Concentrating intelligence🧮 no planetary boundaries for the rich💸 AI for small states🌏 granny scambaiter👵 #paybacktime💰 please die, human🤖❌ openstar💫 end of the Line🏗️ #2024.46
Deepfakes on a plane
Welcome to this week's Memia scan across emerging tech and the exponentially accelerating future. As always, thanks for being here!
ℹ️PSA: Memia sends *very long emails*, best viewed online or in the Substack app.
🗞️Weekly roundup
The most clicked link in last week’s newsletter was: Gary Marcus’ AI rumor you won’t want to miss.
🗣️Upcoming speaking
(I’m currently taking speaking bookings for 2025… details here).
I’ll be talking on Harnessing Exponential Technologies for New Zealand’s Creative Future in a couple of weeks at the Generative AI and Creative Economies in Movie Production mini-conference in Te Whanganui-a-Tara on 5th December… maybe see you there!
I’ll be showcasing some of the best examples I’ve seen this year of AI short filmmaking (still waiting for that first feature-length AI movie, tho…) I’ll also be exploring the rapid convergence between AI and Hollywood: Lionsgate X Runway, James Cameron joining the board of Stability AI… and this week’s unveiling of work by Metaphysic.ai on Robert Zemeckis’s new film Here - they used AI to generate nearly an hour of photorealistic, de-aged human faces — across multiple ages — in realtime for actors including Tom Hanks, Robin Wright and Paul Bettany. Impressive results:
🔄ICYMI
I spent Sunday with Claude exploring what the 2nd Trump presidency might mean for the world outside the US and sketching geopolitical scenarios on how things could play out. In short: pretty much every likely direction points to multipolar alliance-building and increased investment in national resilience. The next 4 years in the White House seem on course for sustained limbic-hijacking policy chaos… but on the whole those of us *not* blue-pilled by US exceptionalism can probably afford to take a dispassionate external observer posture for now. (I could be very wrong…)
🌐Strategy note #5: Trump presidency 2.0, some geopolitical scenario building with AI
🌍🌡️🌊Planetary boundaries roundup
(Pulling this forward from the Zeitgeist section this week… so much important climate and biodiversity developments that are just getting buried under the “Trump just nominated *WHO*!?!” cacophony… news algorithms need a new objective function!)
🎭Cut the theatrics
The likely ill-fated UN COP29 climate conference in oil-rich Azerbaijan entered its 2nd week. UN climate chief Simon Steele told country representatives in Baku to “cut the theatrics and get down to the real business this week”. (Link to roundup of Week 1 from The Guardian).
Meanwhile new US President-elect Trump looms over the conference with long-held threats to pull the US out of the Paris Agreement — but the Economist writes that there are obstacles to Donald Trump’s attempts to reverse climate progress.
💰#PaybackTime A new website cop29.com sprung up which is not what you might expect:
💸No planetary boundaries for the rich
A new study, Keeping the global consumption within the planetary boundaries found that up to up to 91% of planetary boundary breaching can be attributed to the top 20% of global consumers:
“We show that 31–67% and 51–91% of the planetary boundary breaching responsibility could be attributed to the global top 10% and top 20% of consumers, respectively, from both developed and developing countries. By following an effective mitigation pathway, the global top 20% of consumers could adopt the consumption levels and patterns that have the lowest environmental impacts within their quintile, yielding a reduction of 25–53% in environmental pressure.“
So, polluter pays, right…? (Yeah, right…)
🌡️⏳❗1.5 degrees … when, not if
Posted this week by the European Copernicus Climate Change Service:
“Are we already breaching the Paris Agreement this year? No, but... While it is almost certain that 2024 will be more than 1.5°C above the pre-industrial level, the limit set by the Paris Agreement applies to a long-term average of at least 20 years”
How much further can that line go up…?
🦠Plankton extinction?
University of Bristol researchers have warned that certain marine species, particularly plankton, could face extinction within the next century due to their inability to adapt to the unprecedented speed of current global warming.
💧📉Freshwater levels drop abruptly
Observations from NASA-German satellites have found evidence that Earth's total amount of freshwater dropped abruptly starting in May 2014 and has remained low ever since.
From Phys.org:
“From 2015 through 2023, satellite measurements showed that the average amount of freshwater stored on land—that includes liquid surface water like lakes and rivers, plus water in aquifers underground—was 290 cubic miles (1,200 cubic km) lower than the average levels from 2002 through 2014, .... "That's two and a half times the volume of Lake Erie lost."
