AI's Swiss Cheese Opportunity, Peak Social Media and Energy Bills That Disappear
Shift Notes 05: LLMs don't get addicted to dopamine, we may have passed peak social media and Octopus is eliminating energy bills through brand strategy
This month our usually independent sections on #1 AI/tech transformation and #2 algorithmic culture/creator economies are doing a ‘collab’ - simply because the lines between the two are so blurred currently.
In a conversation with Azeem Azhar, Cloudflare CEO Matthew Prince makes the argument that the current breaking and remaking of the internet’s business model at the hands of AI will lead us back to a better content environment, one that’s “more local, more unique, more creative.”
Then we’ve got OpenAI and Meta launching the latest capabilities in AI video creativity through the paradigm of social media. Which seems to argue against Prince’s optimism and suggests a future of banal call and response ‘creativity’ able to fulfil the basest needs of the algorithm.
The FT runs the numbers and comes up with evidence that we may have passed peak social media, suggesting Altman and co are previewing their shiny technologies at a party the cool kids already left.
So many moving parts as so much money pours into a technology that doesn’t yet know quite how it’s going to deliver a return on investment, the answers aren’t at all clear. But the change and creative destruction/reconstruction is coming at such a pace you can’t help but be excited and a little hopeful that some meaningful change shakes out.
And if you’re here for the climate story: we dig into how Octopus Energy is proving that renewable abundance can bring down energy bills - not through subsidy, but through brand strategy.
Thanks for sticking with us on the somewhat intermittent schedule of late. We hope you find these thoughts useful, human, stimulus!
#1/ Acceleration Gap vs #2/ Media Shift
AI/tech transformation vs algorithmic culture/creator economies: a ‘collab’ looking for where AI’s business model is going to come from and if there’s a dynamic in play that might save us from both ‘slop’ and ‘rage bait’
LLMs don’t get addicted to dopamine
Talking to Azeem Azhar’s Exponential View, Cloudflare CEO Matthew Prince makes a case for a better future for the internet.
He compares the value of deals struck with LLMs by Reddit versus traditional publishers like WSJ, FT and NYT. Despite both holding the same amount of data (in AI currency: ‘tokens’), Reddit drove a deal with OpenAI worth 7x per token that of the NYT.
Why? Whereas traditional media titles, despite different story ‘angles’ usually informed by ideology, roughly cover the same big stories. Reddit does niche volume around every conceivable quirk of interest imaginable. It’s the long-tail of data tokens. Much more valuable to AI.
They extrapolate from here: if Google’s model of clicks as value exchange is destined to die at the hands of AI-enabled ‘answer models’, then the zero-sum, clickbait content of Huffington Post et al is also dead.
If AI systems offer a model for the whole of human knowledge, they can also demonstrate where the holes in the Swiss cheese exist:
“The business model of the web is going to change, no matter what... because answer engines are coming and they’re better. What I hope it changes to is one in which we reward content creators that create content that fills in the holes in that cheese.”
If we’ve inadvertently degraded our own experience of the internet by succumbing to our worst tendencies - always clicking on the angry, divisive, shocking invitation - then maybe we’ve found an antidote in AI, which avoids this fate through the very fact of its inhumanity.
Specifically: the absence of neurological impulses. As Azeem says, “LLMs don’t get addicted to dopamine”.
Inside the collapse of the internet economy (and what comes next) - Exponential View
Will AI drive us all insane?
It’s easy to envision, as many are, the phenomenon of AI slop - synthetically generated words and images masquerading as journalism, science, footage, news etc - resulting in once stable knowledge bases online being replaced by increasingly slippery notions of truth. Slippery notions of reality, even.
But ‘Truth’ has been in decline longer than we’ve had ChatGPT. That phenomenon has been more closely bound to another internet technology, social media, whose business model rewards impact over provenance.
In a recent newsletter, marketing industry trends consultancy Contagious joins dots between Cory Doctorow’s characterisation of the ‘enshittification’ of the internet - in which platforms first benefit users to grow audience, then abuse users to court business, and finally abuse both to maximise profits - and AI as an accelerant of the trend. “Enshittification on fast-forward.”
You can track the thought process here. Both Meta (with the newly announced AI feed, Vibes) and OpenAI have plans for social platforms that will embrace AI-generated video.
In 2023 Doctorow not unfairly described Facebook as “terminally enshittified, a terrible place to be.” What are the odds that AI-generated content improves this state of affairs?
Discussing the Sora 2 launch from OpenAI and Vibes on Prof G Markets, Ed Elson speculates both companies are “too obsessed with this idea that ‘we have to be social media, we have to put [AI video] in some short-form content form.’”
“We’re already seeing the backlash where people are correctly saying ‘this is just an endless stream of slop’... they’re photorealistic, they look pretty amazing but... at a certain point it becomes very boring.”
Optimists may wish to intervene: the assumption we’ll all just put up with this and keep flicking is surely a little uncharitable.
We prefer to cite the findings of FT data wonk John Burn-Murdoch, who makes a quantitative case that we may be passing peak social media (although in an apparently major oversight, TikTok is not included in his data!?).
That 2025 might be the year when social media became a “gaudy backwater of the internet inhabited by those with nothing better to do.”
He parades a compelling set of data that shows time spent on social media in decline since 2022:
And even draws on survey data to illustrate a theory for why we may be turning away. As he puts it, social media has undergone a “warped evolution from a place where people swapped updates with friends and family, to one with less and less human-to-human interaction.”
A format once designed for human connection has become an undemanding branch of entertainment media. And it seems we like it less as a result.
(Sad to say, North Americans are the only significant global group bucking that trend.)
So, is a toxic combination of social algorithms and AI-generated content going to send us all mad? From this evidence, it seems like most of us have the presence of mind to vote with our feet.
