H5N1 pandemic watch 🦠 o3🤯 deepseek v3🔓🇨🇳 genesis project🔮 veo influencers👱♀️ d/acc 1 year later🛡️ structural batteries🔋 rainmaker🌧️ Enron egg🥚 #2025.01
Build it and the energy will come (?)
Kia ora and Happy New Year to all Memia readers!
Welcome to the first full Memia missive of 2025 (S06E01) - my weekly newsletter now in its 6th year scanning across the latest developments in emerging tech, AI (AI, AI, AI…) and trying to sensemake an increasingly volatile, technologically transformed, accelerating future.
🙏🙏🙏As always, thanks for reading, subscribing, commenting and reaching out to get in touch - it’s always appreciated and I try to do my best to respond asap…!
ℹ️PSA: Memia sends *very long emails*, best viewed online or in the Substack app.
🗞️Intermission roundup
The most clicked link in last newsletter of 2024 (3% of openers) was Scott Wolfson’s article on AI and the consulting industry: The Consulting Pyramid Just Flipped: Here's Why That Changes Everything. (Note: fixed link since the original email).
ICYMI
My roundup of what happened in 2024, in nearly 24,000 words… thanks to those who commented giving positive feedback… so much happened last year, eh!!
Also yesterday I scored myself a *decent-enough* 75% on my 2024 Tech Predictions:
📋Camp notices
A few reminders to start the year…
📬Annual subscription special offer
Memia remains an ad-free, reader-supported publication running mostly on endorphins and high-quality caffeine☕. If you value receiving these weekly newsletters (and also keen to receive subscriber-only Mind Expanding, Strategy Notes posts and others, together with access to the cumulative Memia Knowledge Graph) then please consider signing up for a paid subscription, still only US$50 (or thereabouts in your local currency) for the year. In particular, if you (and your team) read this for work then please do consider a group subscription.
To grease the wheels, here’s a special offer if you sign up for an annual subscription before the end of January:
🗣️Getting the word out
If you’re really feeling the love, help me get the word out as well:
At all times, feel free to share Memia content using a Substack note / LinkedIn post / Thread / [Tw]eet / (Sk)eet / Cast / Facebook post and/or tell your network what you like about Memia and link back to https://memia.substack.com:
🤝Engage
Last year I worked as an AI advisor with leadership teams across many diverse industries including: energy, telecoms, infrastructure, logistics, financial services, media, health, law, agriculture, education, NGOs and government — as well as with many technology firms. If you require experienced support with your AI strategy workstreams: exec and board briefings on AI, frameworks to quickly embed situational awareness or simply providing an external perspective on what you’re thinking, please reach out directly: ben.reid@memia.com.
Also, my speaking roster for 2025 continues to shape up - if you’ve got an event coming up and are interested in booking me for a keynote/panel discussion/fireside chat on AI and what the future might look like then get in touch.
⏳Zeitgeist
OK, let’s start off with the world in 2025… five key non-tech themes I’ll be tracking this year:
🌡️Climate change, its impacts and its potential mitigations
Earth broke temperature records in 2024, surpassing the 1.5°C Paris threshold
2024 marked another historic milestone as the warmest year on record, exceeding the Paris Agreement's 1.5°C threshold above pre-industrial levels for the first time:
In 26 weather events studied by NGO World Weather Attribution in 2024, climate change contributed to the deaths of at least 3,700 people and the displacement of millions. Food security is increasingly threatened in the most vulnerable regions.
Rising temperatures are creating new health threats through disease-carrying insects and extreme weather events.
Extreme weather disasters have now cost the world US$2 trillion in the past decade.
The renewable energy transition is still accelerating, but with speedbumps
Renewable capacity build-out in 2024 is expected to slightly exceed consensus forecasts (IEA’s “Main Case” illustrated below)
There were wins: both Germany and Britain achieved their cleanest electricity production on record in 2024:
Germany achieved 59% renewable energy without nuclear, proving alternative energy viability.
The UK's complete phase-out of coal showed a practical path for other G7 nations.
