It started small.
What if long articles had RSVP? Add audio. There are twelve languages, so add language learning. If a tutorial teaches a skill, add practice. If a comparison depends on conditions, add a calculator. If an official site already solves the problem better, just send people there.
At that point, this is no longer merely an article-site discussion.
The deeper question is:
Why are people searching in the first place?
Usually not because they are desperate to consume an article. They want Wi-Fi fixed, a move completed, a work task finished, a purchase decided, a language used, something interesting discovered, or simply to understand what they should do next.
The real product is therefore not content. It is progress from a current state toward a desired state.
We do not need to rebuild the internet. The internet already exists.
What is missing is a Progress / Resolution Layer on top of it.
1. Reading an article is rarely the final goal
| Where the person is stuck | What they actually want | What the system can provide |
|---|---|---|
| They do not know what they do not know | What to look at first | Orientation / Situation Map |
| Their goal is vague | A clearer desired state | Goal discovery |
| They cannot find the right information | A relevant entry point | Route / Discovery |
| The material is too long | Only what is necessary | Summary, audio, RSVP |
| Terminology is difficult | Explanation from their level | Plain language, examples |
| They do not know if it applies to them | A conditional answer | Branching, local guidance |
| Too many choices | A reasoned decision | Comparison, calculator |
| They do not know the order | The next task | Task Graph / Plan |
| They understand but do not start | A first action | Small tasks, If-Then |
| They get stuck midway | Another route | Troubleshooting / Wizard |
| They stop and return later | Resume without re-planning | Saved progress |
| They do not know whether they are done | Completion criteria | Verification |
| They did it once but cannot repeat it | Independent ability | Practice / Retrieval |
| They cannot explain it to someone else | Transferable explanation | Explain / Share |
| Information changes | Fresh state | Change detection |
| They have no immediate task | Discovery and fun | Explore |
Audio, RSVP and quizzes are not the purpose. They are interfaces chosen according to the purpose.
2. Start before the search query
Search works brilliantly when people can already phrase a query.
But many people are earlier than that.
Kuhlthau's Information Search Process describes early search as a period of uncertainty. More information can temporarily increase confusion until a useful focus is formed.[1]
So the entry point can be as simple as:
"I moved." "I changed jobs." "My computer is acting weird." "I want to organize my money." "I am going abroad." "I have no idea where to start."
The system maps:
Situation → possibly relevant domains → problem framing → goal.
For a move, it may surface address, insurance, tax, vehicles, mail, utilities, telecom, waste rules and employer-related changes, then remove branches that do not apply.
The job is to turn unknown unknowns into known unknowns.
That is the Orientation Layer.
3. One Progress Journey for many products
SITUATION What is happening now?
→ ORIENT What deserves attention?
→ FRAME What is the actual problem?
→ GOAL What state do you want?
→ UNDERSTAND What must you know?
→ DECIDE What should you choose?
→ PLAN What happens in what order?
→ DO Take the actions.
→ VERIFY Did it actually work?
→ LEARN Can you do it next time?
→ EXPLAIN Can you communicate it?
→ MAINTAIN What must stay up to date?
A user can enter at any stage. W3C's Multiple Ways guidance likewise recognizes that people benefit from multiple ways to reach content.[2]
4. Sending people elsewhere is not failure
If the official site is the best place to apply, send the user there. If a specialist has the best explanation, link there. If the purchase happens at a retailer, route there.
The system's value is not owning every destination. It is composing the route.
Possible decisions:
- ROUTE_EXTERNALLY
- SYNTHESIZE
- EXPLAIN
- COMPARE
- TOOL_REQUIRED
- ACTION_GUIDE
- LEARN
- NO_VALUE_ADD
Google Search Central explicitly says external links should not be feared when they make sense and can support trust, especially when citing sources.[3] Its people-first guidance also asks whether users leave with enough information to help achieve their goal, and warns against mass-produced summaries that add little value.[4]
Sometimes the highest-quality article is the one you decide not to manufacture.
5. The creator does not need to personally experience every branch
The dangerous approach is to ask a model to imagine the complete world and then treat every generated route as fact.
Instead, generation creates a candidate map.
Each edge carries a state:
- HYPOTHESIS
- SOURCE_SUPPORTED
- CASE_SUPPORTED
- VERIFIED
- UNKNOWN
Case-Based Reasoning stores past problems, solutions and outcomes and reuses similar cases for new problems.[5]
The founder does not need to experience everything.
The system needs to accumulate experience.
Common routes can be researched and verified first. Long-tail branches expand just in time when users encounter them. Unknown branches go to a Frontier queue, are researched, and successful or failed outcomes return to Case Memory.
It is not a finished world map.