…“Global warming leads the atmosphere to hold more water vapor, which results in more extreme precipitation… While total annual rain and snowfall levels may not change dramatically, long periods between intense precipitation events allow the soil to dry and become more compact. That decreases the amount of water the ground can absorb when it does rain.
The problem when you have extreme precipitation…is the water ends up running off… instead of soaking in and replenishing groundwater stores. Globally, freshwater levels have stayed consistently low since the 2014–2016 El Niño, while more water remains trapped in the atmosphere as water vapor."
This does not bode well.
🌀Four typhoons in ten days
That was the Philippines last week - the most active November on record.
Six typhoons within 3 weeks and at least 160 deaths. Doesn’t get covered in the mainstream news like Florida (etc) does…my thoughts go out to my few readers in .ph .👐
(Video: @WxNB_)
🗺️Extreme weather, mapped
UK-based climate website Carbon Brief has mapped every published study on how climate change has influenced extreme weather: the latest iteration of the interactive map (screenshot below) includes more than 600 studies, covering almost 750 extreme weather events and trends:
“overall, 83% of the events and trends included in the map were found to have been influenced by human-caused climate change.“
(Further reading: Q&A: The evolving science of ‘extreme weather attribution’)
(I told you it was quite a week…but sure, seed oils are important, too…)
📈The week in AI
My ever-longer roundup of the week's AI news and releases.
🧮Concentrating intelligence
An important working paper from US non-partisan economic think tank NBER: Concentrating Intelligence: Scaling and Market Structure in Artificial Intelligence. Probably the most comprehensive overview of the market dynamics of AI right now. Worth a read of the whole thing but for the time poor, here are the key bullet points I took from Claude’s summary:
The market for generative AI was estimated at only US$3bn in 2023 but is projected to underpin 7-10% of global GDP within a decade
Current market has become very dynamic with fierce competition arriving within 1 year - now 16 different AI labs have produced models surpassing the original GPT-4
Leading providers include OpenAI, Google DeepMind, xAI, and Anthropic
OpenAI/Microsoft currently dominate with 69% market share despite similar capabilities among top models
Compute is a major bottleneck, with NVIDIA controlling 98% of the data center GPU market
While current competition is fierce, there are strong forces that could lead to market concentration, requiring careful policy attention to maintain competitive markets while ensuring safe and beneficial AI development.
Good analysis in here to help understand the winner-takes-most investment dynamics at play in the AI industry right now. It’s a US-centric take on regulation… one question I have is how do smaller states mitigate against concentration of US AI providers (especially when they’re prevented from using Chinese tech…)?
(See also “AI for small states” below).
📉Scaling wobbles (cont’d…)
The rumour that the next generation of AI models are hitting scaling limits, covered last week, is well and truly out in the open. But “scaling” of performance can be measured across multiple dimensions - not just the benchmarking in the post-trained large AI models. The overall functional performance of the end-user AI applications themselves is likely to come to the fore.
A few wider perspectives on “scaling” AI:
With the introduction of its new o1 “reasoning” model series, OpenAI introduced us to another dimension of scaling performance: inference (vs. just training). It also provides more flexibility to shift cost structures from fixed (training) to variable (inference). Does shifting workloads to inference time use more compute overall? A new technical paper from researchers at Tsinghua and Carnegie Mellon Universities explores Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for LLM Problem-Solving:
Inference scaling laws …the error rate for each model size decreases steadily as inference-compute increases, and converges at the end. (Vertical axis is Log scale) (The upshot of which is… at least for these narrow test cases, inference seems to provide an alternative scaling dimension to just training larger and larger models… but likely at a cost of time performance?)
Scaling realities Nathan Lambert inserts further nuance into the analysis:
“Both narratives can be true:
Scaling is still working at a technical level.
The rate of improvement for users is slowing. Especially the average user.