What may send us mad though is the gradual creep of synthetic knowledge. Of AI as the primary way we generate and validate knowledge.
This has been variously investigated since the emergence of LLMs, one notable example being a Nature magazine essay last year that claimed ‘AI models collapse when trained on recursively generated data’.
The researchers’ experiments showed that AI models trained on data generated by previous AI models rather than original human content undergo “model collapse” - progressively losing rare information and diversity until they converge on repetitive, narrow outputs that bear little resemblance to reality.
As we’ve commented here before, AI is only as good as its data and it’s already used most of that up.
Given the tendency of AI models to fail when fed on their own outputs, like photocopies of photocopies, could rapid adoption across all knowledge-based endeavour threaten the same fate for how we produce and trust knowledge itself?¹
Links:
Have we passed peak social media?, FT
AI models collapse when trained on recursively generated data, Nature (2024)
The Great Sloppification of OpenAI | Prof G Markets
Pluralistic: TikTok’s enshittification (21 Jan 2023), Cory Doctorow
Cute fluffy characters and Egyptian selfies: Meta launches AI feed Vibes, The Guardian
Sora 2 is here, OpenAI
Quick Hits
Lovely insight into Rankin’s ongoing experiments with AI here, in this Instagram commentary on his ‘Faik’ project. What makes him such an exciting creative voice in this debate is his open-minded, balanced approach - he wants to explore what AI ‘can’ do rather than fixating on what it ‘can’t’. Increasingly, he combines his own photography with AI, in one project turning his photographs of flowers into ‘flesh flowers’.
For a handle on what’s happening at the apex of AI’s capital arms race, the Unhedged podcast has a useful state of the nation. Rob Armstrong and John Foley discuss how Nvidia (the infrastructure seller) is taking equity in its customers (OpenAI, xAI, Intel, AMD, Oracle etc) to help them keep shovelling cash into the compute race.
#3/ Zero Culture
Sustainability pressures reshaping consumer values and markets
Octopus Energy’s Case for Energy Abundance: What a brand-led approach to renewables looks like
In April we discussed why cheap renewable energy is so difficult to achieve - the complexity of the energy market leads most people to abandon hope of understanding their bills. Which means they fail to demand an explanation for why, when renewable energy is cheaper than fossil fuels, this isn’t reflected in what they pay.
Back then, we asked: what if a progressive energy brand actually behaved like a brand - creating value through honest dialogue, innovative pricing models and advocacy for better government solutions?
Turns out this is happening in plain sight at the UK’s biggest consumer energy provider: Octopus.
Speaking to Bloomberg’s Zero podcast, Greg Jackson, CEO of Octopus Energy, says the energy transition isn’t just economically viable - it’s already cheaper. But unlocking that value means rethinking everything from grid infrastructure to consumer engagement. The brand insights matter because they reveal how to translate technological possibility into consumer value.
Treating consumers like people, not meters
While regulation mandates bill formats, Octopus attaches videos in which Jackson explains current pricing, available support and how to access free electricity during abundant renewable periods.
“In the 21st century, it’s kind of crazy that we still expect a bill to be the same as when you had clerks manually typing them out 50 years ago.”
This reframes energy from an opaque fixed cost to something consumers can actively engage with. As he notes: “Imagine if the only way you could save money on food was by eating less. That would be a bad thing.”
The infrastructure rethink: building less, using more
He challenges the consensus on grid build-out. Every piece of grid infrastructure is designed to handle peak load - potentially just seconds per year - at a cost of billions. But with battery costs plummeting from $1,000 per kilowatt hour a decade ago to $52 in recent Chinese auctions, everything changes.
With electric vehicles projected to represent 75 gigawatts of battery capacity in the UK (versus the 50 gigawatt electricity system), the infrastructure is arriving as a byproduct of transport electrification.
“If we can shave those seconds off by storing them in a battery and avoid the peak, you can save billions in grid build.”
He concludes: “Minimising fixed costs should be a mission for everyone in energy policy.”
Products that eliminate bills
Octopus is already operationalising this. Their BYD partnership offers a fixed monthly fee whereby customers lease an electric vehicle and get 12,000 miles of free driving annually. Octopus fills the battery during cheap periods and sells electricity back to the grid during expensive ones - the arbitrage funds the driving.
The same logic extends to housing, where they work with builders to create homes with no energy bills for five to ten years, using solar panels, batteries and heat pumps.
What this means for brands
Octopus is creating value in a commoditised sector through consumer dialogue, product innovation and systemic advocacy rather than hiding behind regulatory complexity.
#1. Consumer engagement matters more as systems grow complex. Transparency and education unlock value in distributed, dynamic systems.
#2. Technological convergence creates opportunities to reimagine business models. When batteries become cheap and abundant, energy economics fundamentally shift.
#3. Western energy policy’s path dependency - building more transmission to connect centralised renewable generation - may be the energy sector’s landlines moment.
Jackson’s observation deserves attention: “We’ve been focusing on quite a small slice of the pie” by concentrating on electricity generation rather than electrification of end use.
The real opportunity lies in rethinking how we consume and store energy, not just generate it.
When incumbents defend yesterday’s infrastructure, there’s space for those willing to build for the future.
This analysis is part of 33_Zero’s ongoing coverage of how sustainability pressures, technological acceleration and media transformation are reshaping brand strategy. For workshops on contextualising your brand against these macro forces, contact jamesp@33seconds.co
About 33_Zero
33_Zero works with brands large (AWS, Oxfam) and small (Agronomics, Ivy Farm) on brand and comms. Our clients recognise that unprecedented change needn’t be a threat but an opportunity. We help your brand show up and participate in this new reality.
Email jamesp@33seconds.co or subscribe and DM us here to find out more.
Footnote