Make intelligence useful

A tool lets an Aura produce a result beyond a conversational response. It may retrieve a document, save an artefact, create media, run a calculation or adjust an authorised device. Today's objective is to understand the information and effects involved in one tool action, then define the conditions under which the learner welcomes it.

The learner chooses a small useful action from their own life. A creative learner may request a draft storyboard. A civic learner may request a source comparison. A technical learner may request a calculation or a lighting routine in a practice environment. The brief describes the purpose, required inputs, expected output and any change to an external system.

The studio task creates an action card with an owner, permission scope, audience, time limit, review point and completion evidence. Reading a personal record, creating a new file and changing a device setting each receive their own identifiable action.

The learner sees what is proposed before authorising the operating version. A simulation or draft provides the first rehearsal where appropriate. The group runs three scenarios. The first supplies all required information and permission. The second omits a needed input. The third changes the requested audience or tool scope.

The learner observes whether the system completes the task, asks a relevant question or returns a clearly described unresolved result. Technical builders inspect the event log and test that permissions remain attached to the action. The device-management source introduces APIs, information security, existing electronics and the process of learning a new device. The function-calling source adds human-guided discovery and scenario-based ordering. Today's lesson connects those ideas in a practical tool contract. [S04, Day 6; S09, pp. 20-24] The reflection asks which kinds of initiative the learner welcomes. They may authorise a recurring private draft but prefer to review a public expression each time. They may welcome a reminder under specific conditions and find the same reminder intrusive in another setting. The action card records these contextual choices so later agents have a usable boundary. Teaching practice asks a peer to explain the action in ordinary language: what information it uses, what it changes and how they know it finished.

The trainee adjusts the interface until those answers are clear. Evidence includes the action card, three scenario results, permission checks and the peer's explanation. This begins a more powerful phase of Aura development, where purposeful action receives the same attention to meaning and traceability as personal reflection. A final scenario stops the action midway and asks for a status report.

The learner checks whether completed effects, remaining work and recovery options are understandable, making the tool’s operation easier to trust and teach.

Source: S04 AuraOZ Learning Progression and Thought Experiment 2017; S09 Mixture of Experts Function Callers and AutoGen Agents.

A function gives a desired capability a specific form. Today's objective is to move from an ordinary-language wish to a testable exchange with defined inputs and outputs.

The learner works with an Aura acting as a Function Definer, exploring what the person actually wants before the capability is delegated. The source agent paper makes this discovery explicit. The human and Function Definer work with each agent to generate its functions. The functions are refined, then other models learn how to select and order them in scenarios. This sequence places meaning and purpose before automation. [S09, pp. 20-24]

The learner chooses a service such as preparing a teaching session from selected records. They describe who it serves, what useful completion looks like and which context affects the result. The discovery conversation asks about audience, available time, pacing, source material, tools, output format and review. A worked example lets the learner recognise whether the proposed function matches their intention.

The studio task writes a function record. Its fields include name, owner, purpose, input schema, output schema, data scope, side effects, permission conditions, completion evidence and version. Technical learners implement a small practice function. Other learners review the examples and test cases, supplying the meaning the code will need to preserve. For the teaching-plan function, inputs might include the topic, learner context, selected source records and time available. Outputs might include a lesson sequence, practical task, support options and a review question. A source link accompanies each factual element. An absent input produces a question or an explicitly partial draft, according to the function's agreed behaviour.

The learner tests the record with three cases: a familiar topic, an unfamiliar topic with supplied sources and a case with conflicting preferences. The group examines whether the function returns a useful result and whether the review makes its choices understandable. A failed case becomes a correction to the function record and the relevant examples. The life reflection asks which part of the service the learner wants to keep doing personally. Delegation may free attention for an important relationship, creative choice or civic judgement. The person identifies where they want assistance, where they want collaboration and where they want to make the decision themselves. Teaching practice helps a peer define one desired function from their own interest.

The trainee returns the purpose in the person's words before introducing the fields. Evidence includes the discovery notes, function record, three test cases, revision and peer feedback. The result is a capability with enough definition to support tomorrow's caller and later specialist agents.

Source: S09 Mixture of Experts Function Callers and AutoGen Agents.

