Help people do more
Optimise for what people can learn, build, repair, organise and become, not merely what they can buy or scroll past.
Manifesto v0.1 · July 2026
Intelligence must make us more capable.
A simple question for a complicated age: does progress give every child a real chance to learn, make things, feel secure and improve the world around them?
Official Run 1 available · research completed 16 Jul 2026 · next run eligible 16 Oct 2026
Six years old · somewhere in the world
“How do I change the world?”
When a six-year-old asks how to change the world, the answer should open a real path forward, not an advert, a locked door or a comforting lie.
The Test asks what a child could actually do with the best knowledge available today. Faster machines and larger fortunes only count when they create more choices for ordinary people.
What we believe
Progress should widen people’s choices.
This is not a promise that technology will save us. It is a set of practical rules for judging whether new tools are making life safer, fairer and more useful.
Share what progress creates
Cheaper energy and production mean little if access is enclosed by monopoly, geography, debt or subscription.
Public money should create public benefit
Publicly funded intelligence applied to publicly funded machinery must produce publicly accessible benefit.
Show the evidence
Publish assumptions, measurements, failures and uncertainty. Update the programme when reality disagrees.
Help people through change
Income, services, taxation and bargaining power must evolve before job displacement outruns social stability.
Open where safe
Default to open science and fair access while containing weapons, dangerous biology and genuinely hazardous processes.
No birthplace lottery
A child’s country, language or family wealth must not determine whether they can use civilisation’s best tools.
People choose the direction
AI may search the possibility space. People must still choose the goals, limits, ownership and acceptable risks.
Getting from here to there
The future only matters if people can reach it.
Technology can change much faster than wages, public services, energy systems and democratic decisions. We need to build the crossing before asking people to walk across it.
Make useful things cheaper
Drive down the physical cost of intelligence and production.
- Abundant clean energy and stronger grids
- Efficient models, chips and memory
- Cooling, conductors and heat reuse
- Robotics, materials and flexible factories
- Precision agriculture and circular resource flows
Watch what could break
Track where acceleration could fracture the economy or concentrate power.
- Speculative overinvestment and financial contagion
- Job loss arriving before cheaper essentials
- Tax-base erosion and weakened public services
- Ownership concentrating faster than productivity
- Energy, water, mineral and grid bottlenecks
Make sure the gains reach people
Convert productivity into security and agency for households.
- Modernised benefits, including UK Universal Credit
- Universal basic income or social dividends
- Universal essential services
- Taxation that follows capital and automated value
- Public compute, open standards and fair licensing
A practical proposal
Imagine if scientific progress belonged to everyone.
The Common Foundry would be a shared network of scientists, automated laboratories and workshops, working on public problems and publishing what they learn for everyone to use.
Not one giant laboratory. A shared workshop for discovery.
Specialist sites would share a charter, data fabric, safety rules and experimental interface. One site explores materials. Another fabricates. Another performs destructive testing. A separate site must reproduce major results before the work is treated as real.
The Foundry optimises the whole chain, including the machines and processes needed to manufacture a discovery at scale. A miracle material that requires impossible production is not abundance. It is a very expensive paper.
The rule: publicly funded intelligence applied to publicly funded machinery must produce publicly accessible capability.
Define a public challenge and hard constraints
Generate hypotheses grounded in evidence
Simulate candidates and failure modes
Synthesise materials or fabricate prototypes
Measure, stress, break and learn
Redesign both product and production process
Replicate independently at another site
Publish, license fairly and support deployment
Proposed mission 01
Thermal Abundance
Jointly optimise thermal materials, chip efficiency, cooling systems, power electronics, heat exchangers, waste-heat reuse and low-energy manufacturing. A thousand compounding efficiencies may arrive sooner than one miracle.
What is already real
Pieces of this future already exist.
None of these examples proves that everything will work. They show that important pieces already do. Evidence checked on 16 July 2026.
Autonomous materials synthesis
36 / 57
Berkeley Lab’s A-Lab realised 36 target compounds during 17 days of continuous autonomous operation.
Berkeley Lab publication record ↗AI-controlled chemistry
Plan → run
Coscientist demonstrated an LLM-based system that could design, plan and perform complex chemistry experiments using laboratory automation.
