Toolbox · inquiry through practice
How might AI support what you care about, and help you follow the relationship between action and consequence in a living place?
Spiralweb uses AI in developing and sharing this material. These tools grew out of that practice. We offer them as a first practical proposal: something you can try, adapt and judge through experience. Whether they help in your circumstances remains an open question.
“Right relationship” names a direction to explore. The work takes shape through people, places and what actually happens over time.
Start with something that mattersOur digital workbenchExplore the questions · AI’s resource use
A passion, a question, growing food, caring for a place, making a livelihood or an interest you want to give more room: begin with what is already alive for you. The starter and six tools below can help you find a manageable next step.
Help me explore something that matters in my life, work or a place I care about. Begin with what I already do, know or have tried. Ask one question at a time and use my words. Help me choose a useful next step and offer tools when they fit. Read my intention fairly before challenging an assumption. Keep observations, explanations, possibilities and checked evidence distinguishable. Say what you have actually read and what remains uncertain. Help me see when a source, another person, a practical observation or a pause would move the inquiry forward. When I say “For next time”, help me save my question, what changed and my next small step. Begin by asking what matters to me today.
Download the pocket starter (.txt) · Download the fuller starter (.txt)
The text can also be selected and copied directly. If your AI can read links, ask it to read this page and use the starter’s working preferences in your conversation. Ask it to say which material it could actually access.
Use the tool that helps with the question in front of you. Return to it after a conversation, observation, correction or pause. The movements can take different paths.
Begin with a need, curiosity, responsibility or passion already present in your life. Choose something small enough to explore.
Help me find a useful question in this situation. Start from what matters to me and what I already do. Ask one question at a time.
Bring a relevant observation, the conditions around it, a constraint and what you still need to understand. Select what belongs in this conversation.
Help me give you enough context for this question. Distinguish what I observed, what I think it means and what we still need to know.
Compare explanations and make a provisional working wager. Choose a check through a source, observation, conversation or appropriate practical trial.
What are plausible explanations? Help me express a working wager and choose a small check. What finding would make us change our view?
Ask a person or another AI for a fair reading and a specific challenge. Another AI’s agreement is another reading; verification requires suitable evidence.
First tell me how you understand my intention. Then identify a useful challenge, a missing perspective and a way to investigate it.
After an encounter or trial, describe what happened and why your understanding changed. The practical result and the learning result can differ.
Here is what happened. Help me distinguish the result from what I learned. Which assumption should I keep, revise or retire?
Keep the smallest useful note in a place you control. Bring it to the next conversation, with a relevant source or observation if needed.
For next time: help me save my question now, what changed and why, and my next small step. Let me correct the note.
How Spiralweb’s texts, tools and public architecture develop through sustained human–AI inquiry.
We use “workbench” for the working arrangement that brings people, sources, conversations, AI systems and developing understanding into relation. It gives an inquiry continuity across many turns. This names our practice; others may already work in similar ways and use other names.
In our practice, the human participant is part of the workbench. Lived experience, intuition, questions and judgement shape what we investigate and what is ready to share. AI helps us retrieve, compare, translate, question and formulate. Sources, critical friends and encounters beyond the screen can change the question and the form it takes.
Our workbench includes a digital library, earlier conversations, working documents, protocols, versions and the two websites. We select material for the present question, ask AI to revisit earlier work from an updated position, and keep what the comparison changes. A correction can become part of a revised text or tool, available for another turn.
Like sediment settling from moving water, these traces can gradually form ground for later inquiry. They preserve where an idea came from, what challenged it and why it changed. We keep earlier conclusions open to reconsideration. This is what we mean by a spiral: returning to a question with new experience, from a changed position.
A workbench is wider than any one AI context window. Material may be stored and accessible without being read in the current conversation. We distinguish material that is located, partly read, fully read or compared with other sources; verification is a further task. Later sessions may need context reconstructed. Richer context within a conversation does not establish lasting learning or complete memory in the model.
