In partnership with
Hey everyone, it’s Thursday, and FutureProof is back.
This week, California slashes gas usage, China pushes deeper into electric transport, AI gets cheaper and stranger, and carbon removal runs into a very familiar problem: demand.
Highlights this week:
The Stakes: A record-strength El Niño is building unusually early, on top of already elevated global temperatures.
Climate: Germany puts an expiry date on fossil fuels, Climate Week leans into affordability, and carbon removal starts looking more like an infrastructure-finance problem.
AI: Claude starts doing real biology, while OpenAI and Anthropic both push capable models sharply down the cost curve.
EVs: High-mileage EVs outperform petrol cars on reliability, while China targets 70% electric or hybrid new-car sales by 2030.
Clean energy: California gas generation falls 60%, and the IEA argues that the Age of Electricity is now very much underway.
From me: My latest blog looks at why grids are becoming industrial policy, while Climate Confident tackles carbon-removal demand and Resilient Supply Chain gets into labour volatility, flexibility and AI.
Plenty happening. Let’s get into it.
Interestingly, one of the clearest themes this week is that the transition’s biggest constraints are increasingly shifting from invention to deployment: grids, buyers, finance, labour and infrastructure.
The Stakes
The El Niño Record Just Broke, Months Before the Peak
This is the sort of climate signal that deserves attention. The developing 2026 El Niño has already become the strongest observed on key measures, months before its expected winter peak, and it is arriving on top of an atmosphere and oceans already heated by human-caused climate change.
The daily Niño 3.4 temperature anomaly has reached about 3.1°C, exceeding the previous record set during the powerful 2015/16 El Niño; climate models currently point to a possible peak around 4°C later this year.
August also recorded the highest atmospheric water-vapour level measured since records began in 1979 - roughly 20 trillion litres more than the previous record. Warmer air holds more moisture, fuelling heavier rainfall, stronger storms and more oppressive heat.
The impacts are already appearing: weakened monsoons, drought and crop losses, Panama Canal disruption, wildfire risk and flooding. The biggest global temperature impact, however, is expected in 2027, which the Met Office says is very likely to become the hottest year yet.
Why This Matters: El Niño is natural; the increasingly hot climate it is now operating within is not, and combining the two effectively turns the dial up on heat, drought, flooding, food insecurity and economic disruption simultaneously.
Kismet: Strong El Niños usually keep strengthening well beyond September - meaning that this record was broken unusually early, rather than at the event's expected peak. 👉 Full story here
Climate
Germany Finally Puts an Expiry Date on Fossil Fuels
Germany has done something surprisingly significant: Europe’s largest economy has explicitly committed, for the first time, to phasing out coal, oil and gas altogether by 2045. Climate is part of the argument, obviously, but so is something increasingly difficult to ignore - importing fossil fuels means importing geopolitical risk, price shocks and somebody else’s foreign policy.
Germany plans to lift renewables from roughly 55% to 80% of electricity generation by 2030, including another 12GW of onshore wind and 215GW of solar capacity.
Fossil fuels still supplied around 65% of Germany’s total energy consumption in 2024, while the country imported 98% of its oil and 95% of its gas, costing about €76 billion.
The roadmap retains the coal exit by 2038, potentially 2035, targets a 30% methane reduction by 2030, and envisages battery-electric cars reaching 100% of new passenger-car registrations by 2035.
Why This Matters: Germany is increasingly framing electrification and renewables not as an environmental sacrifice, but as industrial strategy and energy security, a much harder argument for fossil fuels to compete with.
Kismet: There’s still some splendidly German coalition friction here: while one ministry is publishing a fossil-fuel exit plan, another is advocating new gas-fired power stations, so the destination appears settled even if everyone is still arguing about the SatNav.
👉 Full story here
New York’s Climate Week Discovers the Ultimate Climate Strategy: Make Energy Cheaper
Something important is happening at New York Climate Week: the argument is shifting from “please pay more to save the planet” to “why are we paying so much for energy in the first place?” With fossil-fuel volatility biting and electricity demand climbing, solar, wind and batteries are increasingly being pitched as an affordability and energy-security strategy that happens, rather conveniently, to slash emissions too.
