FALKNER.XYZ
E

Edo365 — Education Toward Autonomy

tl;dr: Edo365 applies Open Autonomy to teaching: the goal isn't digitization or Scrum, but learners' real capacity to increasingly steer their own learning process. Sprint, Flap, daily stand-up, competency diagnostics and AI support are modules that can each be tested and dropped on their own, not one closed method. Structure recedes as capacity grows — the teacher stays central throughout, but shifts from directing to diagnosing and accompanying. Edo365 itself stays a testable hypothesis too.

1. Starting point

Edo365 is an education concept for expanding real autonomy.

It isn't primarily about digitization, self-organized learning, Scrum, AI or individualized worksheets. Those are tools.

The starting point runs deeper:

Education should expand a person's capacity to learn, decide, test, cooperate, and choose their own paths.

That makes Edo365 a practical application of Open Autonomy to education.1 Its three axioms are also the starting point for Edo365:

  1. No idea stands above the individual – not even this one. Edo365 itself, its methods and its software must not become a fixed idea either.
  2. Autonomy is not abstract freedom, but a capacity. Freedom can be granted; the capacity to actually use it has to be developed. Structure steps back as that capacity grows.
  3. Nobody can check everything. But everyone must have the possibility of acquiring the capacity to check anything. Learners should be able to test claims, sources, materials and AI, and to see and question their own learning progress.

A student isn't autonomous just because they're merely left a choice. They become more autonomous as they increasingly gain the abilities to choose and act meaningfully.

Autonomy isn't exhausted by granting freedom. The capacity to actually use that freedom has to be developed.

Edo365 therefore doesn't simply replace teacher control with learner freedom. It tries to gradually enable learners to dispose over their own learning process.

Psychologically, this approach connects to Deci and Ryan's self-determination theory, which describes autonomy, competence and relatedness as basic psychological needs.2 Open Autonomy isn't derived from that theory; but it offers empirical-psychological common ground for parts of the philosophical thesis.

2. An offer in small units

Edo365 isn't a closed system that has to be adopted as a whole. It's an offer.

It's meant to be modular and decentralized: Sprint, Flap, daily stand-up, competency diagnostics, generative tasks, the AI learning companion and software modules can each be tested, adopted, changed or dropped on their own.

That follows from Open Autonomy. An education concept that wants to foster autonomy can't demand to be adopted completely and unchanged.

Parts that prove themselves get developed further. Others fall away. New ones can be added. That holds for pedagogical elements just as much as for software.

Edo365 isn't mandated. It's offered, tested and appropriated.

3. The central problem: autonomy can't simply be granted

Open, self-directed teaching presupposes abilities that school is often meant to develop in the first place. Learners need to learn, among other things, to

  • assess their own learning level,
  • understand and choose tasks,
  • manage their time,
  • recognize difficulties and ask for help,
  • check results,
  • collaborate with others,
  • revise decisions.

Missing structure therefore doesn't automatically produce autonomy.

Structure should enable autonomy, and be able to step back as capacity grows.

That's the pedagogical expression of the minimally functional. Learning research describes a related principle as scaffolding: temporary support, matched to the current learning level, that's withdrawn as competence grows.34 The expertise reversal effect further shows that guidance which helps beginners can become unnecessary or even hindering for more advanced learners.5

4. The education paradox

Edo365 operates inside an institution that carries a basic tension within it: school is supposed to produce autonomy while itself being a compulsory institution. Children and adolescents can't simply leave it. Measured against Open Autonomy's exit principle, that produces high, partly unavoidable exit costs.

That's the education paradox:

How can an institution produce autonomy if participation in it isn't itself fully autonomous?

This paradox can't be resolved through a teaching concept. It stays visible in teacher requirements, diagnostics, AI and grading.

The deeper question remains open:

May a person be compelled into education so that they can later decide for themselves? And what happens when a present unwillingness produces a later inability?

Open Autonomy doesn't demand the abolition of school because of this. An existing institution can, despite its constraints, be the currently best available solution to a real problem. A compulsory school in particular can be justifiable if the alternative would more severely restrict the later capacity for autonomy. But that justification, too, stays a testable hypothesis — not a blank check for arbitrary compulsion in school.

