Signals, experiments, and things worth looking at — 12 November 2026
Some developments are immediately relevant to public services. Some are proper strange. I’ve been collecting both: changes in who can act, what can be connected, and what we might need to notice.
A funded experiment isn’t an emerging social practice; an interesting scientific finding isn’t necessarily a sign of the future. Some of what follows belongs on a watchlist. Some deserves attention simply because it makes the world a little less familiar.
Who gets to act—and who can get through?
The UK is considering how payments regulation should accommodate AI agents acting for people and businesses. A Treasury answer on 4 September described consultations and sandbox work on agent-led payments. The ITU and World Bank’s September account addresses the corresponding identity problem: how does a public authority know who authorised an agent, what it may do, and when that permission expires?
That is delegated authority, not legal personhood for software. The practical questions are substantial enough: can I revoke the mandate, inspect what happened, and challenge a transaction? Public services will need to distinguish someone acting through an authorised agent from a machine making an unauthorised demand.
Treasury answer on agentic payments, 4 September
ITU and World Bank: identity and authority for agents
Meanwhile, Citizens Advice is arguing for the right to talk to a human and to move between online and offline support. Its September report describes people trapped by channel restrictions and poor digital design. These are proposed principles, not a newly enacted general right.
The two developments belong together. Institutions are preparing to recognise software acting for a person while some people struggle to get themselves recognised without using software. A service’s access arrangements determine who can exercise a right, even when the right itself hasn’t changed.
Citizens Advice: Behind digital walls
Electricity demand becomes something we can organise
The Electricity Demand Turn Up trial asks whether flexibility providers can increase consumption when renewable electricity would otherwise be curtailed. Its initial expression-of-interest window closed on 17 September. That is an experiment to follow, not an open invitation to apply.
There is now a further development. On 10 November, Innovate UK opened a competition offering up to £15 million for interoperable bidirectional vehicle charging. It explicitly brings together transport, energy, digital, and user interests. Applications close at 11:00 UK time on 27 January 2027.
The change worth watching is the coordination required. An electric vehicle can be transport, stored energy, and a potential contributor to grid flexibility, but those uses depend on compatible vehicles, chargers, network arrangements, tariffs, and customer consent. The battery’s existence doesn’t establish that its owner can export to the grid. Who can promise which capacity, at what time, without leaving someone unable to make their next journey?
UKRI: Electricity Demand Turn Up trial
Innovate UK: Scaling Interoperable V2X
Buying something before it is a finished product
The Sovereign AI R&D Procurement Scheme offers another institutional experiment. Its October explanation says there are no minimum turnover, net asset, or cash reserve requirements. Suppliers retain their intellectual property, and upfront payments may be available where cash-flow constraints would prevent participation.
Government is offering to be an early customer for demonstrators. That can help create a market, but the procurement design is only the start. I would want to follow who wins, whether smaller suppliers can actually participate, what gets learned in use, and whether a successful demonstration has a route into ordinary services. Announcing a different buying process doesn’t yet show that those things have happened.
Sovereign AI: Making procurement work for Britain’s AI startups
Some of the stranger experiments
ARIA’s Accelerated Adaptation programme is exploring whether adaptation in wild species can be accelerated responsibly. Funded work includes tree protection and immunisation, amphibian resistance, and pollinator resilience, alongside ethics and governance. The programme describes these as possible additions to conservation, while reducing the pressures driving decline remains essential. The question is how an intervention’s effects would be judged across an ecosystem, and who would have standing to challenge it.
ARIA: Accelerated Adaptation funded projects
At Maynooth, researchers have demonstrated a computer made from interacting DNA strands. The intended answer is encoded in a thermodynamically favoured arrangement: the molecules assemble towards it. Their Nature paper reports ten programs, including arithmetic and parity detection. This is laboratory molecular computing; the prospect of calculations embedded in chemical environments remains something to explore, not a replacement laptop to order.
Stérin and colleagues: A thermodynamically favoured molecular computer
Cornell researchers have combined temperature sensing, electronic communication, and artificial cilia to pump fluid. The resulting action changes the surroundings being sensed. It’s a concrete feedback loop. The demonstrated array uses leader–follower circuitry; it isn’t evidence of a free-roaming swarm with no coordinating structure. That detail matters when describing what has actually been achieved.
Cornell: microscopic robots that sense and reshape their surroundings
Space Solar’s Project HAWK is developing radio-frequency power beaming from the ground to high-altitude aircraft, with an integrated flight demonstration planned. Separating an aircraft from its energy source could change the possibilities for persistent communications platforms. For now, it is a research programme with a target, not an operating replacement for towers or satellites.
The apparatus changes what can be learned
A brain–computer-interface study examined two macaques learning to control a cursor through an adaptive decoder. As training progressed, task-relevant information became concentrated in a smaller neural population. The researchers’ modelling suggests decoder adaptation contributes to that change. Assistance alters the conditions of learning; it doesn’t simply read an intention that exists unchanged outside the interaction. This is evidence from monkeys and a particular experimental setup, not a demonstrated account of all human–AI collaboration.
Assistive algorithms influence neural representations in motor brain-computer interfaces
NASA’s work on sporadic E layers offers a different problem of observation. Metallic ions originating partly in burnt-up meteors can form ionospheric layers that redirect radio signals. A rocket and four released probes sampled one layer simultaneously, revealing structure that a single track couldn’t describe. The layers aren’t newly invented hazards; the new development is a better way of observing them.
NASA: a multi-point look inside radio-disrupting clouds
Two discoveries worth keeping, even if they predict nothing
A message in a bottle looks like a very simple instrument for studying currents. Yet its recorded destination depends on where someone finds and reports it. A 2023 study examined 2,225 bottles released through Students On Ice: 112 were recovered, 92% in Europe. Its model favoured coastal waters on the western side of the Atlantic. Unequal access and reporting offer one explanation, but the model also has limitations. You can’t settle the discrepancy just by declaring either the people or the simulation wrong.
Ridenour and colleagues: drift-bottle records and ocean circulation
Then there are sea slugs that shed their bodies, hearts included, and regenerate from the surviving head. Mitoh and Yusa’s 2021 observations establish that extraordinary capacity in a small number of animals. The suggestion that captured algal chloroplasts supply the energy for regeneration remains untested by that study. Separate experiments in another species show photosynthetic products reaching reproductive tissues. Those are connected findings, but joining them into one explanation requires evidence that isn’t supplied by either alone.
Mitoh and Yusa: extreme autotomy and whole-body regeneration
Cartaxana and colleagues: stolen chloroplasts and reproductive fitness
What I’d keep watching
Familiar distinctions—customer and provider, human and machine, organisation and environment—shape how we organise things. These examples each change something more specific: who can act, how an action is authorised, what can be connected, or how a result becomes visible.
For the actual horizon-scanning items, I’d look next for evidence of use: a revocable agent mandate honoured across institutions, a demonstrator becoming a service, or bidirectional charging working across manufacturers through changing seasons. For the science, I’m happy to leave room for curiosity without demanding a lesson for management.
Which of these would you follow—and what evidence would make you revise your view?
