

Zitron is arguing that if we all pay the true cost once the subsidies go, we'll either stop paying for the value we get today, or we'll find AI to be less useful than we do today, and those businesses won't be sustainable, and then <pop> goes the bubble.

I enjoy keeping an eye on https://isaiprofitable.com to see just where we are overall. Interesting reading if you like to think about such things.

I think both sides might be cherry picking the pieces that make their arguments the strongest - Pearkes has pulled out a few paragraphs, but there is a lot in Zitron's writing, so it's easy to select 6 bits you don't like/agree with or find fault with, but I'm not sure it undermines the rest.

I actually think Zitron would be better off focusing on those deals where there is clearly something weird going on (SpaceX IPO, Oracle/OpenAI/NVIDIA triangle, the neoclouds, and so on), than picking a fight with the one company that actually seems to be trying to play it straight in this sector (Anthropic), and I can see how he's covering the whole field with the same cynicism which might not be appropriate.

That said, I'm not entirely convinced that some of Pearkes' analysis deserves us all to align to the rosy bullish view he takes either.

Time will tell, it always does, but at least you've put forward some data unlike anyone else, so thank you for that.

He's not wrong about the numbers, and you may be seeing what you want to see a little, here.

Fable scoring OK on some benchmarks does not the mean the investment has financially paid off, because that's not how ROI works.

You're right that no bubble has deflated, but I think we can all see that a) there seems to be a bubble - I'm old enough to remember the dot-com era bubble, and this definitely feels like that, b) the only company clearly making a profit on AI right now is NVIDIA, and c) the net economic value of the sector as a whole (i.e. money out > money in), is still unproven

It's not even obvious to me as somebody who is up on comp-sci, that this isn't all hype, that the progress is/will be steady, and that reaching brain equivalent levels and beyond using these architectures will be economically viable.

I can create a perfectly good human brain in 9 months, train it in ~20 years, and have it pay off economically in a more proven way than [waves hands] all of this.

For some reason we've decided paying a hyperscaler the equivalent of a year's salary to do a job in a a day that a similarly paid and skilled human could do in a month is value for money.

You could probably give everyone in the US free healthcare for life and a free ride through college for less money than has been pumped into AI in the last 5 years, and have a more proven economic model.

Sure, time value is a thing, but given the error rates, the energy issues... this direction is not exactly a slam dunk as a net benefit, and I think that's all he's called out in any of the writing I've seen of his.

I also happen to agree with his take on Oracle, as it happens - they're only going to survive if it turns out they're a part of critical national infrastructure deep in the bowels of the US government. They look totally over-leveraged otherwise.

Can you provide evidence of where he's been wrong?

"We" tried this in the late 1990s. I remember writing an exim config to do this a couple of ways back in my ISP times:

1. "The recipient will not get this message until you authenticate yourself as a legitimate sender at this URL [...]" - resulted in people never getting mail from no-reply addresses they cared about (banks, e-com, etc.)

2. "This sender has sent you an email, with this subject and this first paragraph. Click here to whitelist, here to blacklist" - created a lot of churn on first setup that users didn't like as a UX

Baking it into the protocol - even now in SMTP and IMAP land - is doable, but would be resisted by email providers who, for example, make their money selling data to marketing agencies. Like, you know, Google, Microsoft, Yahoo...

I've seen other approaches proposed - even at IETF level - over the years, including digital postage stamps (sender pays to send email, your server gets paid, and may reject non-stamped/paying email), but it's the usual network effect: until it becomes widely supported it won't "stick". The stuff that has been adopted has basically been stuff that Google likes, period.

Devil's advocate argument: if they were giving you something in return, and that could be cash, but it might also be "you can have this content for free".

In the UK a few news sites have changed cookie banners to "you can accept and see this stuff for free, or you can sign up for a subscription, which would you prefer?". It's the only time I hit accept (and then clear browser history).

If blanket preferences from browser signals became the norm, a segment might open up where you would configure preferences and a data broker would make sure you get something in return for your data. At minimum it might force paywalled publishers to consider that as a "lite" subscription option.

I think you'll find that's Google and Meta.

Just step back and ask yourself what each side of that debate is trying to achieve and why. What is motivating them? Why are they motivated in that way?

Don't just recite what you "know", think, look, research, figure it out. It might sound good to have a one-liner like this in your back pocket, but do you really believe it after looking at the publicly available information that it is their real intention to conduct a "crusade to destroy the internet"?

As a general rule, you should take the train to Gatwick if you can, but this is reasonably priced and next time I'm flying from Gatwick I'll probably drive and do this - the train route for me is a bit fiddly, and the worst bit of any London airport for a local is long-stay parking. I'm much closer to Heathrow but the flights from LGW are so much cheaper, it more than covers two weeks of parking at this price.

> Or do they have to do that and eliminate their dependence on Boeing, Lockheed Martin, Northrop Grumman, Intel, AMD, nvidia, Apple, Mastercard, Visa, Google, Microsoft, AWS, Oracle, Facebook, Youtube, Netflix, Reddit, Twitter, Instagram, Juniper, Cisco, Qualcomm, Broadcom, Adobe and Uber?

Some of those are not like the others. Some are B2C and will face (indeed, are facing), tougher regulations. Some are B2B or B2G and there are already companies squaring up to take some of that pie. Many are ultimately reliant on ASML tech which I think Europe is prepared to weaponise in a trade war if they need to.

And yes, it'll take leadership, and constitution of the United States, or constitution of the physical body: one of them is taking that man out of the White House in the next 10, 20 or however many years.

But the tone has been set. Trust takes years to build and seconds to destroy. The direction of travel has been created, and while it might take a generation to get to an end goal, the process has started already.

US tech does not understand or comprehend the amount of damage they’ve self inflicted in the last couple of years. There will be a steady drip of this in years to come, I suspect.

Allows you to move stuff from capex to opex while remaining in compliance and governance lines.

They’re not idiots. A lot of thought went into it, and the idea it’s all just incompetent BS dreamt up on a golf course is madness.

The UK straddles the 55th parallel North. Due to the jet streams (and formerly, trade winds) carrying warm air from the Caribbean, it has often benefitted from a milder climate than other regions at similar latitudes: Alaska, Siberia, the Baltic states.

But this is wild. I am only in my 40s but I've seen the climate change significantly and it having a major impact on flora and fauna.

When I was a kid, English wine was a joke and now it's winning prizes because we have the same climate French vineyards did when I was a child. The idea of growing olives in the UK was laughable, but now in the South East of England it's starting to look commercially viable. Only 10 years ago I remember days with highs over 30C being rare, but we just had two weeks of them, accompanied by tropical nights (not dropping below 20C), for the first time I remember.

And now... bananas. Which is, pun intended, utterly bananas.

It feels too late to stop the worst of the damage, but the idea of this accelerating and getting worse is just astonishing to me, that within my lifetime I'd witness such a huge swing in climatology.

I've been on loops for people with a PhD in ML and associated fields to work on R&D roles for the FAANG I work for, so I think I have more of a chance of deploying this view in reality than most.