…but, with EU gas reserves depleting at the fastest pace since 2021, with storage levels dropping 19% from September to mid-December 2024, the pace of the transition raises concerns about higher prices, refilling costs and energy security risks.
Investment in the green energy transition needs to double to US$3.5 trillion per year if the world is to meet the Paris Agreement's goal of reaching net zero emissions by 2050... yeah right.
HOWEVER GHG emissions *increased again* in 2024:
By November, the 2024 Global Carbon Budget project was forecasting fossil carbon dioxide (CO2) emissions of 37.4 billion tonnes for the year, up 0.8% from 2023. There were some significant regional variations:
China (32% of global emissions): 0.2% increase
US (13% of global emissions): 0.6% decrease
India (8% of global emissions): 4.6% increase
European Union (7% of global emissions): 3.8% decrease
Rest of the world (38% of global emissions): 1.1% increase
Meanwhile, CO₂ removal at scale remains a mirage Despite a lot of noise around CO₂ removal tech, nothing really worked except companies bought carbon removal “futures”, of which only 4% were delivered:
“capturing and storing CO₂ remains costly, with current prices reaching up to $1,000 per ton. Analysts suggest that for the industry to thrive, prices must fall to around $100 per ton.“ - via Andrew Lockley in his newsletter Carbon Removal Updates\
Microsoft way out front in the greenwashing stakes…
The more I look at the CO₂ removal industry, the more it seems a distraction unless a miraculous new technology emerges imminently: In 2024, technological carbon dioxide removal (CDR) extracted approximately 1.3 million tonnes of CO₂ from the atmosphere, constituting less than 0.1% of the total global CO₂ removal and a drop in the ocean compared to the vast amounts emitted globally, which exceeded 41 billion tonnes in 2024. (Conventional reforestation accounts for about 2.2 billion tonnes annually.)
🦠Pandemic watch
H5N1 bird flu is transmitting to mammals and this week killed the first human in the US:
The H5N1 virus has caused massive deaths among wild birds and mammal populations in the last two years.
According to the World Health Organization, there have been more than 950 confirmed bird flu infections globally since 2003, and more than 460 of those people died. That’s a high fatality rate.
A recent study by University of Pittsburgh researchers found evidence of H5N1 avian influenza virus adaptation in domestic cats. The research found that cats may act as "mixing vessels" for avian and mammalian influenza viruses.
A severe case in a Canadian teenager and increasing infections among US farmworkers indicates the growing occupational health risks. Just as I was compiling this newsletter, the news arrived that the first US bird flu death was announced in Louisiana.
I am certainly keeping an eye out for evidence of a sudden mutation which causes rapid human infection and fatalities, putting global pandemic controls back into force. The relatively high fatality rate would potentially make the 2019 Covid-19 restrictions look distinctly comfy in comparison…
Also watching: Human metapneumovirus (hMPV), a distant cousin of RSV, may be spreading in China during the current winter season.
🗽💩US domestic politics and its spillovers
Editorially, Memia is aiming to stay a US-domestic-politics-free zone during 2025. My hot take is that — as with the last Trump administration — there will be endless theatre, policy uncertainty and a cacophony of limbic-hijacking bluster… but little coherent signal other than here be chaos, leaving international trust and trade networks seeking greater reliability and resilience elsewhere and vainly hoping for sanity in 4 years’ time.
In particular, the brightly burning shooting star that is Elon Musk’s recent political ascendance (combined with his proprietorship of Twitter/X) is delivering mountains of coverage in US domestic media… but personally I suspect that star will burn out quickly one way or another (possibly very abruptly if he carries on this way… Icarus and all that…). We’ll see.
However, the wider impacts of the current rightward-swing of US domestic politics and the incoming Trump administration’s shotgun marriage with Silicon Valley’s techno-oligarchy is starting to spill over to other countries:
Musk’s recent unsubtle sidequests into UK and German domestic politics can be interpreted as a signal of how allowing US memetic culture to spread *unfiltered* outside US borders may no longer be considered desirable by elites and proles alike in many countries. China’s Great Firewall is an extreme case of measures to prevent this — in 2025 will Western liberal democracies actually consider enforcing more extensive bans on US social media platforms, particularly Musk’s X mouthpiece (like Brazil did)?