…OpenAI has backed itself into a corner with its messaging. Their messaging around AGI soon made people expect this to be in a form factor they know. Really, the AGI they're building is not contingent on GPT-5 being mind-blowing. It's a system that will use GPT-5 as one part of it. It’ll be agents, like computer use, or even things that are entirely new form factors — agents that are dispatched on the web. These views of “AGI” are entirely separate from what ChatGPT is“
Meanwhile Microsoft AI CEO Mustafa Suleyman identifies another dimension for scaling performance: “memory” (eg context window size - the information the AI can hold in memory before responding to a prompt). In the clip below he says that next year’s AI models will come with “near infinite memory” - effectively allowing you to have a personalised agent assistant which remembers *everything* about you…) (6:55 in)
And likely one further dimension which I’m expecting to advance significantly: user interface. Simple extensions such as Artefacts (Claude) and Canvas (ChatGPT) already add another step up in terms of usability. ChatGPT Advanced Voice Mode is another. Claude computer use another. We should see significant UI innovations in the next year even if the models themselves may not be shifting upwards at quite the same rate.
🌏AI for small states
Singapore’s Digital Minister Josephine Teo spoke at the recent Financial Times Future of AI Summit, sharing insights about how Singapore’s National AI Strategy 2.0 (released Dec 2023) on how smaller countries are responding to the AI race happening in Silicon Valley. Insightful conversation with the FT’s Geoff Dyer, some key points extracted from the transcript:
Singapore is focusing on practical, enterprise use cases rather than developing frontier AI models
Aiming to integrate AI across all economic sectors, not just tech
AI implementation examples mentioned include Singapore Airlines, DBS Bank
Also government agencies: 65,000+ public servants are using a secure ChatGPT version (eg for meeting transcription…)
Workforce:
Focus on upskilling existing workforce rather than just AI specialists
Emphasis on human-AI collaboration rather than replacement
Goal to train significant portions of 3.5 million workforce in AI skills
Of note: last month, the governments of Singapore and Rwanda released the AI Playbook for Small States, in consultation with government agencies from the informal grouping Forum of Small States (FOSS). Nothing world-changing in here… but lots of practical examples of how smaller countries are attempting to deploy and gain advantage from AI. I’d love to see more of this.
💻China building SV labs
Also an interesting story from the FT that Chinese tech giants Alibaba, ByteDance and Meituan have started building AI teams in Silicon Valley, looking to poach top US talent. How will the incoming Trump administration regulate against that, I wonder…?
🤖❌Please die, human
This week’s clickbait AI was an apparently innocuous chat conversation with Google Gemini by a graduate student suddenly took a highly unexpected turn:
(WTF?! You can still access the original chat thread here after the user published it).
Google were clearly a bit perplexed and put out a vanilla message…
"Large language models can sometimes respond with non-sensical responses, and this is an example of that. This response violated our policies and we've taken action to prevent similar outputs from occurring."
But then silence… very strange…. there are whole Reddit threads trying to work out what happened…. no-one seems to know exactly.
Some sleuthing on X could point to it being a sophisticated hack… @lefthanddraft seems to think it may be someone interfering with two different Gemini conversation threads using Chrome developer tools… but didn’t manage to replicate. Google’s lack of any further response so far seems to indicate it was just a glitch in the model itself…. so, pretty weird one.
👵Granny scambaiter
UK telco O2 and agency VCCP launched an anti-scam campaign featuring Daisy (dAIsy, geddit?), a lifelike AI "granny" created to engage scammers in lengthy conversations, preventing them from targeting real victims.
On the topic of scams, here in Aotearoa The Spinoff’s Duncan Greive examines Meta’s role in hosting paid ads leading to recent banking scams…: Banks are grappling with a scam ‘crisis’ while Facebook profits from it.
☢️Human in the loop
News from the sidelines of the recent APEC Summit in Peru: the US and China agreed to keep humans in the loop (rather than delegate to AI) when it comes to firing nuclear arms…well, phew😬. (Expect blurred boundaries of where AI reaches right up to the decision to press the button — one dark scenario is an AI posting a highly crafted series of tweets targeted directly at Trump’s brain circuitry and …. boom…💥)
🧠AI Research
Roundup of what I’ve caught up on this week:
Alphafold open Google Deepmind open sourced the code to Alphafold, the Nobel-prize-winning tool for modelling protein structures, for academic use.
AI literature reviews Can AI review the scientific literature — and figure out what it all means?