Nature, 2023 ↗Hypothesis engines
2026
Google’s multi-agent Co-Scientist generates, critiques and evolves scientific hypotheses, with ideas already being moved into experimental validation.
Nature, 2026 ↗National autonomous labs
$34m
US programmes are funding AI-paired autonomous laboratories for catalyst development, while the DOE’s Genesis Mission explicitly targets AI-driven labs and advanced manufacturing.
US Department of Energy, 2026 ↗Data-centre electricity
945 TWh
The IEA projects global data-centre electricity use could more than double to roughly 945 TWh by 2030.
International Energy Agency ↗Fusion ignition repeated
11×
LLNL reported an eleventh NIF ignition in June 2026. This is a research milestone, not yet commercial power.
Lawrence Livermore National Laboratory ↗Work will change
≈40%
The IMF estimates AI may affect almost 40% of jobs globally, complementing some work while reducing demand for other roles and potentially worsening inequality.
International Monetary Fund ↗Factories are not yet general
Islands
Labs can close narrow discovery loops, but no public institution yet integrates general hypothesis generation, materials, machine design, scale-up, independent replication and open deployment.
DOE autonomous laboratory programme ↗Design-to-production loop
Roadmap
NIST’s 2026 smart-manufacturing roadmap tracks AI across autonomous systems, additive manufacturing, digital twins, robotics, logistics and sustainable production.
NIST, 2026 ↗The honest bit
What could break the thesis.
A manifesto that cannot be falsified is just branding in a lab coat. We track both the failure modes and the developments that could neutralise them.
Reasons this could go badly
01
Compute investment outruns useful revenue and triggers financial contagion.
02
Energy, cooling, water and grid delays keep intelligence expensive and geographically concentrated.
03
Automation weakens wages and tax receipts before living costs fall.
04
Scientific systems optimise proxies, conceal failure or produce unsafe dual-use capability.
05
Patents, platforms and capital ownership capture the gains before democratic institutions respond.
Reasons the constraints may soften
01
Smaller models, specialised chips and better scheduling sharply reduce energy per useful task.
02
Thermal materials, liquid cooling and waste-heat reuse improve infrastructure economics.
03
Fusion, fission, renewables, storage and grids expand together rather than waiting for one saviour technology.
04
AI accelerates science, manufacturing and deployment, compounding thousands of modest gains.
05
Social dividends, universal services and reformed benefits preserve demand, dignity and political stability.
How we keep ourselves honest
Measure what changes for people.
Every claim should connect to something we can inspect. These questions guide the Test without pretending that missing evidence is a number.
Can people do more?Can ordinary people solve more meaningful problems?
Is everyday life safer?Are housing, food, healthcare, energy and income becoming more dependable?
Who gets to use the tools?Who can use the best tools, compute, education and manufacturing?
Who receives the gains?Where do productivity gains, ownership and bargaining power land?
Do we waste less?Useful output per joule, litre, kilogram, hectare and hour.
Can good ideas become real?Time from hypothesis to replication and deployable production.
Can systems survive shocks?Can systems withstand shocks without sacrificing the least powerful?
Can people shape decisions?Can the six-year-old participate, create, question and choose?
Version 0.1
Nothing happens unless we start making it happen.
This is an invitation to scientists, engineers, economists, policy designers, funders, artists, sceptics and citizens. Challenge the assumptions. Improve the evidence. Help turn the Common Foundry from a good sentence into public machinery.
Sources and evidence notes
Berkeley Lab: A-Lab autonomous synthesis record
Nature: Autonomous chemical research with large language models
Nature: Accelerating scientific discovery with Co-Scientist
US DOE: Achieving AI-driven autonomous laboratories
US DOE ARPA-E: autonomous catalyst-lab programme
NIST: AI and machine-learning roadmap for smart manufacturing
IEA: Energy demand from AI
LLNL: Achieving fusion ignition
IMF: AI and the global economy
Evidence is presented as a dated snapshot and should be periodically reviewed. Scientific milestones are not the same as commercial readiness. Occupational exposure is not identical to job loss.