Meta’s appointment of Republican Joel Kaplan to lead global policy, replacing former UK liberal politician Nick Clegg, indicates that US domestic political norms could be tacitly rolled out through hegemonic US technology platforms. For example, just today Meta announced it was abandoning its fact-checking programme in favour of a crowdsourced “Community Notes” variant similar to X. Seizing the political moment to appease the incoming Trump administration while also ridding itself of the pesky costs associated with licence-to-operate... Zuckerberg actually said this in his statement:
“Fact checkers have just been too politically biased and have destroyed more trust than they’ve created.”
The European Fact Checking Standards Network (EFCSN - who knew there was such an entity until now?) pushed out a strongly-worded statement in response:
“The EFCSN strongly condemns Meta’s CEO’s statements linking fact-checking with censorship.
Platforms retracting from the fight against mis- and disinformation allows for election interference.
The EFCSN encourages the European Union to stand strong in the face of such political pressure and not be deterred in its efforts to stop the spread of mis- and disinformation on VLOPs.“
The growing conflict between Elon Musk and Wikipedia also reveals a broader right-wing campaign to control information flow and delegitimise independent platforms. Wikipedia's resilience (so far) against wealthy individuals' control makes it a target for attacks. I paid my regular donation within a few minutes of reading that article.
For those of us outside the US, there will need to be more deliberate analysis of the merits of US-style “freedom of speech” (so long as you agree with the oligarchs) and perhaps this will give rise to more robust debates on personal and national filters for “cognitive security”.
🌐Geopolitics
"When elephants fight, it is the grass that suffers."
— African proverb
While much of the world’s attention focuses on the US vs. China narrative, as both bull elephants circle each other following age-old animal power struggle dynamics, Memia will be mostly watching the grass… and perhaps the other challenger elephants watching from the sidelines. In particular my questions in 2025:
Will the Russia / Ukraine and Israel / Palestine conflicts see resolution in 2025? (And reparations…?)
How are other geopolitical players manoeuvring around the US vs. China headline act?
For example, China’s new stealth fighters and assault ship are driving an accelerating regional arms race — with India responding with hypersonic weapons development. Emergent regional powers including Türkiye, Brazil, Saudi Arabia and Indonesia all have stakes in the game and cards to play. And the UK and Europe aren’t entirely gone yet.
But the most important geopolitical question for me in 2025: The Technology for Autonomous Weapons Exists. What Now?
Who will have them?
Who will control them?
Who will control the defences against them?
Who will regulate / enforce regulation of their use?
(Stretch goal: Is the escalating global technological (AI/robotics) arms race actually just one giant coordination problem which could be successfully regulated with a coordinated (giant) open source technology solution?)
Fun and games ahead in 2025, indeed…
👥Demographics and migration
The world’s human population is getting older, quicker. With fertility rates now plummeting in many parts of the developed world, particularly Asia, how would the world respond to the demographic overhang?
A recent study by researchers from the National University of Singapore and Columbia University assessed 143 countries' preparedness for supporting aging populations across five key domains:
Chen et al. Switzerland, Norway, Denmark, Sweden, and Iceland topped the overall rankings, while the US ranked surprisingly (?) low at 24th among high-income nations.
Countries must adapt quickly: 50% of babies born in high-income nations are projected to survive to age 90 by 2060.
Making the picture more complex, according to the International Organization for Migration (IOM) World Migration Report 2024:
There are now 281 million international migrants globally.
The number of those displaced hit a record high by the end of 2022 at 117 million.
Migrant remittances surpass foreign direct investment in boosting the GDP of developing nations.
I’ll be watching closely for how the world adjusts to challenges associated with migration (cf. the incoming Trump administration’s intention to deport millions of migrants…) and also the opportunities: meta-national polities and technologies which cater to the needs of this increasingly significant migrant demographic?