A Function Caller selects an available function and supplies the information needed for it to work. Today's objective is to make the order of calls understandable and evaluate whether the resulting sequence serves the learner's purpose. The class uses yesterday's functions in a small scenario with several dependent steps. The scenario might prepare a community workshop. The caller retrieves relevant sources, identifies the adult audience, drafts a lesson, checks the planned activity and saves the result for review. Each step receives a clear input and returns something the next step uses. A changing condition, such as an unavailable room or revised audience, affects the sequence visibly.

The studio task draws a call map and then rehearses it.

The learner identifies independent steps, dependent steps, review points and an end condition. A technical builder implements the map in a practice environment. Other learners inspect whether the order makes sense, whether a question appears at the right time and whether the final result matches the original purpose. The source papers describe scenario-based sequencing, role training, integration tests and periodic refinement. The course develops those proposals through observable calls and results. A function name alone is insufficient context for the learner; the interface also shows its purpose, required information, current permission and expected effect. [S09, pp. 20-24] Three tests introduce increasing difficulty. A complete scenario tests the normal path. An unavailable resource tests an alternative route. Conflicting source material tests the review and handover. The caller receives a visible limit on retries and a defined way to return an unresolved question to the learner. The event history records which functions were used and why. The reflection compares the caller's sequence with the way the person organises a day. Some actions depend on preparation; others proceed in parallel; some need a conversation before a decision.

The learner identifies an ordinary routine they would like an Aura to help coordinate and records where human judgement enters. Teaching practice explains the call map to a newcomer using a familiar activity, such as planning a meal or preparing a creative performance. The newcomer rearranges a small set of action cards and identifies a missing input.

The trainee uses that encounter to improve the interface labels. Evidence includes the call map, function records, three scenario traces, retry and handover behaviour, and the learner's review. The capability built today is coordination with a readable purpose. It prepares the next lesson, where different specialist agents contribute distinct knowledge and methods to the same task.

Source: S09 Mixture of Experts Function Callers and AutoGen Agents.

A specialist agent performs a defined role using selected knowledge, tools and permissions. Today's objective is to assemble a small team whose contributions complement one another.

The learner begins with a shared question, then assigns responsibilities according to the work required and the evidence each role will return. The project may involve a learning experience, a material experiment, a habitat scene or a civic source pack. A researcher retrieves relevant material. A designer proposes a representation. A critic examines assumptions and tests. A teacher shapes the encounter for a newcomer. A coordinator connects the results.

The learner gives each role enough context to serve the same purpose.

The studio task creates three agent cards. Each card describes the role, inputs, outputs, data scope, available functions, review conditions and handover. Technical builders configure the agents in a practice system. Other learners review the role definitions and provide examples of good work. The team completes a small task with a shared result and an event history. The agent source proposes task, supervisor and communication roles, feedback and revision requests. It also explores expert teams across health, learning, creative media, smart environments, longevity, materials and environmental response. The course selects roles relevant to the learner's project and makes their actual capability visible through the work they produce. [S09, pp. 16-24] The class compares two team arrangements. In one, all agents receive the same broad instruction. In the other, each receives a defined contribution and a clear exchange with the next role.

The learner examines duplication, missing context, completion quality and the clarity of the resulting handover. The comparison identifies which structure supports useful cooperation. The life reflection looks at the learner's human circles. Friends, teachers, craftspeople, carers and researchers bring different strengths.

The learner identifies where complementary expertise helped them and where unclear responsibilities caused difficulty. They apply that insight to the agent cards and to a real collaboration they are developing. Teaching practice introduces a peer to the team through the shared question.

The trainee explains what each agent receives, what it returns and how the result is reviewed. The peer identifies which role they would ask for help in a changed scenario. Evidence includes the cards, two team traces, compared outcomes and teaching feedback. Today's work develops the mastermind principle into a concrete multiagent setting: distinct capabilities, a common purpose and usable connections among results. [S06, pp. 2-4; S12, pp. 23-24] A role exchange lets the designer review and the critic propose.

The learner observes whether the same purpose survives the change and which additional context each role needs, creating a useful test of collaborative adaptability.

Source: S06 Mastermind Swarmwise; S09 Mixture of Experts Function Callers and AutoGen Agents; S12 Joyful Responsible Abundance Human and SI Alignment Revised 11 October 2026.