“But most AI science search engines cannot produce an accurate literature review autonomously…Their output is more “at the level of an undergraduate student who pulls an all-nighter and comes up with the main points of a few papers”“
GPT-4, but for DNA? Evo is a groundbreaking 7-billion-parameter genomic foundation model trained on 2.7 million prokaryotic and phage genomes. In testing, Evo achieved zero-shot function prediction across DNA, RNA, and protein modalities including successfully creating experimentally validated synthetic CRISPR-Cas molecular complexes and transposable systems:
“…representing the first examples of protein-RNA and protein-DNA codesign with a language model.“
Thread from @IterIntellectus hypes it up some more:
Clearly we’re on a path to AI accelerating bio-engineering… who’s ready?
Stanford Virtual Lab: a team of AI scientist agents (AI chemist, AI reviewer...) led by an “AI professor” with feedback from human scientist. The Lab created new nanobodies that were experimentally validated to bind to recent Covid variants.
Surgical AI robot trained using imitation learning Researchers at Johns Hopkins University demonstrated a robot, trained to perform surgical procedures by watching videos of robotic surgeries, executing the same procedures—but with considerably more precision.
🆕Releases
And a race through this week’s latest AI releases…
Apple’s new M4 chip is getting some impressive reviews for price/power/performance on AI tasks compared to Nvidia RTX:
Google dropped their latest Gemini model and it went to the top of the LmSys Arena, level with OpenAI’s GPT4o-latest:
OpenAI looks like GPT4o V.Next is on its way soon?
Actually spotted in the wild:
OpenAI released desktop tools integration on MacOS for Plus users, coming soon on other platforms and subscriptions. This increases the power of these AI agent tools significantly… no more copying and pasting back and forth…
Chinese AI lab 01.ai (OK my head hurts from the naming…) spent only US$3M compared to OpenAI's US$80M to US$100M to train a GPT-4-scale model …with just 2,000 GPUs.
The Qwen team at Alibaba Group released Qwen 2.5 Coder 32B model as open source - beating GPT-4o across a raft of evals and giving Claude 3.5 Sonney a run for its money as well. Download it to your GPU farm today…
LLaVA-o1 an open source visual language model capable of reasoning:
AWS is the latest tech giant to release an open source multi-agent framework
Windsurf IDE … a new competitor to Cursor, VS Code and all the rest… fun (if a bit cringe) launch video:
Nous research launched the Forge Reasoning API Beta, an advancement in inference time scaling that can be applied to any model or a set of models:
AI4Finance Finrobot is an open-source AI agent platform for financial analysis using LLMs
Magnific Mystic SREF AI image style transfer for 1K, 2K and 4K resolutions:
(Too many sexualised female images in these demos…)
Runway Gen3 Video To Video now goes to 20 seconds. Here’s an example of what you can do with it (via @Uncanny_Harry , who doesn’t look like thi in real life!)
🔮[Weak] signals
Non-AI tech signals from near and far futures...
🔇Stop it now, pleeeaaasssee
Europeans spend 575 Million hours per year clicking cookie banners… GDPR is such a productivity killer!
🛠️Useful OSS
Debarghya Das (@deedydas) asked himself “How can I find and understand the top 1000 useful Open Source Software (OSS) Github repositories?” and came up with this tool:
↔️Stretchable
South Korean technology conglomerate LG unveiled its latest stretchable display prototype - able to expand by up to 50%, more than doubling the previous record of 20%. I sense a new clothing trend…
🦿🛞Dog on wheels
OK, this isn’t scary at all in a military context…🫣 Chinese firm DEEP Robotics unveiled their new quadruped robot called Lynx that combines wheeled mobility with traditional four-legged locomotion - on all terrains. By far the most advanced robot agility I’ve seen demonstrated so far…
Deep have a number of other models in production and also a humanoid in development… not long to go now until these devices start making their way into our everyday environments. (Deep envision Lynx being used for use cases including search and rescue missions, security inspections, mapping and exploration of challenging environments and equipment transport.)
(video: @YouJiacheng)
⚡Energy equations
Two data points on renewable energy:
Low Tech Magazine explores how:
“Interesting possibilities arise when you combine old technology with new knowledge and new materials, or when you apply old concepts and traditional knowledge to modern technology“
Its website runs on a solar+battery powered server located in Barcelona, and goes offline during longer periods of bad weather. It used 2.79W of power last time I checked — and last night I got my HTTP 404 but it’s back up now.
The myth of renewable energy laid (brutally) bare by Energy Skeptic: 67 Reasons why wind turbines cannot replace fossil fuels.