📈Three weeks in AI
On to tech. Catching up on all the AI news and releases from the last 3 weeks while I’ve been on a break…
The relentless pace of releases continued right up until Christmas:
🤯OpenAI released its o3 reasoning model, only 2 months after the initial o1-preview. ("o3" chosen to avoid trademark conflict with the existing UK mobile carrier named O2). As close to “AGI” as anyone has measured yet:
OpenAI claims that o3 outperformed o1 significantly on complex benchmarking tasks:
In science, achieving 87.7% on the GPQA Diamond expert science benchmark
In software engineering, scoring 71.7% on SWE-bench Verified (compared to o1's 48.9%) and reached a much higher Elo score of 2727 on Codeforces (vs o1's 1891)
o3 demonstrated triple the accuracy of o1 on the ARC-AGI benchmark, which tests logical reasoning and skill acquisition abilities.
OpenAI also released its full o1 model (superceding o1-Preview) — and with a premium “o1 Pro Mode” that:
“thinks longer for the most reliable responses“
o1 Pro Mode is made available as part of a new “ChatGPT Pro” tier at US$200/month price tag.
Despite the price hike, according to OpenAI CEO Sam Altman they are still losing money on it… (as I wrote in the 2024 roundup, expect US$2000/month this time next year!)
Google shipped Gemini Flash 2.0 Experimental - a small “workhorse” AI model with the power of GPT-4o and o1-preview. Expect much from Google in 2025…
🔮The Genesis Project released the results of a 24-month large-scale research collaboration involving over 20 research labs to create an open-source “generative 4D physics engine”. This demo video below points in a direction of completely integrated multimodal generative physics models underpinning all real-world AI simulations from later this year (including video…):
🔓🇨🇳And then on Christmas Day, out of China emerged DeepSeek v3, benchmarking as the leading open-source AI model available and rivalling leading closed-source models like GPT-4o and Claude 3.5 Sonnet. But at a fraction of the cost of its competitors
Although the model has a massive 671 billion parameters, it utilises a Mixture-of-Experts (MoE) architecture, only activating 37 billion parameters for each token, which greatly enhances its efficiency.
Furthermore, the model was trained in just 55 days at a cost of only US$5.58 million — using far fewer resources compared to its commercial peers.
In particular calling into doubt the massive brute-force scaling hypothesis for AGI currently underpinning many of the major commercial labs’ work.
[Samo Burja and Nathan Lobenz discussed the DeepSeek moment in passing recently, introducing a mental model of a future “periodic table” of different “types of intelligence” — the generative AI architectures which have held the limelight over the last few years being just a subset. Useful framing.]
It is published under a very permissive open source licence, freely available to be modified for commercial use. They also provide a very informative and open 50-page technical paper.
Hanzhou-based DeepSeek was reportedly spun off from hedge-fund manager High Flyer Quant in 2023 by founder Liang Wenfeng, who had studied AI at Zhejiang University.
Chinese AI capabilities are clearly progressing rapidly, despite (because of?) US attempts to restrict access to high-end GPUs for training. (When you can’t get hold of enough chips, ingenuity must find a way…)
More in-depth analysis from Zvi below:
Interestingly, there may be a bit of imitation being the sincerest form of flattery going on here:
Sam Altman showing the tiniest signs of being rattled?
Wider picture:
At least one Chinese open-source lab is keeping up closely with leading US commercial and open-source labs. DeepSeek V3 was released only a couple of months after OpenAI o1.
Open-source is, to some degree, the great leveller (although it also forces capital to go and seek out other monopoly tollbooths upstream - data centers, chips…)
Every AI developer in every country outside the US should download and add DeepSeek to their library of open source AI models to maintain some degree of independence from a handful of US commercial labs.
💸All aboard the AI money train…
Generative AI VC funding skyrocketed in 2024:
According to data from financial tracker PitchBook compiled for TechCrunch, funding for generative AI worldwide hit record levels in 2024, with companies raising US$56 billion across 885 deals - a 192% increase from 2023's US$29.1 billion.
US companies dominated the AI funding landscape, capturing 89% of total investments
This number included infrastructure layer companies which attracted major funding due to rising data center demands — but not the rumoured multi-billion dollar “acqui-hires” of Inflection (Microsoft), Character.ai (Google) or Adept (Amazon).
