Not everything needs to scale.

// The Situation

I'm watching and hearing a lot about Claude at the moment - Claude being the AI alternative to ChatGPT, not my mums old next door neighbour (he wasn't even called Claude, his name was Richard but for comedy effect, I'm calling him Claude)

I've infected my own Youtube algorithm with repeated videos around Claude automations and the news is rife with Anthropic taking the American government by storm because they called Claude 'a national security risk' - I've even gone so far as to download Claude onto my Mac to have a play and I'll be honest, I'm blown away.

Now, I'm not a novice with LLM's - far from it - I use ChatGPT quite regularly to organise and sort data, reframe words into a more cohesive structure and get it to write and edit long passages of text for work purposes - it's something I've grown to love and understand and I'm able to create projects, customGPT's and all the like but when I downloaded Claude - I thought it would be the same.

It's not - there are so many layers to understand with Cluade and I'm sure it's the same as ChatGPT under the hood but I feel like Claude is a lot more powerful and has the groundworks in place to be really useful.

The biggest thing (which is being addressed) is that A.I LLM's are racing for adoption by the many and it echoes of the VHS/Betamax wars of the 80's - for those of you who are under 30, this might also apply to CD's and Minidisks or iPods and Microsoft Zunes.. pick your fancy.

They are all looking to be the model for the masses and therefore, capture market share but Claude has done something cool - it's created a skill within it's platform to take your chats, your thoughts and the memories within ChatGPT and 'move them' into Claude - it's becoming the place to transfer the data too.

When you change (which I never have, and never will) from Apple to Samsung, there is a bit of a faff in moving all your data etc - the company you're leaving want to make it as hard as possible and the company you're joining want to make it as easy as possible - Claude are in the latter camp.

// The Scale

I did that this week - I created an export of all my ChatGPT data and downloaded it and then, I used a skill within Claude to unpick all that data and make sense of it.

It took me about 20 mins of getting Claude to run the scripts in the background and get everything aligned but it did it and Claude now knows who I am, how I interact with it, what my knowledge level is and everything I've ever asked for help with (even how to lace up my shoes in a particular way) - it's quite remarkable the level of data that these LLM's now hold.

But, once Claude had run the script and sorted all my ChatGPT data - it started to offer to automate tasks which I carry out frequently.. it offer to become a 'co-worker' for me so that we could SCALE.

SCALE.

It wanted to help me SCALE my operations.

But what if I don't want to scale the things it was asking me to scale? What if, the videos that interest me the LEAST on Youtube are those that offer to help me to scale the things that I do regularly - the things that make me human.

It offered me the 'Top 3 Co-working' opportunities - the first being the CustomGPT that we created at Yammayap to generate 'generic' social post ideas - it offered to run this task every 2-4 weeks and deliver those results to me.. well, thats Grace out of a job. Nope. Don't want to do that..

*for reference, Grace is our marketing apprentice.

The second - as it could see that I used to use ChatGPT to write my newsletter preview text and subject lines - it offered to generate subject lines for me on a weekly basis from which I could 'write a blog' - but that defeats the object and the purpose of this blog - I (YEP, ME) write the blog and then ask an LLM to shortcut the SEO and newslettery bits for me (the bits that I don't really like or have to do).

Finally, it offered to run an automation EVERY YEAR to remind me about updating the 'seasons logic' in my football app - I need to update the season every year and it offered to remind me - I really don't need reminding about that either..

// The Obsession

LLM's are supposed to be here to make our lives easier and save us time and how I use them DOES save me time but they definitely shouldn't be replacing the thinking we do.

To forget is human, to create is human and to use an LLM to generate generic posts, well I guess thats human too.

But the AI world (and the tech world around it) is becoming obsessed with scaling operations at a pace which just feels unsustainable - how fast and how far can we go? At what point does the world stop and realise that the rate of change is so fast that actually we can't keep up and it's going to be waiting on us.

Everyone is talking about scaling - it's nothing new because I remember during COVID when I found a group of people who wanted to 'scale' their instagrams and gamed the system to attract a ton of new followers - but there was a finite amount of resource available to do that and it would only work if you had the talent.

Scaling with an LLM requires very little talent and it means the 'cream' won't rise to the top - it's going to be churned up in amongst everything and everyone else.

This newsletter does not need to scale - it's my creative outlet and I am very privileged to be able to write it every single week to a select amount of people who (I think) enjoy my writing. If I began to 'scale' it and increased the output or increased the volume of writing, it's lost its spark.

I liken my newsletter to a home cooked meal - if you're cooking up a roast dinner every Sunday and it's enjoyed by a few people - its very hard to scale it, you can of course, you can scale 'cooking' that home cooked meal but what it becomes is a canteen and you lose the personality of a home cooked meal, you lose the intimacy and you lose the fact that you're doing it for a select few.

Yes, you might be able to feed many, many more in your canteen but so what? The enjoyment was the cooking and the serving - not the 'scale'.

// My Worry

I think I'm worried for my children in this new AI world because the rate and speed of change is such that the things they're learning now will be outdated in a year or two because everything's 'scaling'. We're moving at such a pace that the home cooked meal will become a thing of the past and the appreciation of the short video, the impromptu message and the voice note will be lost to volume.

Volume will be the answer to everything in company's pursuit of 'scale' - messaging at SCALE, video production at SCALE and mass marketing - all done to reach as many people as possible.

I don't think we need to scale - I think we need to be the ones who reject the scaling model and stick with the personal.

The connection we feel when you're speaking to someone one on one and being around them means more than the 100 pieces of content that you could produce with your new LLM friend and the automations saving you more and more time (Claude told me that these automations could save me 4-5 days per year.

Those 4-5 days are 4-5 days of learning, connection and mistake making time that I will never get back - optimising and forever looking for the next thing to optimise until we've outsourced our entire existence and decision making capability to an LLM in a vein attempt to get back more time.

I'm happy using LLM's to help me but I don't want them to live for me. I want to do the living and they can just come along for the ride and if we 'scale' then great but stop telling me and selling me the dream.