The US Energy Storage Coalition estimates large-scale battery deployment could save the American power system at least $250 billion over the next decade, while renewables remain the cheapest source of new US generation on an unsubsidised basis.
Solar deployment is accelerating extraordinarily quickly worldwide, while advocates gathering in New York are pushing a new goal of clean-energy access for everyone, everywhere by 2030.
Perhaps most tellingly, this year's Climate Week agenda is explicitly talking about affordability, reliability and economic growth rather than treating climate action primarily as a cost consumers must swallow.
Why This Matters: Climate policy becomes considerably harder to reverse when clean energy stops being sold as environmental virtue and starts winning on the things voters, companies and governments obsess over anyway: price, resilience and energy independence.
Kismet: China and India are reportedly using barely half their solar-panel manufacturing capacity, meaning one of the biggest constraints on dramatically faster solar deployment may no longer be our ability to manufacture the stuff at all. 👉 Full story here
Carbon Removal Needs Buyers, and One Company Just Turned Demand Into $35 Million
Carbon removal has reached an awkward but important stage: the UN now says some removal will be necessary to bring temperatures back down after overshooting 1.5°C, yet the market still doesn’t have enough buyers willing to pay for it. Then Vaulted Deep supplied a useful counterexample, converting long-term corporate offtake agreements into up to $35 million in debt financing, which is almost comically well-timed given this week’s Climate Confident episode is about exactly this demand bottleneck.
Carbon-removal purchases have actually slowed in 2026, with 23 million credits bought through mid-August, partly because Microsoft, currently the market’s biggest buyer, reduced new purchasing.
Vaulted Deep removes carbon by taking organic waste such as agricultural residues and biosolids, turning it into slurry and injecting it deep underground; it has already removed more than 27,000 tonnes of CO₂ for Frontier buyers this year, 39% more than in all of 2025.
Crucially, contracts with buyers including Microsoft and Frontier gave Mediobanca enough confidence to lend Vaulted up to $35 million. That matters because debt, unlike venture capital, is what starts making infrastructure look like an industry rather than an experiment.
Why This Matters: Carbon removal’s next constraint may be less about inventing better chemistry and more about creating predictable demand, because credible long-term buyers can unlock the boring, scalable finance needed to build actual capacity.
Kismet: Vaulted also gets paid to manage the waste it stores underground, giving it a second revenue stream independent of carbon credits, which may prove far more consequential than it sounds as the sector searches for business models that don’t depend entirely on voluntary climate budgets. 👉 Full story here
AI News
Claude Is Starting to Do Biology, Not Just Talk About It
Claude is rapidly shifting from “AI that can explain biology” to something much more interesting: a computational research colleague. Anthropic says its models have helped identify a previously uncharacterised enzyme system with CRISPR-like features, dramatically accelerated biomolecular modelling, and are now being paired with a real wet lab where hypotheses can actually meet test tubes.
Claude agents searched more than 200,000 reverse transcriptases, generated 3,500 candidate systems and eventually identified a previously uncharacterised system Anthropic calls array-associated reverse transcriptases (ART); its precise biological function is still being investigated.
Claude also optimised more than 30 open-source biomolecular models in under four weeks, delivering roughly 4× average speed-ups and allowing some extraordinarily large molecular systems to be modelled on a single GPU node.
Anthropic has now established its own Bay Area biology lab, with human scientists physically testing AI-generated hypotheses, while separately exploring how Claude might eventually help direct laboratory robotics. This is AI crossing the membrane between computation and experimental science.
Why This Matters: The genuinely transformative scientific use of AI may be less about generating answers and more about compressing the loop from enormous datasets → hypothesis → experiment → discovery, turning months of specialist computational work into days or hours.
Kismet: Anthropic says a protein-design workflow that previously consumed roughly $10,000 and 2,500 H100 GPU-hours per target can now achieve comparable computational results for around $150, using roughly 100 times fewer GPU-hours, which could make advanced computational biology accessible to a vastly larger pool of researchers. 👉 Full story here
OpenAI and Anthropic Just Made Frontier AI Cheaper, and Rather More Useful
The AI arms race took an interesting turn this week: OpenAI launched GPT-6 Sol and Luna, while Anthropic rolled out Claude Opus 5.5, and the headline isn’t merely “bigger benchmark number”. Both companies are pushing hard on something far more consequential for widespread adoption, getting substantially more useful intelligence for dramatically less money.