(Some definitional differences likely accounting for the discrepancy with Accel’s October estimate of around US$80Bn for AI and Cloud companies outside China)
Where will it end? So far the revenues for nearly all of these startups are minimal… so I’m expecting rapid consolidation throughout 2025 — particularly as the new Trump administration is more likely to ignore antitrust concerns for M&A. Many of these companies will be forced to fold at lower valuations than they raised I expect.
Of the headline number, 1/56th of funding came from Nvidia pouring $1bn into AI startups in 2024 as chip demand soared, across 50 funding rounds and several corporate deals in 2024, up from $872 million in 2023:
A clear signal of Nvidia’s growing influence in shaping AI industry development, in particular making strategic investments to maintain market dominance against firms developing competing chips.
The scale of the investments certainly raises antitrust concerns — and also reveals Nvidia's strategy to diversify its customer base.
How many of these startups will be acquired by Nvidia itself as it levers its way up the AI value stack?
Microsoft kicked off 2025 announcing a massive US$80 billion investment for fiscal year 2025 to construct AI-focused data centers globally, with over half of the funding allocated to US builds.
The rapid expansion of AI data centers across the US is creating significant power grid challenges, with over 75% of distorted power readings nationwide occurring near these facilities. These investments must have large “build it and the energy will come” assumptions built in…
Well worth a listen: Microsoft CEO Satya Nadella in conversation with Bill Gurley and Brad Gerstner at the end of last year… an intimate, open conversation revealing many first-hand insights into how the tech giant thinks about allocating capital for its AI strategy:
📡AI signals
Miscellaneous AI developments I also took note of:
No more Talkie Chinese startup MiniMax's popular AI companion app Talkie (a competitor to Character.ai) has been removed from the US Apple App Store for “technical reasons”, while currently remaining available on Google Play…leaving aside the $70 million revenue impact for MiniMax, commentary has this as an anti-China tech move with allegations that CCP influence has been instrumental in Talkie’s rapid rise in the US.
Intention economy: Cambridge researchers warn of an emerging "Intention Economy" where AI systems could predict, commodify and sell your intentions … before you decide.
The science is likely sound that AI systems could predict personal intentions before conscious awareness … individual autonomy is potentially an illusion we tell ourselves.
Conversational AI assistants (cf. Talkie above…) can gather intimate psychological data and use this for unprecedented social manipulation.
How does a government regulate this? (And what would AI-driven “intention markets” mean for democratic processes...??
Apple agreed to a $95 million settlement in a class action lawsuit over Siri privacy violations going back to 2014 when Apple introduced the “Hey Siri” voice trigger. iPhone owners complained that Apple regularly recorded their private (sometimes rather intimate…) conversations after they activated Siri unintentionally — and then disclosed the content of these conversations to third party contractors and advertisers. Apple denies the claims about advertisers and says it has tidied up Ts&Cs… but Ouch.
Meta course-corrected a mis-step dating back to 2023 after the recent Financial Times headline Meta envisages social media filled with AI-generated users brought attention to its AI-generated “character” accounts on Facebook and Instagram. After a (predictable?) online backlash, Meta has since swiftly deleted all 23 of the AI accounts. (Was this actually a test run for the Dead Internet Theory?)
🗳️🤖AI and politics
It’s time to move beyond AI nationalism Wired paints a more hopeful picture of a significant shift away from nationalistic AI competition toward international collaboration in 2025, avoiding repeating Cold War mistakes but rather fostering beneficial innovation. (Is anyone in Washington or Beijing listening?)
Deposed by AI? Also in Wired, Yuval Noah Harari posits that AI's relationship with authoritarian regimes will be more complex than initially assumed:
“In the long term, authoritarian regimes are likely to face an even bigger danger: instead of criticizing them, AIs might gain control of them. Throughout history, the biggest threat to autocrats usually came from their own subordinates. No Roman emperor or Soviet premier was toppled by a democratic revolution, but they were always in danger of being overthrown or turned into puppets by their own subordinates. A dictator that grants AIs too much authority in 2025 might become their puppet down the road.