It is a map that grows as people travel.
6. Internally, the product becomes several connected graphs
- Situation Graph: what happened?
- Relevance Map: what may matter?
- Goal Graph: what state is desired?
- Task Graph: what must be done?
- Resolution Graph: under which conditions does which action move the state forward?
- Evidence Graph: why trust this route?
- Case Memory: what actually happened in prior cases?
- Frontier: what is still unknown?
Hierarchical Task Network planning has long represented high-level goals as decompositions into lower-level tasks.[6]
Human errands often look surprisingly similar.
"Travel abroad" decomposes into passport, transport, lodging, connectivity, insurance and entry requirements, each of which decomposes again.
The article becomes one renderer for the graph.
7. Generate only the affordances that help
| Need | Output |
|---|---|
| Deep explanation | Article |
| Fast orientation | Summary |
| Hands occupied | Audio |
| Focused reading | RSVP |
| Position tracking | Karaoke Follow |
| Choosing | Comparison |
| Numeric decision | Calculator |
| Sequential work | Checklist |
| Conditional paths | Wizard |
| Novice imitation | Worked Example |
| Skill building | Practice |
| Recall | Retrieval Quiz |
| Reuse | Template |
| Communication | Explain Card |
| Execution elsewhere | External Route |
| Continue later | Reminder / Progress |
| Changing information | Notification |
An entertainment essay does not need a mandatory quiz. A moving guide benefits from a checklist. Excel benefits from worked examples. Product decisions benefit from calculators. Troubleshooting benefits from decision trees.
Adding every tool to every page turns a helpful product into an aircraft cockpit.
8. UNDERSTAND and LEARN are different
Understanding now is not the same as reproducing later.
Multimedia learning research supports learner-paced segmenting and pre-training as ways to manage essential cognitive processing.[7]
Retrieval-practice research has repeatedly found that practice tests can improve later performance compared with additional exposure alone.[8]
Worked-example research shows that novices can benefit from studying solved examples before or alongside problem solving, while the same support may become unnecessary for experts.[9]
A practical Skill Mode can therefore follow:
Example → Modify → Do → Verify → Retrieve
For action, implementation intentions use concrete If-X-then-Y plans. A 2021 meta-analysis of mental contrasting with implementation intentions found a small-to-medium effect on goal attainment, while also noting possible publication bias.[10]
"Try harder" becomes: "If this screen appears, do this." "When this is complete, do that." "If you hit this state, take this branch."
9. RSVP and audio are common I/O, not the king of the product
A review of speed-reading research finds real limits to claims that people can greatly increase speed while fully preserving comprehension.[11]
So the promise should not be "everyone reads insanely fast."
The promise is choice:
normal text, short summary, detailed version, audio, follow mode, RSVP, diagrams, interactive tools and conversation—all sharing the same progress position.
The Article Player becomes a Universal Consumption Layer across UNDERSTAND, DO and LEARN.
10. The global background: information exists, but usable information is uneven
| Background | Public evidence | Product implication |
|---|---|---|
| Literacy | UNESCO: 739 million adults lacked basic literacy skills in 2024 [12] | Do not make long text the only interface |
| Disability | WHO: about 1.3 billion people, 16% of the world, experience significant disability [13] | Multiple modalities belong in the core |
| Adult skills | OECD: 18% average across 31 participating countries/economies were below the most basic level in all three surveyed domains [14] | Do not assume specialist-reading ability |
| Work change | WEF: 59 of 100 workers would need training by 2030 [15] | Connect learning with actual work output |
| Language learning | Duolingo 2025: English was #1 in 154 countries; Japanese ranked fourth globally [16] | Turn interests and work into learning material |
| Migration | UN: around 304 million international migrants in 2024 [17] | Separate interface language from jurisdiction |
| Connectivity | ITU 2025: 2.2 billion people remained offline [18] | Lightweight and selectively downloadable formats matter |
These are not market-size numbers. Not every person with a disability needs audio, and a web service cannot directly reach people who are offline.
The point is narrower:
Information existing is not the same as information being usable.
11. At least forty application examples
| Domain | Examples |
|---|---|
| Get work done | spreadsheets, quote comparison, handover, support responses |
| Use languages | extensive reading, listening, job phrases, explain-back |
| Finish daily tasks | Wi-Fi diagnosis, moving, decluttering, cooking workflow |
| Buy and choose | total cost, repair vs replace, new vs used, compatibility |
| Protect money and information | household visibility, fees, scam checks, suspicious jobs |
| Live abroad | housing, banking, telecom, public procedures |
| Family and cooperation | chore handover, care notes, requests, device instructions |
| Relearn | foundations, worked examples, checks, certification study |
| Enjoy and discover | science, history, culture, hobby simulators |
| Public and organizational use | multilingual guidance, onboarding, disaster prep, SOPs |
Useful does not have to mean boring.