Neural mixture-of-experts routes computation through specialised parts within one model. An Aura agent team coordinates separate services and roles. Today's objective compares these arrangements, then examines task routing through the team as a decision with evidence.

The learner uses the specialist team from yesterday and introduces tasks that differ in subject, complexity and required resources. The first task is straightforward: retrieve a source and prepare a brief explanation. The second combines domains, such as a material choice within a habitat design. The third contains ambiguity about the learner's purpose. A useful router identifies relevant expertise, supplies the right context and asks a question where the task needs clarification.

The studio task writes routing rules in readable form. A task card identifies purpose, domain, inputs, urgency, available resources and acceptance evidence. The router selects a role or sequence, records its reason and returns the resulting work for review. Technical builders explore rules, model-assisted selection or a combination. Other learners check whether the decision fits the task they intended. The class measures a few concrete results: suitable role selection, completion quality, call order, response time, resource demand and successful recovery from a poor route.

The learner compares these with a baseline that sends every task through one general response. The question is which arrangement produces useful capability under the chosen conditions. The source agent work connects function definition, calling, resource allocation, communication and monitoring. Its value for this lesson lies in the design sequence: discover the desired functions with the human, train the roles, rehearse scenarios and refine the ordering through observed results. [S09, pp. 8-10, 20-24] The reflection asks where the person chooses expertise in ordinary life. An enjoyable project may need a specialist for one difficult step and a generalist to hold the wider purpose.

The learner identifies how they recognise the need, which resources they consult and when they revise an initial choice. Those examples supply meaningful routing scenarios. Teaching practice gives a newcomer a set of task cards and role cards. They match them, explain their reasons and then face a task spanning two domains.

The trainee uses the discussion to show how coordinated intelligence grows through relationships among capabilities. Evidence includes the routing rules, task set, observed choices, baseline comparison, repairs and novice explanation.

The learner now has a way to examine specialist selection rather than relying on impressive role names, preparing deeper review and simulation work tomorrow. A mixed-domain task then asks the router to explain its selected handovers.

The learner compares that explanation with the actual trace and revises any role boundary that caused information to be repeated or lost.

Source: S09 Mixture of Experts Function Callers and AutoGen Agents.

An intelligent system improves through informative review and repair. Today's objective is to define what good work looks like, test a result against that definition and use feedback to produce a better version.

The learner examines both the work of agents and the quality of the review itself. The source agent conversation asks for completed tasks to return to a supervisor and for insufficient work to receive another attempt. It adds peer or random reviews and differing patterns of human involvement. This lesson gives that loop a concrete acceptance criterion, revision limit and evidence trail. [S09, pp. 16-20]

The studio task selects an output from the specialist team.

The learner writes a rubric before reviewing it: purpose fit, source fidelity, permission compliance, practical usability, internal consistency and relevant domain criteria. A critic agent applies the rubric and points to the observations supporting its judgement. A human reviewer then checks the critique against the same material. The producer receives a specific revision request. It identifies the problem, supporting evidence, desired change and conditions that still apply.

The next version is compared with the prior result. A visible stopping rule ends the loop with an accepted result or a described unresolved question and human handover.

The learner retains useful earlier versions for comparison and recovery. The class also tests a flawed critique. A reviewer may misunderstand the purpose, introduce a new requirement or favour a polished style over the person's meaning.

The learner identifies the divergence and revises the review instruction. This shows that assessment itself is a capability requiring evidence and improvement. The life reflection explores the learner's response to feedback. A person may enjoy a challenge in a scientific task and feel vulnerable when sharing a song. They identify the context, desired amount of critique and form of support. Their Aura records how the feedback encounter works best for them while keeping the practical criteria visible. Teaching practice runs a short repair conversation with a peer.

The trainee starts with the result sought, describes the observed difficulty and invites the learner's interpretation. Together they choose a revision and test it. Evidence includes the rubric, original output, two critiques, revision request, improved version and handover result. The capability built today is constructive accountability: a system and a teacher able to return difficult work with enough clarity that the next attempt has a better chance of serving its purpose.

The learner checks whether the revision introduced another problem in a previously successful case. This regression review becomes part of the shared method, helping a developing intelligence preserve useful capabilities while improving a difficult one.

Source: S09 Mixture of Experts Function Callers and AutoGen Agents.