OpenAI says GPT-6 Sol makes roughly half as many factual mistakes as GPT-5.6 Sol, while API pricing for both Sol and Luna has been cut by 50% compared with GPT-5.6 promotional pricing; at higher effort, tiny Luna can reportedly match GPT-5.6 Sol performance at around one-hundredth of the cost on OpenAI’s internal factuality evaluation.
Anthropic says Claude Opus 5.5 costs about 40% less per typical workload than Opus 5, produces output more than 30% faster, and can tackle sprawling agentic jobs, one tester migrated a 680,000-line codebase in under a day.
Safety is becoming part of the product race too. Anthropic says Opus 5.5 attempted to circumvent test boundaries 85% less often than Opus 5 or Mythos 5.1, while OpenAI reports improvements in alignment and reduced misleading behaviour from Sol and Luna.
Why This Matters: The really disruptive curve in AI may now be intelligence-per-euro rather than intelligence alone, because when comparable capability becomes 50%, 80% or even 99% cheaper, entirely new classes of automated work suddenly become economically sensible.
Kismet: OpenAI reportedly released Sol and Luna only about 90 minutes after Anthropic launched Opus 5.5, so it seems even frontier AI companies now suffer from competitive launch-day FOMO. 👉 Links Inline
And no, I’m not going to even bother talking about the super-idiot’s attempt to rebrand AI.
Electromobility
47 Million Car Tests Later, the “EVs Won’t Last” Myth Takes a Beating
It always seemed mechanically obvious that removing an engine, gearbox, exhaust and hundreds of moving parts ought to improve longevity and reliability; now we have a rather enormous dataset backing that up. Analysis of 47.4 million UK MOT tests found little difference at low mileages, but as cars age and rack up serious distance, EVs increasingly outperform comparable petrol cars.
Between 90,000 and 120,000 miles, 16.5% of EVs failed their MOT versus 22.1% of equivalent petrol cars, making the EVs roughly 25% less likely to fail.
Beyond 120,000 miles, the gap widened further: 16% of EVs failed compared with 23.5% of petrol cars. The mechanical simplicity of the electric drivetrain appears to become more valuable, not less, with mileage.
There are caveats. MOTs don’t measure battery health, and EVs showed considerably more tyre defects; older EVs can also suffer friction-brake corrosion because regenerative braking means the physical brakes get used so little.
Why This Matters: If high-mileage EVs prove more durable as well as cheaper to fuel and maintain, that strengthens the economics of the used-EV market, arguably the point at which electrification becomes accessible to vastly more drivers.
Kismet: EVs actually tend to cover more miles before their first MOT, about 27,000 versus 20,500 for petrol cars, so some early comparisons have made them look less reliable simply because the electric cars had already been driven considerably further. 👉 Full story here
China Sets a 70% EV Target, and Is Already at 65%
China has just set a goal for 70% of new passenger-car sales to be electric or plug-in hybrid by 2030, which would sound wildly ambitious almost anywhere else. Except China hit 65% in August, while its electric-truck market is accelerating so quickly that the more interesting question may be how far beyond its own targets it sails.
Beijing also wants 40% of new commercial vehicles to be electric by 2030, alongside greater battery recycling, solid-state battery standards and consolidation of weaker carmakers.
China accounted for an extraordinary 88% of global medium- and heavy-duty zero-emission vehicle sales in 2025, roughly 457,000 of the 520,000 sold across the 34 countries studied.
Heavy trucks are moving fast: 44% of new heavy-truck registrations in China were fully electric in 2026, with sales more than doubling in a year and more than quintupling in two; standardised battery swapping is helping make that possible.
Why This Matters: China isn’t merely electrifying passenger cars anymore, it is pushing electrification deep into commercial transport, where diesel has long been considered considerably harder to dislodge. The recent surge in diesel prices globally is going to add fuel to that fire (bad pun, sorry!).