A good system can help somebody finish taxes and then send somebody else down a rabbit hole about octopus intelligence.
Both are progress.
12. Rollout priority is not moral importance
A practical validation order is:
- Work and digital task completion
- Interest-based language learning
- Everyday resolution guides
- Multilingual corporate manuals and training
- Conditional comparison and calculators
- Audio media and relearning
- Migrant / foreign-resident life guidance
- Scam and digital-safety guidance
- Rules, opportunities and change notifications
- Family, care and shared-life explanation
- Foundational learning outside school
- Access to health, disaster and rights information
Lower rank does not mean lower social value. High-stakes domains deserve slower automation and stronger routing to official or professional help.
13. Globalization should be cut by problem × language × situation
Do not reduce the world to "English market" and "Japanese market."
Separate at least:
- display language
- target learning language
- applicable country / jurisdiction
- current situation
- goal
A person may read Japanese procedures in Vietnamese, learn English through Japanese, or compare a German product in English.
The core international problem is not translation.
It is separating language from jurisdiction and state.
14. High-stakes domains should route safely
For health, law, finance, safety and rights, simplifying information and making an individualized final judgment are different tasks.
The safer role is:
- clarify what to check
- make official information understandable
- prepare questions
- route to the right professional or authority
- avoid pretending uncertainty is certainty
Generation may be cheap. Maintaining accuracy is not.
Rules change. Interfaces change. Prices change. Translation needs QA. Local exceptions remain.
Cheap generation does not imply cheap governance.
15. Move the KPI from pageviews to progress
Longer time on site is not always success.
If a user reaches the correct government page in forty seconds and completes the task, that can be a perfect outcome.
Measure:
- Orientation Rate
- Goal Formation Rate
- Route Start Rate
- Task Completion Rate
- Resolution Rate
- Time to Progress
- Wrong Route Rate
- Recovery Rate
- Learning Transfer
- Trust Return
The metric moves from "did they read?" to "did they move?"
16. The article factory becomes an engine inside a Progress Factory
Today:
Demand → Research → Japanese canonical → 12 languages → Publication → Distribution → Reader signals
Tomorrow:
Situation / Need discovery → Problem / Goal detection → Research → Resolution Graph update → Fill missing canonical knowledge → Generate an article only if needed → Generate useful affordances → Adapt across 12 languages → Observe user progress → Send unknown / failed routes to Frontier → Research again
The article factory is not discarded.
It becomes more important because it is now one production engine inside a larger resolution system.
17. Conclusion: not a search engine, but something that helps you finish
You do not need to build Google, a government office, Amazon, a school and a hospital.
Use the existing world as components.
Fill the gap between:
Current State → Desired State
If people do not know what to ask, show the map. If they do not understand, explain. If they cannot choose, structure the decision. If they cannot start, give the next action. If another site is better, route them. If they get stuck, find another path. When they finish, verify it. If they want mastery, practice.
The ideal user reaction is:
"That was annoying, but I followed this thing and somehow got it done."
Design broadly. Validate narrowly.
Broad in architecture, narrow in validation.
Do not build the whole world first.
Make the roads people actually use more reliable, one route at a time.
References (18)
- Carol Kuhlthau, Information Search Process Theory wp.comminfo.rutgers.edu
- W3C, WCAG 2.2 Understanding Multiple Ways w3.org
- Google Search Central, Link best practices developers.google.com
- Google Search Central, Creating helpful, reliable, people-first content developers.google.com
- Janet Kolodner, Case-Based Reasoning sciencedirect.com
- Georgievski & Aiello, An Overview of Hierarchical Task Network Planning arxiv.org
- Mayer, Principles for Managing Essential Processing in Multimedia Learning cambridge.org
- Testing Improves Performance as Well as Assesses Learning pmc.ncbi.nlm.nih.gov
- Effects of worked examples, example-problem, and problem-example pairs on novices’ learning sciencedirect.com
- A Meta-Analysis of Mental Contrasting With Implementation Intentions pmc.ncbi.nlm.nih.gov
- Rayner et al., So Much to Read, So Little Time journals.sagepub.com
- UNESCO, Literacy facts and figures, 2025 unesco.org
- WHO, Disability and health who.int
- OECD, Survey of Adult Skills 2023 oecd.org
- World Economic Forum, The Future of Jobs Report 2025 weforum.org
- Duolingo, 2025 Language Report blog.duolingo.com
- United Nations, International migration un.org
- ITU, Facts and Figures 2025 itu.int