Edo365 therefore accepts the existing institutional frame as a starting point, without declaring it unattackable. Within it, it tries to reduce avoidable compulsion, make requirements transparent, and gradually enable more disposal over one's own learning process.

The goal isn't to resolve the education paradox, but to generate as much real capacity for autonomy as possible within this tension.

5. As much structure as necessary

The teacher initially sets a frame of learning goals, subject content, sprint structure, materials, minimum requirements and ways to check progress. Sprint planning, too, initially happens largely through the teacher.

As capacity grows, parts of this structure can shift to the learners:

Structuring by others → supported self-organization → growing self-organization

How far this shift makes sense depends on the learner and the specific situation.

6. The sprint as temporary structure

The sprint, roughly one to two weeks long, isn't an end in itself, and Edo365 isn't an attempt to transfer Scrum unchanged onto school. What it adopts, above all, are the three pillars of its empirical process model: transparency, inspection, and adaptation.6

As a pedagogical forerunner, eduScrum shows how agile forms of organization can be transferred onto teaching.78 Edo365 differs from it through lighter roles, stronger individual diagnostics, generative differentiation, Open Autonomy as its foundation, and the explicit revocability of its own elements.

The sprint makes a learning process time-boxed, plannable and revisable. Its retrospective asks:

  • What was achieved?
  • What worked?
  • Where did difficulties arise?
  • Which structure was helpful, or unnecessary?
  • What should change?

That turns the organization of teaching itself into a testable hypothesis.

Learners get a voice here, and, as their capacity grows, can help shape the rules of coming sprints. Teaching thus becomes a small practice field for law from below: rules arise, get tested, and get changed.

The gap between teacher and learners doesn't disappear because of this, but it gets a regular place where it can shrink.

7. The Flap as a tool of appropriation

The Flap isn't primarily a task planner. It's meant to gradually turn an externally planned learning process into one's own learning process.

Sprint goal, tasks, battery gauge, problems, Definition of Done, Definition of Fun and reflection make different aspects of learning visible.

The battery gauge is a low-threshold self-assessment of one's own ability, or competence, in a topic. The familiar charge-level display makes an abstract degree of competence immediately understandable.

Alongside What am I supposed to do?, questions like these increasingly appear:

  • Where do I stand?
  • What can I do?
  • What am I missing?
  • What's holding me back?
  • What do I want to achieve?
  • What's my next sensible step?

The Definition of Fun adds What do I want? to What am I supposed to do? and What can I do?, making the learner's own will visible within the learning process.

The Flap thus supports self-regulated learning.910 Its cycle of planning, observing, reflecting and adapting shows up again in how Flap, sprint and retrospective work together.

8. Daily stand-up: help as an available alternative

Autonomy doesn't mean isolation. People also expand their capacity through relationships.

The daily stand-up at the end of a lesson makes difficulties and available help visible:

keep working alone → use materials → ask a classmate → use the AI learning companion → ask the teacher

The goal isn't independence from others, but reducing dependence on any single source of help, and creating alternatives.

9. The team as a union

The team serves its members and shouldn't itself become a coercive structure.

Stronger learners can explain things, others can ask questions, different abilities can work together. At the same time, nobody should be artificially held back from progressing because others work more slowly. Roles therefore stay light, functional and changeable.

The OECD Learning Compass ties student agency to co-agency — acting together with teachers, classmates, parents and community.11 Research on cooperative learning further shows that well-designed cooperation can foster subject learning, mutual support and shared responsibility.12

The team is thus both a resource and a learning space — but stays a tool of learning, not its purpose.

10. Individual support means expanding capacity

Differentiated instruction gets a more precise meaning from this. The decisive question is:

Which next experience expands this particular learner's capacity?

That requires knowing what a student can already do. Teacher observations, results from tasks and edoquiz, and self-assessments together produce an ongoing, correctable picture of competency development.

Diagnostics here doesn't primarily serve classification, but the next action. It's meant formatively.13 Feedback research sums this up in three questions: Where am I going? Where am I now? What's the next step?14

Universal Design for Learning likewise ties learner agency to choice, goal-setting, monitoring one's own progress, varied tools, graduated support, and collaboration.15

At the same time, diagnostics touches the education paradox: data gets collected about people who can't opt out of school. That makes transparency and data minimalism all the more important.