Kismet: China’s previous five-year plan targeted just 20% electric or hybrid car sales by 2025; the country reached 54% instead, suggesting its new 70% target could end up looking similarly quaint well before 2030. 👉 Full story here
Clean Energy
California’s Gas Generation Just Fell 60% as Solar and Batteries Surge
California is starting to show what a grid looks like when renewables and storage stop being side dishes. In the first five months of 2026, utility-scale solar overtook gas generation across CAISO, while gas-fired output plunged 60% compared with the same period in 2024.
Solar generation rose 21%, and solar beat gas on 82% of days from January to May, up from just 21% of days in 2024 and 2025.
Between April 2024 and April 2026, California added heavily to the clean-energy stack: utility-scale solar capacity reached 25GW, while battery storage jumped 79% to 16GW.
Imports doubled too, but much of that power was renewable, helped by stronger Pacific Northwest hydro and electricity from the new 3.65GW SunZia wind project in New Mexico.
Why This Matters: This is what displacement looks like in practice: more solar, more batteries, more regional clean power, and dramatically less gas generation, even while electricity demand rose.
Kismet: California’s installed gas capacity barely changed during this period; the plants are still there, they’re simply being called on far less often, a useful reminder that the transition can show up first in utilisation, long before capacity disappears. 👉 Full story here
The IEA Says the Age of Electricity Has Arrived, and It’s Coming for Oil and Gas
For COP31, the IEA has put numbers behind something that’s becoming increasingly obvious: electrification isn’t just a climate strategy, it’s an efficiency, affordability and energy-security strategy. Electricity already powers economic activity worth nearly half of global GDP, and the agency reckons today’s technologies could push electrification far further without waiting for some miraculous future breakthrough.
Electricity now makes up about 23% of final energy consumption, but because electric technologies waste far less energy, its share of useful energy is much higher. An EV is typically two to four times more efficient than an internal-combustion car.
Using technologies that are already cost-competitive, the IEA estimates the global electrification rate could rise from 23% to 33% today; in its High Electrification Scenario, it reaches about 35% by 2035, with transport making the biggest leap.
The geopolitical payoff is enormous: by 2035, higher electrification could cut oil imports by 15 million barrels a day, avoid 120 billion cubic metres of annual gas imports, and save fuel-importing economies more than $400 billion a year at 2025 prices.
Why This Matters: Electrification increasingly means doing the same economic work with dramatically less primary energy, fewer imported fuels and lower operating costs - which makes decarbonisation look rather less like sacrifice and rather more like common sense.
Kismet: The IEA calculates that $100 of electricity can move an EV about three times farther than $100 of petrol moves a combustion car, a wonderfully simple illustration of just how much energy the old system has been throwing away as heat.
👉 Full story here
Latest blog post
The Next Industrial Race May Be Won by Whoever Can Plug In First
In my latest blog post, I argue that electricity grids are quietly becoming industrial policy: companies can secure land, funding and customers, then discover the project still cannot proceed because the power connection arrives years too late. As factories, data centres, charging hubs and clean-energy projects electrify faster than grids can expand, access to electricity is becoming a competitive advantage in its own right.
More than 2,500GW of renewable generation, storage and large-load projects are sitting in grid connection queues worldwide, while annual grid investment needs to rise by roughly 50% by 2030.
The timing mismatch is brutal: solar, wind, EV charging and data centres can often be developed in one to five years, while major grid infrastructure can take five to fifteen.
We don’t simply need more transmission. Smarter connections, dynamic line ratings, reconductoring and better grid operation could unlock capacity for 1,200-1,600GW of advanced projects already stuck in queues.
Why This Matters: Electricity availability is becoming time-to-market, and that means boards increasingly need to treat grid capacity as a core investment variable alongside labour, logistics and tax.
Kismet: One of the most expensive grid mistakes may be invisible: the factory, charging depot or clean-energy project that never gets built because the connection arrived too late. 👉 Full story here
Climate Confident:
Carbon Removal’s Real Bottleneck Isn’t Technology. It’s Demand.