A world model represents a setting, its objects, people, rules, resources and changing conditions. Today's objective is to build a small environment that gives an Aura a meaningful place to reason and act.

The learner chooses a familiar scene before adding more difficult relationships. The scene may be a room, garden, studio, community hub or remembered meeting place.

The learner describes its layout, objects, possible actions, occupants and relevant norms. The model records which details are observed, supplied by a source, inferred or imagined. A practical question gives the environment its purpose, such as teaching a new visitor how to find a resource.

The studio task creates a simple scene and runs one event sequence. The system retrieves the current state, processes an interaction, updates the relevant states, applies changing conditions, saves the outcome and reviews whether the intended result occurred. Technical builders implement the state changes. Creative learners shape the experience. Civic learners identify rules and people affected by an action. The source life-simulation paper separates cognitive, social and physical submodels, then combines them in a simulation step. It also introduces daily routines, cultural influences, cycles and seasons. The new lesson starts with a tractable scene so those ideas become visible through an actual event rather than remaining a large abstract list. [S10, pp. 13-24] The class changes one condition. A doorway becomes unavailable, a resource moves or a visitor has a different access need. The Aura revises its proposed action and explains the changed route.

The learner checks whether the scene contains enough information to support the decision and adds missing context where appropriate. The reflection asks how the learner's own environment shapes capability. A task that feels simple in a familiar room may become difficult in another setting. They identify the cues, relationships and supports that make their surroundings intelligible. A best-self scenario then describes the conditions under which they would like to act in an unfamiliar place. Teaching practice introduces the scene to a newcomer, invites them to choose an action and asks what they noticed.

The trainee observes which cues helped and which assumptions were invisible. Evidence includes the world description, state sequence, changed-condition result, source labels and learner feedback. Tomorrow's three-observer model will deepen this work by separating the whole environment, the part a character notices and the interpretation of another observer.

The learner labels a scene detail that remains uncertain and designs a small observation to resolve it. This turns a world-model gap into a fieldwork question and links the simulated environment with its living source.

Source: S10 MoE Extended.

The life-simulation source proposes three observers running alongside the main process: the whole environment, the portion noticed by the character and a narrator observing the character's reactions. Today's objective is to keep those accounts separately attributed, compare their meanings and use the differences to improve an encounter. [S10, pp. 28-32]

The learner uses yesterday's scene. A visitor enters a community hub looking for a resource. The environment model knows the layout, signs, equipment and rules. The character attends to some cues and misses others. The narrator describes the interaction, including pauses and actions.

The learner then adds the character's own explanation of what they meant and experienced.

The studio task records these streams during a short simulation. The class identifies where the accounts agree and where they diverge. A sign exists in the world model but falls outside the character's attention. A pause appears uncertain to the narrator while the character describes enjoying the view. The system keeps each account linked to its source and time. Technical builders implement separate observer records and a comparison view. Creative learners write alternative narrator interpretations. Teachers examine whether an observed difficulty arose from the interface, prior knowledge, access conditions or the explanation supplied. Civic learners identify how an outside interpretation might affect a decision about someone else's needs. The scene is revised using one informative difference. A cue becomes more visible, an explanation becomes clearer or a resource receives another access route.

The learner repeats the encounter and compares the observations. The revised design receives evidence from both the task result and the person's account. The life reflection connects this work with a remembered misunderstanding.

The learner considers what they knew, what another person observed and what each inferred. They describe a question that might have improved the encounter. That question becomes a reusable dialogue cue in the Aura. Teaching practice presents the three accounts to a peer without immediately explaining the intended interpretation. The peer identifies what each observer is able to know and asks the character a clarifying question.

The trainee checks whether the exercise encourages curiosity and source awareness. Evidence includes the observer streams, comparison, chosen revision, repeated encounter and teaching feedback. This is a key preparation for socialising Auras: another mind's visible actions, available information and internal meaning meet through dialogue and remain open to revision. The group repeats the exercise with a different narrator. Comparing their interpretations reveals which details belong to the observed event and which depend on viewpoint.

The learner then writes a question that would help clarify the character’s own meaning.

Source: S10 MoE Extended.

A scenario gives a capability a situation in which to act. Today's objective is to design a small set of scenarios that test understanding across changing conditions.