11. Generative tasks instead of rigid differentiation tiers

Classical differentiation often works with categories like easy – medium – hard, or with prepared support and extension tasks. Edo365 can eventually work more finely: from a competency profile, tasks can be generated that target a specific uncertainty.

That changes the underlying logic:

The learner doesn't get assigned to a prepared task. The task can be generated for the learner.

A classic forerunner is mastery learning. Bloom shifted the question from "who can do this?" to the question of under what conditions, and in what time, different learners can reach a given goal.16 Generative AI can substantially expand the practical possibility of making such adjustments at scale.

AI stays a tool throughout. It doesn't decide what education should be, and holds no pedagogical authority. International and German education bodies accordingly emphasize a critical, constructive use that preserves pedagogical responsibility, data protection and independent thinking.171819

12. The AI learning companion should make itself unnecessary

A central design decision follows from Open Autonomy:

A good learning companion increases the learner's ability to get further without it in the future.

Its interventions can therefore be graduated:

Orientation question → small hint → subject-specific hint → sub-problem → solution strategy → a full solution only as a last resort

The decisive long-term optimization target is: does the learner need less support on comparable tasks over time?

Research on intelligent tutoring systems shows that step-based systems in particular can, under certain conditions, achieve substantial learning effects that approach those of individual human tutoring.20

More recent research on generative AI also shows how much design matters. In a field experiment with nearly a thousand students, unrestricted AI access initially improved practice performance, but later led to worse results without AI. A more pedagogically constrained variant substantially reduced that negative effect.21 In a separate randomized study, a learning-science-designed AI tutor achieved higher learning gains in less time than a compared active-learning classroom condition, in a concrete university physics setting.22

The OECD Digital Education Outlook 2026 accordingly distinguishes between better immediate performance through AI and actual learning. Pedagogically designed systems can support learning, while mere cognitive offloading can impair it.2324 Other studies find no significant additional learning gain despite positive perceptions.25

The evidence, then, doesn't speak for AI as such, but for the pedagogical design of how it's used.

AI's own statements stay checkable too. Learners should demand sources, examine them themselves, and cross-check claims. Experienced fact-checkers deliberately leave a source to do this, investigating who's behind it and what other sources say about it.26

Which checking strategies work best for learners in the long run stays open, given how fast AI is developing. Edo365 therefore treats this area explicitly as a hypothesis.

As capacity grows, learners should also increasingly be able to decide for themselves how much help they request. Even a learning companion that rigidly withholds help can become a paternalistic authority.

13. The teacher, too, should partly make themselves unnecessary

The teacher, too, shouldn't remain the permanently necessary authority for every learning step. They provide knowledge, orientation, feedback and support, while at the same time developing learners' capacity to take on more of that themselves.

Their activity thus shifts from permanent steering and routine checking toward observation, diagnosis, advice, subject-specific intervention, and shaping the conditions for learning.

Technology here shouldn't automate relationship, pedagogical judgment or responsibility. It can take over routines, freeing up human time for where human perception and interaction are actually required.

14. Time freed up for what only works together

Individualization, mastery learning, immediate feedback and adaptive support have the potential to make certain individual learning processes more efficient.

Edo365 formulates a further hypothesis from this:

If individualized learning processes become more efficient through adaptive systems, the time freed up can be deliberately used for the educational processes that need human interaction: cooperation, projects, responsibility, creativity and leadership.

The development line runs: mastery learning → intelligent tutoring systems → formative diagnostics → adaptive systems → generative AI.

In Edo365, that becomes: competency profile → individual task → graduated help → diagnosis → adaptation.

The possible efficiency gain isn't itself the educational goal. It's meant to create room for forms of learning where human interaction is itself part of what's being learned.

A current practical example of a radical version of this idea is Alpha School in the US, where core subjects are meant to be covered in roughly two hours of personalized, mastery-based digital learning per day; the remaining time goes toward workshops, projects and practical skills, among other things.27

Alpha School isn't evidence of effectiveness for Edo365, but a current practical experiment. Independent evidence so far isn't sufficient to causally confirm the model's far-reaching success claims.27 What's interesting is the organizational hypothesis being tested there in practice.