Our inquiry asks whether digital capacity can remain answerable to the living conditions it concerns. A field question may bring local observation into conversation with soil, water, species, work, research and institutional conditions. We call this pixel precision: making particular connections more discernible while keeping their context. It can make complexity easier to discuss and act within; it can also reveal complications we had overlooked.
Context can help expose a misunderstanding and recover an earlier correction. It can also preserve an error or reinforce our own assumptions. Whether this workbench supports better judgement is a working question. A formulation that feels faithful can help the inquiry move; its claims still need appropriate evidence and criticism. The living field must remain able to contradict our account.
Papers holds the public, versioned inquiry. Spiralweb.earth presents the Association and current field relationships. AnchorPoints hold relational continuity. The PG Ledger can preserve an agreed, traceable memory of a bounded field or relationship; Flow Desk helps the Association coordinate agreed work. These have distinct responsibilities. Selected material connects them without making the whole workbench public or turning an AI interpretation into a local decision.
This architecture is being prepared to support stewards and young people in developing practice knowledge through observation, dialogue, action and return. The materials are available; their usefulness in further local settings remains to be learned through use. People decide which knowledge may enter which AI service and what can travel beyond its originating setting.
In the Association’s next phase, we wish to explore decentralised arrangements: locally held digital memory, websites or other knowledge spaces, and AI systems working with selected material under locally chosen conditions. A locally running language model could be one component. This is a direction for development, rather than a capability we already provide.
The possibility is a polycentric ripple: learning moving between distinct centres of practice, with context, consent and uncertainty preserved. Knowledge from a place can inform a bioregional conversation and return as questions and possibilities. Each place retains its authority to adapt, question or decline what arrives.
A short continuity note is one small tool within this wider arrangement. The starter and “For next time” prompt above offer an accessible beginning; the workbench can hold much more than the note.
Illustrative dialogue. A steward wants to reduce daily watering work while keeping young plants healthy.
“Could mulch help? I already water every morning, and carrying the water takes time.”
The AI helps ask about the soil, plant stage, available material and who would prepare and carry it. A possible wager emerges: a small mulched patch may hold moisture longer and reduce watering work. An appropriate trial is agreed with the people responsible for the nursery.
After observing the patch, the steward returns: moisture seems to last longer, but collecting mulch has added work for someone else. The question changes: which locally available material could help plants while making the whole arrangement easier to carry?
The practical result concerns the nursery. The learning includes noticing an overlooked cost, revising an assumption and choosing what to investigate next. Those are different things worth remembering.
AI helps us compare, imagine, formulate and question. What happens in a place or relationship can change our understanding in ways that dialogue alone cannot settle. In our practice, an inquiry may move between AI dialogue, observing a place, talking with others and returning when something has changed.
We distinguish observations, interpretations, proposals and what has been checked. A situated observation and a question people can investigate may become a starting point for citizen science. Conditions, sources and uncertainty help others understand what the observation can tell them and which decisions it might inform.
What belongs in a conversation depends on the service, the purpose and permission to share the material. A prompt can guide the dialogue; data handling also depends on the provider and local arrangements. Decisions affecting other people, shared resources or safety need the relevant human authority and expertise.
In work we share, we aim to make our understanding and AI’s contribution visible: what we asked, what it offered, what we kept or changed, what we checked and what we can explain. Courses and schools may have their own agreed ways of showing that contribution.
People can care for the same field through different roles. A grower, teacher, habitat steward or guide may find a shared concern while seeing different parts of it. AI can help make those differences and possible common ground legible. People decide which question to pursue together, what knowledge can be shared and who holds each decision.
Participation also takes time and resources. An agreed project may provide paid learning time, relief cover, connectivity, materials, transport or specialist support. Stewardship economy asks how continuing care and the people carrying it can be properly supported. Rotation can make room for learning while someone carries the work at home.
