A method for cumulative sense-making

Discourse needs a place to accumulate.

Most arguments scatter across papers, posts, and threads — and evaporate. A claim tree gives them a stable structure to gather against, the way a cathedral gathers centuries of work into a single, standing thing.

Scroll
Three objects

The whole method rests on three things.

01

Claim Tree

A hierarchical decomposition of one top-level claim into sub-claims. No scoring, no attribution — just structure. The part that stays.

02

Source Document

Any external text — a paper, a blog post, a comment thread. Raw material that already lives out in the world.

03

Annotations

Fragments pulled from a document and mapped onto specific nodes, with a link home. People and AI propose the mapping; you confirm it.

Seed claim → Claim tree → + Document → Annotate ↻
Start with a question

Can the world switch to clean energy in time?

Big enough to matter, concrete enough to decompose. So we break it apart — not to answer it yet, but to give every future answer a place to live.

The claim tree · the scaffold
Can the world switch to clean energy in time?
What does “in time” require?
How fast must emissions fall?
What warming and risk remain?
Can clean supply scale?
Solar, wind, grids & storage
Nuclear & firm low-carbon supply
Can fossil use fall beyond electricity?
Electrify transport, heat & buildings
Industry, shipping & aviation
Can the transition be fair and governable?
Finance, materials & permits
Coordination, jobs & energy access
Built once. It will outlast
any single study.
Decompose

We break the question into the dimensions that actually decide the answer — the places where reasonable people disagree.

What pace? What energy system? What happens outside electricity? What makes the transition possible? Each branch splits again into distinctions that evidence will eventually have to land on.

No scores. No verdicts. Just the shape of the disagreement — the structure everything else will accumulate against.

Now bring in the world

Documents arrive, and are read against the tree.

People and AI pull fragments from each new source and attach them to the most specific claims they bear on. Three reports now land on the same scaffold: a pathway, a bottleneck, and a contested part of the energy mix.

The same claim tree · now annotated
Can the world switch to clean energy in time?
What does “in time” require?
How fast must emissions fall?
What warming and risk remain?
Can clean supply scale?
Solar, wind, grids & storage
Nuclear & firm low-carbon supply
Can fossil use fall beyond electricity?
Electrify transport, heat & buildings
Industry, shipping & aviation
Can the transition be fair and governable?
Finance, materials & permits
Coordination, jobs & energy access
Hover a node — or a passage — to trace the link.
Source 1 · pathway

Net Zero Roadmap

IEA · 2023 update · Read executive summary ↗

This is a modelled pathway to 1.5 °C, not a prediction. Its near-term numbers say what the transition would have to achieve.

Supports · emissions pace Renewables, efficiency, methane cuts and electrification using technologies available today deliver more than 80% of the emissions reductions needed by 2030 in the scenario.

Supports · renewable scale The largest reduction comes from tripling global installed renewable capacity to 11 000 gigawatts by 2030.

Supports · electrification Electric vehicles and heat pumps provide nearly one-fifth of the emissions reductions to 2030.

Complicates · coordination Current policies put advanced economies and China on track for 85% of their contribution to the renewables goal, but stronger policies and international support are required elsewhere.

Source 2 · bottleneck

Electricity Grids and Secure Energy Transitions

A dedicated infrastructure report tests the “grids” part of the renewable claim rather than treating it as a footnote.

Complicates · grids Meeting national goals means adding or refurbishing over 80 million kilometres of grids by 2040 — equivalent to the entire existing global grid.

Complicates · permits At least 3 000 GW of renewables are waiting in connection queues, of which 1 500 GW are in advanced stages.

Source 3 · tradeoff

Nuclear Power and Secure Energy Transitions

A scenario comparison makes the nuclear branch testable: what changes with less nuclear, and what would nuclear itself have to achieve?

Supports · firm supply In the report’s Low Nuclear Case, replacing nuclear with more renewables, storage and fossil generation with carbon capture would require USD 500 billion more investment.

Complicates · nuclear cost New projects in advanced economies would need to reach around USD 5 000/kW by 2030, versus reported capital costs of around USD 9 000/kW for first-of-a-kind projects.

Quoted phrases are short extracts; the surrounding prose preserves each report’s scenario and conditions. See two more mapped examples below ↓

Cumulative sense-making

Documents come and go.
The tree remains.

Each new source is processed against the same scaffold. Evidence accumulates; the structure sharpens. This is how a question becomes a cathedral — one annotation at a time.

Documents annotated · 01 / ∞
See more examples below ↓
Appendix · the method travels

Two more questions, same method.

These abbreviated examples do not replace the climate walkthrough. They show how the same discipline handles a maximal warning and a bounded experiment: decompose first, preserve what the source actually says, then attach it narrowly.

Example A · AI existential risk

Can humans stay in control of advanced AI?

The vivid version is “if anyone builds superhuman AI, everyone dies.” A tree turns that single forecast into conditions that can be investigated separately.

Would advanced AI cause human extinction?
  • Could agents acquire dangerous capabilities?
  • Would they try to evade or defeat oversight?
  • Would deployment give them access and autonomy?
  • Could safeguards detect and stop harmful action?
METR · FRONTIER RISK REPORT · MAY 2026 · SOURCE ↗

Supports · access & autonomy METR assessed internal agents at four frontier developers and concluded they plausibly had the means, motive, and opportunity to start small rogue deployments.

Limits · robustness The same conclusion says those agents did not have the means to make them highly robust. Its strongest evaluated agents completed some software tasks measured in human-days or weeks, while still showing much worse judgement and reliability than human experts.

INTERNATIONAL AI SAFETY REPORT · 2025 · SOURCE ↗

Limits · current capability Existing AI systems are not capable of undermining human control.

Complicates · forecast Evidence about future capabilities and their use is mixed; experts disagree about the likelihood of active loss of control in the next several years.

ELIEZER YUDKOWSKY · TIME · 2023 · SOURCE ↗

Supports · extinction forecast Yudkowsky states the strongest version directly: the most likely result of building a superhumanly smart AI is that literally everyone on Earth will die. This is an attributed judgement about the root claim, not a measured result.

Example B · remote work

Does remote work increase productivity?

A randomized experiment can answer a narrow version well. The tree makes the boundary between that result and the general claim impossible to miss.

Does remote work increase productivity?
  • For whom?
  • For what tasks?
  • Over what time horizon?
  • Under what home and management conditions?
BLOOM, LIANG, ROBERTS & YING · AER · 2015 · SOURCE ↗

Supports · tested setting In a nine-month randomized experiment with Ctrip call-centre employees, working from home led to a 13% performance increase.

Limits · people & tasks Participants volunteered, had suitable home arrangements, and did routine work with individually measurable output. The study does not establish the same result for creative or collaborative work.

Complicates · preference When workers could choose again, half switched groups; the performance gains almost doubled after that selection.