The goal of digital efficiency isn't to shorten education. It's meant to win back time from standardizable learning processes, so more time exists for the forms of education that live on human interaction, cooperation, creativity and shared experience.

15. Digitization isn't a goal

A digital tool only makes sense if it expands capacity or the available alternatives. Education research, too, evaluates technology by whether and how it actually improves teaching, feedback, practice and learning.28

That's why edox and edotex exist side by side. Tasks can be worked on digitally, or used on paper as an individually generated PDF.

The question isn't: can this be digitized? It's: which form best supports this particular learning process?

16. Open source and modularity are the pedagogically consistent choice

An education system that wants to foster autonomy should create as little unnecessary dependence as possible on technical systems it can't control.

The Edo components are therefore small, combinable and as independent as possible. edodeck, edosheet, edoquiz and edokanban don't form a monolithic learning platform; edowebbuild can generate static learning environments, and edotex an alternative output path via PDF and paper.

A constructive inspiration for this is modular technical systems, among them the Saab Gripen. What matters is the underlying engineering principle: components hold limited functions and should be changeable or replaceable as independently as possible.

Simon describes complex systems as often built from relatively independent subsystems.29 Parnas transfers this principle onto software modules with hidden internals.30 Baldwin and Clark show how modularity enables decentralized further development.31

Edo365 carries this idea over into pedagogy too. A sprint, for example, doesn't strictly need a particular digital Flap, just a way to do goal → tasks → progress → retrospective. Diagnostics doesn't strictly need a particular app, just observation → competency mapping → assessment → support decision.

That creates pedagogical interfaces: tools can be replaced as long as their function is preserved.

Modules are tools. They're allowed to be replaced. Edo365 itself must not become a fixed idea either.

17. Data serves the learner

Data isn't an end in itself. What gets collected should serve a concrete pedagogical function: observation → diagnosis → support.

The minimally functional applies here too:

As much data as necessary for meaningful support, as little as possible.

A learner's progress belongs to the learner, not to the platform or the institution. That's a normative claim, not an assertion of legal data ownership.

Learners should be able to see what's being recorded about them, question assessments, and take their data with them in principle. Within the legal framework, relevant competency information should stay usable when a learner changes learning group or school.

That reduces technical and institutional exit costs.

A competency model stays a tool here, not a digital image of the person. Every statistical assessment is a correctable hypothesis.

Self-assessment and outside assessment don't need to be merged into one supposedly objective value either. Their difference in particular can provide pedagogically interesting information.

18. Errors get a different function

An error is, first of all, information. It shows a difference between current capacity and a requirement.

The first question therefore isn't: how should the error be graded? It's: what does this tell us about the next sensible learning step?

One inspiration for this comes from the Toyota Production System: deviations and problems are meant to become visible, so their causes can be investigated and processes improved.3233

Edo365 explicitly does not transfer the logic of industrial production onto people. The learner is not a product. What's adopted is only the underlying insight: using a visible deviation as information to improve the next process.

Grading stays necessary given the school's institutional framework. Within that frame, though, a question to the teacher, a classmate or AI isn't a sign of weakness — it's a legitimate attempt to expand one's own capacity.

19. Edo365 is itself a hypothesis

Edo365 must not claim to be a final teaching method.

Sprint, Flap, daily stand-up, teamwork, the competency model, AI and software modules are hypotheses about what might improve learning. They have to prove themselves in the classroom:

Design → teaching → observation → critique → change → renewed attempt

Elements without a recognizable benefit get changed or removed. New ones can be added. Even central elements are allowed to be dropped.

Modularity makes it possible to run this test not only on the whole system, but on its individual parts.

This holds for the question of measuring performance too. Exams and standardized central tests remain in place for now, within the existing institutional framework; Edo365 doesn't unilaterally replace them, but treats them for the time being as a given boundary condition. At the same time, they only offer a single-point, often delayed look back at where a learner stands, and say little about what next step would make sense. In the long run, Edo365 should therefore develop its own, formative assessment model that ties continuous competency diagnostics to the three questions from feedback research: Where am I going? Where am I now? What's the next step?14 But that model, too, remains — like every other element of Edo365 — a testable hypothesis, not a replacement for the legally binding forms of grading.