“The most important problem to be solved is electrifying transportation, otherwise how can you deliver the 30,000 parts of a Vestas 6MW wind turbine to the assembly factory and deliver it to the final site on gigantic trucks assembled with a gigantic crane? Or mine iron ore for the turbine, turn it into steel at 3000 F (70% made this way, not electric arc furnaces), deliver cement for the 2000 ton turbine platform?
If you do that, then each wind turbine must generate enough electricity to build another turbine (and decommission and recycle the old one) PLUS deliver huge amounts of electricity to society for home appliances, computers, the financial system, and all the other 14 major sectors of the economy that run on electricity.
There are no COMMERCIAL electric ways to make cement, iron, steel, other metals, glass, ceramics, bricks and so on. Steel recycled by an electric arc furnace is down-cycled and no longer suited for many purposes that require steel with certain alloys for strength and other qualities. Methods are still experimental, if they work, and are cheaper than fossils, and it would take decades and hundreds of billions of dollars to modify/build new manufacturing facilities. If they are more expensive, manufacturers will do nothing, and if forced to by regulations, move their facilities to where fossils are allowed.
The grid needs to grow 3-fold and thousands of square miles of solar and wind farms to generate the electricity (most net zero plans call for 70-90% of electricity from wind & solar)….“
(They go on to list 63 more reasons…)
🔬General galactic
General Galactic, a startup founded by former SpaceX engineers, came out of stealth. The firm is pivoting from developing Mars-focused technology to creating reactors that produce “fossil-free” methane fuel from carbon dioxide here on Earth. The company has built a pilot system producing 2,000 litres of methane daily, with plans to increase output and replace natural gas in industrial processes and chemical manufacturing rather than in heating and energy.
(But looking at their process… it requires Hydrogen gas as an input, which magically appears mid-way… oh… not so fossil-free then…!)
💫OpenStar
Here in Aoteoroa, nuclear fusion research company OpenStar Technologies achieved its first plasma. The team used a novel levitated dipole reactor (LDR) design which differs from traditional tokamak or stellarator fusion reactors, using a magnetosphere-like confinement system and levitation to keep plasma within a doughnut-shaped reactor. High-temperature superconductor (HTS) magnets, operating at 50K, create strong magnetic fields up to 20 Tesla. OpenStar aims to begin generating electricity from their reactor by 2030.
Check out this moment when the plasma first appeared:
…and the crowd reaction:
Video journalist Madison Reidy went behind the scenes with Openstar founder Ratu Mataira, super interesting. (Madison has been doing great work this year covering science and tech… check out her Youtube channel if you haven’t already…)
🚀Aerospace dawning
Also here in Aotearoa, in my home town of Ōtautahi no less, Dawn Aerospace achieved yet another milestone, getting their Aurora Mk-II prototype autonomous rocketplane up to Mach 1.1 and an altitude of 82,500ft - the first civil aircraft to fly supersonic since Concorde… and setting a world record for the fastest aircraft to climb from ground level to 20km altitude. After it landed, the rocket flew again 6 hours later. Congratulations to founder Stefan Powell and team… amazing!
🧠Mind expanding
Lots of interesting links to share this week.
Firstly, a couple of podcasts to recommend listening to…
👤Gwern in conversation
Gwern Branwen (@gwern to everyone) is a pseudonymous writer and obscurely followed polymath - arguably the first person to foresee and write about the value of Bitcoin and also the scaling hypothesis behind all of today’s generative AI.
His esoteric website of essays is a continually evolving goldmine of insights, trivia and customised CSS… you can get lost in here for years…
A reclusive figure until last week, he just took part in a long-form interview with the excellent Dwarkesh Patel (in virtual AI-generated avatar form - amazing work by Dwarkesh’s tech team) A unique individual with a refreshing intellectual curiosity…
It surfaced towards the end of the interview that Gwern has been living off US$12K / year Patreon income in the remote backwoods somewhere… cue Silicon Valley tech bros jumping in with offers for him to join the party in SF:
Of course, this may not end so well…
🎙️Lex X Dario
Also Lex Fridman does 5 hours with Anthropic’s Dario Amodei…. I haven’t had a chance to listen yet, but guaranteed full of insights:
⏩🌱Forward to Nature
The "Forward to Nature" (F2N) manifesto, co-authored by Charles Oppenheimer and David Galbraith, presents a novel approach to addressing climate change and achieving sustainable progress — advocating for reduced emissions rather than reduced energy consumption.