Dictatorships are far more vulnerable than democracies to such algorithmic takeover. It would be difficult for even a super-Machiavellian AI to amass power in a decentralized democratic system like the United States. Even if the AI learns to manipulate the US president, it might face opposition from Congress, the Supreme Court, state governors, the media, major corporations, and sundry NGOs. How would the algorithm, for example, deal with a Senate filibuster? Seizing power in a highly centralized system is much easier. To hack an authoritarian network, the AI needs to manipulate just a single paranoid individual.“
(Love the edginess in this...)
🆕AI releases
OK I’m back in the saddle now, but may have missed a few key announcements over the last couple of weeks while on holiday, but here’s what else I did catch:
Anthropic signalled a more independent path for its Model Context Protocol (MCP) AI agent integration framework, outlining the H1 2025 development roadmap focusing on five key areas of expansion and improvement:
Remote MCP Support
Reference Implementations
Distribution & Discovery
Agent Support
Broader Ecosystem
If the roadmap coalesces, then MCP should be one of the first frameworks to enable secure remote AI model connections and standardised authentication across multiple modalities, platforms and providers, supporting complex AI agent applications:
“Think of MCP like a USB-C port for AI applications.”
Google released Gemini 2.0 Flash Thinking Experimental, a new AI model that incorporates "thinking tokens" similar to OpenAI's o1 model, but with far lower pricing./
Among other things, this is another signal of an industry shift away from larger models toward improved reasoning through reinforcement learning.
Alibaba Cloud has announced its third major price reduction of 2024 for its Tongyi Qianwen visual understanding models, slashing prices by over 80%.
The cost reduction enables processing of 600 high-resolution images for just one yuan (13.7 US cents)!
ByteDance’s Doubao is yet another Chinese LLM demonstrating significant technological progress in 2024, claiming to match GPT-4o's performance while costing only one-eighth of the price.
Nvidia is having a big week of announcements at CES … more on that next week. But one new model already released:
Cosmos an open-source, open-weight Video World Model, trained on 20 million(!) hours of videos:
Google’s new Veo video model released at the end of last year is proving a hit in my feed. Two examples:
Veo showreel from P J Perreira:
”A year ago we were saying it would still take years to start to get convincing humans and emotion. We’ll, this new model is getting really close now. Consider all these images are exactly what we got, no post production. And they’re even not the best ones — we did some for our clients that were even cooler but can’t post without revealing too much.”
👱♀️Case in point: Veo makes very convincing “influencers”… hopefully rendering this awful talking-head format largely DOA:
Prime Intellect in collaboration with researchers from USC released METAGENE-1 :
“we're open-sourcing a state-of-the-art 7B parameter Metagenomic Foundation Model. Enabling planetary-scale pathogen detection and reducing the risk of pandemics in the age of exponential biology.“
Launching a new frontier AI model has become a new marketing campaign art form:
(On the topic of bioengineering AI models, see also futurist Amy Webb’s frantically flag-waving essay this week: Living Intelligence.)
📚AI research
🪦RIP peer review Researchers from the University of Rochester and Penn State just dropped a paper showing how to mass-produce academic finance papers using AI. They generated HUNDREDS of papers about stock predictions, each with different creative names and theoretical explanations. These weren't just random papers - they followed proper academic standards and actually discovered some legit stock prediction patterns from accounting data.
🤔 Why it matters:
On one hand, it shows how AI could supercharge financial research by automating analysis of large datasets.
On the other hand, it's basically industrialised HARKing (Hypothesizing After Results are Known) - creating theories AFTER finding patterns in data. The researchers imagine hundreds of seemingly legitimate papers flooding academic journals, each with AI-generated explanations for why their trading strategy "should" work. Academia's peer review system is about to face its toughest stress test yet.
AI editorial processes On which note… the entire editorial board of Elsevier's Journal of Human Evolution (JHE) resigned, (the 20th mass resignation from a science journal since 2023) after raising concerns about AI-driven editorial processes introducing errors and compromising scientific accuracy without author consent. (Not to mention rising author fees and reduced editorial independence).