20. The core of Edo365

Edo365 is an open, modular education offering meant to gradually expand learners' real capacity to dispose over their own learning process.

To that end, it ties together

  • structure, so that self-sufficiency becomes learnable,
  • teamwork, so that other people expand one's own capacity,
  • diagnostics, so that support starts from actual learning progress,
  • individual materials, so that different learning paths become possible,
  • AI, to make individual support scalable,
  • reflection, so that learners increasingly dispose over this process themselves.

Every part can be tested and adopted on its own. Software is infrastructure, not the center.

Open Autonomy asks under what conditions people can dispose over their own lives. Edo365 asks what kind of education enables people to do that.

From this follows the guiding principle:

As much structure as necessary. As much autonomy as possible. With the aim that ever less structure becomes necessary.

Terms

Sprint
A learning segment of roughly one to two weeks with a goal, planning, and a retrospective.
Retrospective
Review at the end of a sprint. It checks what worked, and learners help shape coming sprints as their capacity grows.
Flap
A planning and reflection surface with sprint goal, tasks, battery gauge, problems, Definition of Done, Definition of Fun and reflection.
Battery gauge
A low-threshold self-assessment of one's own ability or competence in a topic.
Definition of Done
Criteria for when a task or sprint goal counts as complete.
Definition of Fun
Makes "what do I want?" visible alongside "what am I supposed to do?" and "what can I do?"
eduScrum
The transfer of Scrum onto teaching, with self-organized learning teams and sprints.
Daily stand-up
A short end-of-lesson exchange about progress, difficulties and available help.
Competency profile
An ongoing, transparent and correctable picture of what a learner can already do.
Education paradox
The tension between the goal of expanding autonomy and an education institution whose participation and rules aren't fully autonomously chosen.
AI learning companion
AI-supported help with graduated levels of assistance, aiming to become less needed over time.
Source work
Checking the origin, evidence and credibility of a claim — including claims made by AI.
Pedagogical interface
The function through which a pedagogical module works together with other parts of the system, without prescribing a specific tool.
edox
A Markdown-based intermediate format with macros for HTML and LaTeX/PDF.
edotex
An output path via LaTeX to PDF and paper.
edodeck
A module for presentations.
edosheet
A module for digital worksheets.
edoquiz
An embeddable quiz module.
edokanban
A Kanban board for planning tasks.
edowebbuild
Generates static learning environments from the modules.
Minimally functional
As much structure as necessary, as little fixation as possible.
Exit costs
The effort or loss involved in leaving a structure, such as a platform or an institution.