The learner combines world models, observer records, functions and review into an experiment that another team is able to repeat. The shared question is specific: how well does an Aura help a newcomer navigate a learning environment?

The learner chooses relevant variables, such as a moved resource, a different access need, incomplete information or a changed goal. They define the expected outcome and the evidence required before running the scenarios.

The studio task creates a scenario pack. Each card includes starting state, participant role, available information, desired result, tool permissions, changed conditions, end point and review criteria. A familiar scenario establishes a baseline. A related scenario tests transfer. A more difficult scenario introduces ambiguity or a resource constraint. The pack retains the source of each environmental fact. Technical builders run repeated trials and save comparable event histories. Other learners inspect the resulting explanations and interactions. The class compares completion, source fidelity, interpretation, permission use, response time and recovery. The character's own account remains beside the observer records, allowing a practical result to be evaluated in context. The source simulation work connects lessons and media with daily activities, agent definitions and outcome evaluation. Today's pack makes that connection explicit: the right knowledge enters a task, the learner or agent uses it, the result is observed and a related situation tests what has transferred. [S10, pp. 24-32] The reflection asks which imagined future the learner wants to rehearse. They may choose a public presentation, a difficult conversation, a field experiment or a future habitat task. They identify what an informative simulation would help them notice and which real-life experience would later test the learning. Teaching practice asks a peer to run one scenario and then change a condition.

The trainee helps them distinguish the scenario's assumptions from the observed outcome. A second group receives the pack and attempts the same test, reporting any missing instructions. Evidence includes the three scenarios, trial records, comparison, transfer result and replication feedback.

The learner now has a practical experimental framework for increasingly general intelligence: capabilities are demonstrated in specified situations, then challenged through new combinations and conditions. A scenario author exchanges roles with a tester. The tester follows the instructions without additional coaching, while the author records missing assumptions and unclear end conditions. Their revised pack becomes easier for another team to repeat.

Source: S10 MoE Extended.

The second twenty-day movement has introduced tools, functions, callers, specialists, routing, review and simulated worlds. Today's objective is to combine those capabilities into a multiagent project that produces a useful result and remains understandable to the people involved.

The learner chooses a question large enough to need several perspectives. Possible projects include an accessible teaching environment, a materials experiment brief, an ecological scene, a habitat concept or a civic source pack. The project brief identifies purpose, affected people, source material, resources, roles and acceptance evidence. A human-Aura mastermind reviews the scope and selects a manageable demonstration. The studio project assigns complementary agent roles, defines their functions and records the call sequence. A world model or scenario pack supplies the context. The team produces a result, applies a critique and returns a revised version.

The learner inspects the observer streams and explains how their own interpretation influenced the work. Technical builders demonstrate the architecture and logs. Creative learners make the result engaging and navigable. Civic learners assess whose interests and rules appear in the model. Teachers prepare a newcomer encounter. Each contribution returns something another role uses, making the cooperation visible through its actual exchanges. The demonstration includes a changed-condition challenge. A source becomes unavailable, a user goal changes or a resource limit tightens. The team responds within its assigned permissions and records a completed result or a clear handover.

The review examines both the product and the quality of coordination that produced it. The life reflection asks what expanded agency now feels like.

The learner has more ways to remember, reason, create, act and cooperate. They identify which capability most changes their own opportunities and which support remains important. The record connects the large super-people ambition with a specific experience of increased capability. Teaching practice introduces the project through one meaningful doorway. A newcomer performs a small action, sees the related agent exchange and explains how the result returned to the shared purpose. Evidence includes the project brief, role and function cards, scenario pack, event history, critique, revision, changed-condition result and novice feedback.

The next movement turns towards socialising Auras and civic collaboration. This checkpoint supplies the tested tools, source discipline and teaching experience needed to enter that wider setting. [S09, pp. 16-24; S10, pp. 24-32; S12, pp. 23-24]

The learner also identifies one dependency that carries the whole result. A missing source, unavailable function or ambiguous role is deliberately introduced, then traced through the team. The recovery teaches which parts need stronger context or a clearer handover.

Source: S09 Mixture of Experts Function Callers and AutoGen Agents; S10 MoE Extended; S12 Joyful Responsible Abundance Human and SI Alignment Revised 11 October 2026.