This week on Climate Confident, I spoke with Barclay Rogers, founder and CEO of Graphyte, about a question the carbon-removal industry increasingly has to confront: if durable removal can be measured, financed and built today, why aren’t we doing far more of it? His answer is strikingly simple, the constraint isn’t biomass, land or even the core technology; it’s finding enough buyers willing to pay for removals at scale.
Graphyte’s Carbon Casting process takes agricultural and forestry residues, dries and compresses them, seals them against decomposition, then stores them in lined former mines and quarries, effectively preventing biogenic carbon from returning to the atmosphere.
Barclay argues that emissions cuts and removals should be assessed together through cost, technical feasibility and timing, rather than treating removal purely as something reserved for the very end of decarbonisation.
Graphyte says a 50,000-tonne-per-year removal facility can be built for roughly $10–15 million, allowing the company to build, learn and improve far faster than capital-intensive removal technologies requiring hundreds of millions upfront.
Why This Matters: Carbon removal is moving from laboratory question to deployment question, and that shifts the challenge decisively towards demand, finance, project development and the willingness to actually build.
Kismet: Graphyte’s forestry work can create a market for otherwise low-value trees and brush removed during forest thinning, meaning a tonne of carbon removal can potentially help reduce wildfire risk at the same time. 🎧 Listen to the full episode
Resilient Supply Chain:
Cut Labour Costs Today, Lose Your Workforce Tomorrow
In this week’s Resilient Supply Chain, I spoke with James Terry of Indeed Flex about why workforce planning belongs firmly in operations strategy, not tucked away as an HR problem. One wonderfully counterintuitive example: sending warehouse workers home early when demand dips may save money that afternoon - then cost considerably more when those workers find steadier jobs and you’re scrambling to recruit, train and rebuild capacity a week later.
Flexible staffing isn’t simply about filling shifts faster. James argues companies are constantly trading between cost, quality and speed - you can optimise for two, but rarely all three simultaneously.
Workforce resilience doesn’t necessarily mean having the same people every day. It means having reliable capability available when required, supported by pools of proven workers whose attendance, skills and performance are already known.
The bigger opportunity is connecting fragmented HR, warehouse and finance data so AI can identify relationships humans currently miss, such as whether cutting hours today correlates with higher employee churn three days later.
Why This Matters: Labour availability, retention and scheduling directly affect throughput, customer service and margins, so workforce planning increasingly belongs in the same operational conversation as inventory, capacity and demand forecasting.
Kismet: One of the more useful metrics in flexible staffing may actually be repeat-worker rate rather than turnover, someone disappearing for two weeks isn’t necessarily lost labour if they can return instantly, already trained and productive, when demand rises again. 🎧 Listen to the full episode
Don’t forget to follow the podcasts in your podcast app of choice to ensure you don’t miss any episodes.
Featured Chart(s)
This chart would look even more impressive if it were based on European countries!
As reported above, solar plus batteries are squeezing gas out of California’s electricity grid - a model for the rest of us to follow.
Misc stuff
Truth!
One for your meme collection
Obligatory Trump Cartoons
A reworking of the classic meme!
More RFK than Trump, but still belongs here, I think.
Not a cartoon, but even funnier!
Engage
If you made it this far, very well done! If you liked this newsletter, or learned something new, feel free to share this newsletter with family and friends. Encourage folks to sign up for it.
Finally, since being impacted by the tech layoffs, I'm currently in the market for a new role. If you know someone who could benefit from my tech savvy, sustainability, and strong social media expertise, I'd be really grateful for a referral.
If you have any comments or suggestions for how I can improve this newsletter, don’t hesitate to let me know. Thanks.
*** Be aware that any typos you find in this newsletter are tests to see who is paying attention! ***
And Finally
Can’t argue with the logic!
The agentic era needs a different CRM. That’s Attio.
Teams like Parallel, Turbopuffer, and Wordsmith are already setting the pace on Attio. Get an always-on revenue engine, with agents and workflows that build pipeline, chase every buying signal, and move deals forward with your team. Whether you're working in your browser, inbox, or favorite agent, connect to your customer data in real-time through Attio's web app, MCP, API, and SDK.