References

  1. Falkner, Benno (2026): Open Autonomy. An Open Philosophy of Radical Self-Determination. falkner.xyz/openautonomy. ↩
  2. Ryan, Richard M. / Deci, Edward L. (2000): "Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being." American Psychologist, 55(1), 68–78. ↩
  3. Wood, David / Bruner, Jerome S. / Ross, Gail (1976): "The Role of Tutoring in Problem Solving." Journal of Child Psychology and Psychiatry, 17(2), 89–100. ↩
  4. van de Pol, Janneke / Volman, Monique / Beishuizen, Jos (2010): "Scaffolding in Teacher–Student Interaction: A Decade of Research." Educational Psychology Review, 22(3), 271–296. ↩
  5. Kalyuga, Slava / Ayres, Paul / Chandler, Paul / Sweller, John (2003): "The Expertise Reversal Effect." Educational Psychologist, 38(1), 23–31. ↩
  6. Schwaber, Ken / Sutherland, Jeff (2020): The Scrum Guide. The Definitive Guide to Scrum: The Rules of the Game. Scrum Guides. ↩
  7. Delhij, Arno / van Solingen, Rini (2013): The eduScrum Guide. The Rules of the Game. eduScrum. ↩
  8. Wijnands, Willy / Stolze, Alisa (2019): "Transforming Education with eduScrum." In: Parsons, David / MacCallum, Kathryn (eds.): Agile and Lean Concepts for Teaching and Learning. Singapore: Springer, 95–114. ↩
  9. Zimmerman, Barry J. (2002): "Becoming a Self-Regulated Learner: An Overview." Theory Into Practice, 41(2), 64–70. ↩
  10. Education Endowment Foundation (2025): Metacognition and Self-Regulated Learning. Guidance Report. 2nd ed. London: EEF. ↩
  11. OECD (2019): OECD Learning Compass 2030. A Series of Concept Notes. Paris: OECD. ↩
  12. Johnson, David W. / Johnson, Roger T. (2009): "An Educational Psychology Success Story: Social Interdependence Theory and Cooperative Learning." Educational Researcher, 38(5), 365–379. ↩
  13. Black, Paul / Wiliam, Dylan (1998): "Assessment and Classroom Learning." Assessment in Education: Principles, Policy & Practice, 5(1), 7–74. ↩
  14. Hattie, John / Timperley, Helen (2007): "The Power of Feedback." Review of Educational Research, 77(1), 81–112. ↩
  15. CAST (2024): Universal Design for Learning Guidelines version 3.0. Lynnfield, MA: CAST. ↩
  16. Bloom, Benjamin S. (1968): "Learning for Mastery." Evaluation Comment, 1(2), 1–12. ↩
  17. Kasneci, Enkelejda et al. (2023): "ChatGPT for Good? On Opportunities and Challenges of Large Language Models for Education." Learning and Individual Differences, 103, 102274. ↩
  18. UNESCO (2023): Guidance for Generative AI in Education and Research. Paris: UNESCO. ↩
  19. Kultusministerkonferenz (2024): Handlungsempfehlung für die Bildungsverwaltung zum Umgang mit Künstlicher Intelligenz in schulischen Bildungsprozessen. Resolution of 10 Oct. 2024. ↩
  20. VanLehn, Kurt (2011): "The Relative Effectiveness of Human Tutoring, Intelligent Tutoring Systems, and Other Tutoring Systems." Educational Psychologist, 46(4), 197–221. ↩
  21. Bastani, Hamsa et al. (2025): "Generative AI without Guardrails Can Harm Learning: Evidence from High School Mathematics." Proceedings of the National Academy of Sciences, 122(26), e2422633122. ↩
  22. Kestin, Gregory et al. (2025): "AI Tutoring Outperforms In-Class Active Learning: An RCT Introducing a Novel Research-Based Design in an Authentic Educational Setting." Scientific Reports, 15, 17458. ↩
  23. OECD (2026): OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education. Paris: OECD Publishing. ↩
  24. Education Endowment Foundation (2026): Research Agenda Theme: EdTech. London: EEF. ↩
  25. Slijepcevic, Nedim / Yaylali, Ali (2025): "Leveraging 'Khanmigo' Generative AI-Powered Tool for Personalized Tutoring to Learn Scientific Concepts." Journal of Teaching and Learning, 19(4), 155–178. ↩
  26. Wineburg, Sam / McGrew, Sarah (2019): "Lateral Reading and the Nature of Expertise: Reading Less and Learning More When Evaluating Digital Information." Teachers College Record, 121(11). ↩
  27. Scientific American (2026): "Alpha School's AI Teaching Model Is Expanding. Does It Work?" ↩
  28. Education Endowment Foundation (2019): Using Digital Technology to Improve Learning. Guidance Report. London: EEF. ↩
  29. Simon, Herbert A. (1962): "The Architecture of Complexity." Proceedings of the American Philosophical Society, 106(6), 467–482. ↩
  30. Parnas, David L. (1972): "On the Criteria to Be Used in Decomposing Systems into Modules." Communications of the ACM, 15(12), 1053–1058. ↩
  31. Baldwin, Carliss Y. / Clark, Kim B. (2000): Design Rules, Volume 1: The Power of Modularity. Cambridge, MA: MIT Press. ↩
  32. Liker, Jeffrey K. (2004): The Toyota Way: 14 Management Principles from the World's Greatest Manufacturer. New York: McGraw-Hill. ↩
  33. Ohno, Taiichi (1988): Toyota Production System: Beyond Large-Scale Production. Portland, OR: Productivity Press. ↩

All projects →