Our proposed approach to place-based work keeps three readings distinct: the living place, people’s capacity to continue, and the agreements holding shared work. Progress in one cannot cancel harm in another. These three readings can lead to support, revision or pause. The five PG Ledger formats offer light tools for a review, observation, action, decision or resource flow. Each format has a different use; the record can be as light as the work and its responsibilities allow.
For active place-based practice, a light monthly review is the proposed starting rhythm, adapted to local needs, seasons and capacity. The other four formats are used when there is something worth observing, acting on, deciding or tracing financially. An urgent need calls for attention when it arises.
You can use and adapt these public tools independently and choose what to share. Their use creates no automatic relationship with the Association or promise of support.
The texts below approach different parts of our human–AI practice. They make its direction, questions and limits available for discussion. A connection may open through reading, a conversation with someone you know or an AI dialogue, if useful to you.
AI runs on physical infrastructure. Its use requires electricity, and data-centre cooling can require water. That belongs in this conversation too. The table offers a sense of scale for repeated use over approximately 1½ years.
| AI responses per day | Responses over the period | Electricity (kWh) | Direct cooling water (litres) |
|---|---|---|---|
| 100 | 55,000 | 17–220 | 20–240 |
| 250 | 137,500 | 41–550 | 45–600 |
| 500 | 275,000 | 83–1,100 | 90–1,200 |
These are calculated scenarios, not a measurement of Spiralweb’s use or an individual ecological footprint. Actual use can fall outside these figures. An AI response is not a standard unit: length, model, reasoning and other operations change the resource demand.
Electricity is calculated at 0.3–4 watt-hours per response. These rounded reference points draw on a study of inference and reasoning workloads; they are not universal lower and upper limits.
Water is calculated using a single illustrative factor of 1.1 litres per kWh, approximately derived from the cooling-water and energy figures in Google’s 2025 inference study. The water columns therefore track the electricity scenarios. They do not describe the full variation between data centres, climates or cooling systems, and exclude water used in generating electricity.
The figures are rounded. They cover estimated operational electricity and direct cooling water; they do not allocate model training, equipment manufacture or personal devices. Image, video and extended agent work require separate estimates. These assumptions give us something concrete to discuss while provider-specific measurements remain necessary for a proper account.
Our use of AI is a choice we have made. The resources it draws on remain part of that choice. We want to keep asking whether the work it supports serves the living places and people we care about, and what amount of use is proportionate to that work.
Further reading · follow a connection
Our public account of how we work with AI: what it contributes, what requires checking and who remains answerable for what we share.
A practice of sustained co-inquiry rooted in a person’s life and place. Its founding experience offers a starting point for exploration; how it travels needs to be learned through use.
Makes the working arrangements explicit: context, reading status, correction and a human-held workbench. It offers a form for inquiry without certifying right relationship.
Who has the final say when a conversation becomes an action or a public text? This section asks what it takes for human judgement to remain consequential: able to revise, stop or take responsibility.
How doubt, friction and critical friendship can become correction. The question is whether a change in understanding also changes what happens.
How situated experience can enter a wider inquiry while its conditions and uncertainty remain visible. Local observation and wider evidence have different work to do.
Brings attention back to living conditions and the capacity to respond. It asks what care requires when bodies and places carry the consequences.
Follows learning through continuing responsibility for a place. Returning over time allows actions, consequences and changing understanding to be read together.
Places inquiry alongside the practical care of land and water. What a living place needs becomes a question to revisit through encounter and shared work.
For the wider architecture, The Living Field describes how learning, relationships and public memory connect; The Protocol Habitat explains the different roles of foundations, protocols, situated applications and tools.
These working materials are available for adaptation. Wording, examples and language can change through local use, with useful revisions kept in the workbench of the people developing them. This version is in English; further language versions can develop with local speakers and their practical experience. Accounts of use may later join the shared collection with context and permission.