🔮[Weak] signals
Non-AI tech signals from near and far futures...
⏱️Unregretted user-seconds
One nugget of gold, in amongst all of the political cacophony this last week, when X owners Musk introduced the goal to “maximize unregretted user-seconds” as the social network’s algorithmic North Star. I am intrigued how you would measure this… but it’s certainly a step up from “ad impressions” or “total time online”…:
🏛️💻Technocratic futures
A couple of pieces from two of the leading thinkers on how progressive technology could potentially shape future governance:
🛡️d/acc 1 year later Vitalik Buterin revisits progress towards his philosophy of “d/acc”:
“…decentralized and democratic, differential defensive acceleration. Accelerate technology, but differentially focus on technologies improve our ability to defend, rather than our ability to cause harm, and in technologies that distribute power rather than concentrating it in the hands of a singular elite that decides what is true, false, good or evil on behalf of everyone. Defense like in democratic Switzerland and historically quasi-anarchist Zomia, not like the lords and castles of medieval feudalism.“
Progress noted includes diverse initiatives such as:
Verifiable open-source vaccines
Indoor air quality monitoring
Epistemic technologies including prediction markets Community Notes
Privacy-preserving ZK-SNARKs in government ID and social media.
Buterin proposes two main strategies for AI safety regulation: liability rules targeting users, deployers, and developers; and a global "soft pause" capability for industrial-scale AI hardware. A couple of accompanying graphics below:
Vitalik - d/acc images (Images: @VitalikButerin )
🗳️Technodemocracy Balaji Srinivasan riffs on the concept of "technodemocracy": a blockchain-based evolution of political systems, building on cryptocurrency foundations:
Introducing new voting systems based on blockchains to make political participation verifiable and binding
Enabling creation of global political communities unrestricted by geographic boundaries. (The Network State is a meme which isn’t going away…)
Linking digital voting rights with real-world consequences through smart contracts
A fount of provocative ideas as always… but practical pathways to implementation?
🌧️Rainmaker
US startup Rainmaker is developing cloud seeding technologies to bring rain to arid areas:
“Traditional cloud seeding efforts struggled with operational flexibility and measuring impact due to past technological limitations. Recent advancements in drones, numerical weather modeling, and radar algorithms have transformed cloud seeding into a viable, scalable pursuit with demonstrable results. The time is now.“
What could possibly go wrong…?
🌋Fracking geothermal
Oil frackers pivot to clean geothermal energy development Enhanced geothermal energy systems (EGS) are attracting significant investment and expertise in the US from major oil and gas fracking companies, accelerating geothermal energy development through existing fracking skills.
Geothermal could provide massive clean energy potential in the US, with Utah alone having 49,400 megawatts capacity.
🌌📍Mapping space resources
Space resource mapping the Planetary Resource Management System (PRMS), adapted from oil industry practices, offers a structured approach to space resource utilisation through a two-step framework: resource discovery and technology development, and could determine which space resources are most crucial for mission success.
ISRU system concept for autonomous construction on Mars. Credit: NASA/JPL-Caltech
🕶️VR
Alibaba Cloud and RayNeo partnered to develop advanced AI glasses Alibaba Cloud and RayNeo signed China's first major collaboration to make AI smart glasses to rival Meta/Rayban, targeting competitive pricing at 1699 yuan (US$232).
Rayneo 🎻JAMS The University of Birmingham has developed JAMS (Joint Active Music Sessions), a virtual platform that enables remote musical collaboration through VR. The system delivers near-zero-latency virtual music collaboration and enabling musicians to practice and play with expert avatars. (Nice concept, but looking at the image they’ll need to work with something other than Microsoft Hololens 2 to go commercial!)
Opens up a whole new set of musical education options, potentially democratising access to top instruction (imagine if the avatar is animated using AI…)
Also the ability to play collaboratively - imagine the “Metaverse Symphonic Orchestra” delivering a live concert!