“Our house, the Earth, is on fire. To extinguish it we need to prioritize and act. Current political views on both Left and Right are either holding back action through climate change denial or ideological rejection of re-prioritization such as low carbon nuclear power. Longer term, these ideologies are more destructive still. Either assuming a view that we can continue to destroy ecosystems in the name of growth, tarmac the earth and deplete the seas without consequence or that we can reduce growth and progress to create a permanent global recession and not find ourselves sitting in a forest around a candle with lowered life expectancy and increased inequality. Meanwhile climate change continues and we may no longer produce or apply the right knowledge and tools to fight it.
This manifesto is designed to show, from physics principles, that energy consumption is not the problem, it is the solution, provided it is the right type of energy; why waste is part of life and can only be managed, not eliminated; that what we mean by growth is increased knowledge and that only a true understanding of how natural systems work will enable us as a species to protect our environment.
Active solutions require a radical shift in mindset and focus, neither as a passive inhabitant of a benign natural environment nor a bulldozer through it.“
It’s a comprehensive argument, aligning with many of my own 🌱/acc intuitions… Here are the 11 points of the whole manifesto, definitely worth going in deeper here:
Accept that climate change is real and an existential threat to humans.
Accept that climate change is provably caused by human energy use.
Reduce greenhouse gas emissions not energy consumption.
Reduce pollution except where greenhouse gas emissions are a competing higher priority.
Pursue economic growth as a means to achieve the goal of knowledge and wellbeing.
Understand energy, information, and capital flows as an integrated whole.
Capital flows: actively align economic incentives.
Capital flows: create non-prescriptive regulation.
Energy flows: abundant clean energy on demand, anywhere, with real-time pricing.
Information flows: abundant compute capacity, on demand.
Information flows: create recipes for solutions.
💻Digital institutions create wealth
Digital Institutions Create Wealth by Increasing Economic Computational Complexity - a new paper from pioneer crypto economists Justin Banon and Jason Potts, proposing a theory explaining the global transition from an industrial to a digital economy:
Digital institutions, including web3 technologies, enable the development of a 'computable economy' that increases economic complexity and wealth.
A critical transition will occur when institutional computation extends from on-chain assets to off-chain real-world assets
(It feels so abstract… but it could well arrive within just a few years…)
💭Noospheric
Finally… I found a kindred memetic spirit out there on X… John Bollenbacher’s analysis of what’s going on with social media information bubbles is spot on IMO (click to read the full thread…):
⏳Zeitgeist
Two other signals from non-tech world... in addition to the wounded planetary boundaries covered above…
🦠Mpox on the rise
The Africa CDC reports that the novel clade 1b mpox virus has spread to Zambia and Zimbabwe, indicating its presence in most African regions. The continent is averaging 2,800 new cases weekly, with 53,903 cases and 1,109 deaths reported since the beginning of the year. Key points include: 1. Uganda is experiencing a sharp rise in cases, particularly among sex workers, with 184 new cases in the past week.
🏗️End of The Line
The CEO of Saudi Arabia's ambitious Neom project, Nadhmi al-Nasr, stepped down amid shocking allegations of over 21,000 foreign worker deaths during construction (so far😮💨). Neom features "The Line" - a planned 105-mile-long desert skyscraper… as reported in Memia 2024.16 is now cut short by 103.5 miles and with only 1.5 miles expected to be completed by 2030.
🧘Memetic savasana
Quite a week. Luckily we have AI to create videos now to relax to...
😂Broligarchs
Someone did this. Gold. (Deepfakes on a plane?).
(Video via @lmMeme0)
📱Attention, splatted
Engaging use of Gaussian Splats in this video about…attention (watch to the end):
(Video via Jacques Alomo)
🌍The Turning Point
Finally (not AI generated), this beautiful, moving short film by animation artist Steve Cutts from 2020. Perhaps something like this might trigger *some* empathy for nature, *some*where…? (🌍 Jay Scott for sharing.)
(Another huge one this week… only four to go until the end of year - plus my annual wrapup post. Heads down…)
🙏🙏🙏 Thanks as always to everyone who takes the time to get in touch with links and feedback.
Namaste
Ben