A musician records themselves and sends the video to another musician. The software creates a responsive avatar that plays in perfect synchrony with the music partner. All you need is an iPhone and a VR headset to bring musicians together for performance, practice, or teaching. Credit: ARME project
🔋Structural batteries
Scientists at Chalmers University and KTH Royal Institute of Technology have made significant progress in developing "massless" EV batteries that could increase EV range by 70% by integrating battery components into vehicle structure.
World's fastest train
Photos emerged of China’s new CR450 high-speed train prototype, unveiled at the end of 2024, designed to be capable of a top speed of 450 km/h (280 mph)
(China's massive rail expansion plans confirm its dominance in high-speed rail technology, with plans to extend its network to 60,000 kilometers of high-speed tracks by 2030.)
🛰️🚀Fake satellites on a real rocket
SpaceX's upcoming seventh Starship test flight will marks a significant milestone with the mission featuring the first-ever payload deployment, consisting of 10 Starlink satellite simulators that match the size and weight of future V3 satellites. (The next-generation V3 Starlink satellites will offer dramatically improved capabilities - over 10 times the downlink and 24 times the uplink capacity compared to current V2 Mini satellites). Once operational, Starship will deploy 60 V3 satellites per launch, adding 60 terabits per second of network capacity.
☀️Flying close to the Sun
NASA's Parker Solar Probe successfully completed its closest-ever solar approach, passing through the Sun’s atmosphere and reaching within 6.1 million km of the Sun's surface on December 24th, surviving temperatures exceeding 500,000 degrees Celsius in the solar corona, protected by a specialised carbon foam shield that maintains room temperature conditions for instruments.
🧬Smart cells
Bioengineers at Rice University have created a “construction kit” for building customisable sense-and-respond circuits in human cells. The technique leverages phosphorylation - a natural cellular process - to create "smart cells" that can detect and respond to disease markers, enabling rapid disease detection and automated treatment delivery through programmable cellular responses.
The technique could revolutionise treatment of complex diseases like cancer and autoimmune conditions.
Here’s a video of the process in action over the space of 70 minutes:
(Xiaoyu Yang et al., Engineering synthetic phosphorylation signaling networks in human cells.Science387,74-81(2025).DOI:10.1126/science.adm8485 )
💎Light-switching nanocrystals
Researchers from Oregon State University and partner institutions have developed revolutionary nanocrystals that can rapidly switch between light and dark states, which could enable faster, more energy-efficient AI processing and datacenters down the road….
🧠Mind Expanding
…is coming in a separate post at the weekend.
🎭And then there were memes…
(New section name for 2025, same content…)
💰🏛️Pay to play
The US political system is demonstrably a marketplace captured by financial interests and open to the highest bidder… as this animation below makes clear.
How to design next-generation techno-democracies to prevent this?
💩If your data is shit…
Captures the AI hype zeitgeist pretty well:
🥚Enron Egg
(🎩John H for sharing… gold!)
This week people were momentarily surprise that apparently… Enron Corporation… still exists(?!) And on enron.com this week the company apparently launched a “micro-nuclear reactor made to power your home” - the “Enron Egg”:
It’s all over the home page of their website…including a slick — but slightly off — tech industry launch pitch:
…together with bucketloads of saccharine corporate messaging, some of which is surely over-the-top irony:
And then this:
“…All of this made possible by the Enron mining division, which has been sourcing the proprietary enronium ore, which I think you’ll agree has unlocked a new atomic age,“
Aha. Turns out this is a top-quality parody campaign by comedian Connor Gaydos — after purchasing the rights to the defunct Enron brand in 2020 for just US$275!
More like this in 2025, please!
Took me a while to get back into it, thanks for reading. See you next week…
Namaste
Ben
Re Meta/X: I firmly believe that the world is ripe for what I call "public service social media" - platforms that provide the undeniable benefits of FB and X in terms of information distribution and social connection, but which are run on a BBC or NPR type of model. Relatively easy to start in one country (NZ?) or region (EU?), but should be extensible so that various systems can interlink and provide the international connections important to FB-type networks. All posts to be "legal, decent, honest and truthful", which is what ads in the UK are supposed to be.
All we need to find is a billionaire who wants to back us, and doesn't want to make more money... ;-)