# ON2100 — A few lines. A world ahead.

Research and product blueprint · 9 November 2026

## Recommendation

Build a **future-impact journal**: a fast news outlet, a searchable record of technological progress, and an explicitly pro-AI advocacy platform. The advantage is the connection between a development, the people behind it, its evidence, and the future it could change. Short summaries alone are easy to reproduce; a well-maintained record of claims and outcomes is harder to replace.

The opening audience should be curious builders, founders, researchers, and informed non-specialists. Launch in English with a global perspective, including Africa and other regions often treated as peripheral to the technology story. Start with AI, robotics, and AI-enabled science as the core reporting beats. Keep the broader taxonomy ready without trying to staff every topic immediately.

The proposed brand position is **“Unapologetically pro-AI. Committed to human progress.”** Support rapid discovery, open experimentation, access, and meaningful industry self-regulation. “At any cost” is a weaker public promise: it leaves unclear whose costs count. Human progress should be the criterion for the advocacy, not an assumed consequence of every company announcement.

## What changed in media

Reuters Institute’s 2026 Digital News Report finds that social/video networks are used for online news by 54% of respondents, compared with 51% for publishers’ own sites and apps. Its online survey covers 48 markets. These findings support a distribution strategy that travels beyond the website; they do not establish that any particular format will succeed for ON2100. [1]

My inference: use short videos and recognizable presenters to attract readers, then give them a reason to return directly: personal briefs, followed people, source trails, and forecast updates. An AI interface could later answer questions over the verified archive, but that archive needs to exist first.

| Reference | Useful pattern | ON2100 adaptation |
|---|---|---|
| Axios / Smart Brevity [2] | Important information first; clear, short reading structure | A headline and three lines, followed by optional depth |
| Techmeme [3] | Software discovery with human editorial decisions | Group duplicate reports into one event and show multiple sources |
| TLDR [4] | Repeatable daily summaries | A finite daily brief that readers can finish |
| The Verge’s 2022 redesign [5] | Short posts and links alongside deeper work | One flexible feed that can carry text, visuals, and context |
| Metaculus [6] | Explicit, resolvable forecast questions | Record the claimant, target, criteria, revisions, and eventual outcome |

These are patterns to learn from, not evidence that ON2100 has no competitors. The initial opportunity is a distinctive combination; validate it with readers before claiming a new market category.

## The information unit

Target 45–75 words for the default brief, excluding metadata. This is an editorial target, not a rigid limit when nuance is essential.

1. **The signal:** what happened, or exactly who made a future claim.
2. **Why it matters:** the concrete implication for people, science, or industry.
3. **Keep in view:** a constraint, uncertainty, comparison, or next milestone.

Every item also has a short headline, story type, publication and update timestamps, impact horizon, topics, people/organization links, and source links. A company announcement proves that the company made the claim; it does not independently prove the claimed performance.

Depth works in layers: headline scan → three-line brief → context and milestones → primary sources and full reporting. Readers should not have to leave the page simply to understand a term or see why a horizon was assigned.

Use distinct labels: **News, Analysis, Forecast, Scenario, Opinion, Advocacy, Sponsored.** Corrections, updates, and disputed claims are separate status fields. A story type is not a numerical confidence score. “News” should not visually imply that every assertion in an announcement has been independently verified.

## Two clocks: publication and impact

Do not use a future year as the publication date. The impact horizon is a lens for the main development or attributed target.

| Primary horizon | Initial coverage window | Typical editorial use |
|---|---|---|
| 2026 | 2026 | Current developments and present adoption |
| 2027 | 2027 | Near-term targets and deployment plans |
| 2028 | 2028 | Pilots, expansion, and approaching milestones |
| 2030 | 2029–2030 | The end-of-decade transition |
| 2040 | 2031–2040 | Longer adoption and infrastructure cycles |
| 2050 | 2041–2050 | Large changes in energy, health, and society |
| 2070 | 2051–2070 | Long-range scenarios and attributed forecasts |
| 2100 | 2071–2100 | Century-scale questions and possibilities |

Assignment rules:

- Use an explicit source target first, storing the exact date separately from the bucket. A 2029 target belongs in 2030 while remaining visibly “source target: 2029.”
- Preserve date ranges. A 2035–2055 range should remain visible with related horizons; choose a primary browsing bucket with a written editorial rationale. Do not replace the range with its midpoint as if the source gave a point estimate.
- When timing is unknown, use an **Undated** editorial holding status. If an editor supplies a horizon lens, label it as an editorial lens, with its rationale. Unknown timing is preferable to false precision.
- Dates after 2100 need a **Beyond 2100** tag. Do not silently redefine 2100 as “all later years.”
- Store secondary horizons without duplicating the event across the main feed.
- Do not rewrite the original story or forecast when a target slips. Add a dated revision and connect it to the original. Maintain a current-view index and an immutable historical record.
- Preserve old year pages. Add new near-term year pages as the calendar advances and version the displayed windows. A 2026 story must not automatically become a 2027 story in January.

Near-term news should carry most of the publishing volume. Distant horizons earn space through strong attribution, meaningful scenarios, and periodic synthesis. Do not fill 2100 with daily speculation just to keep every column busy.

## Navigation from first principles

A reader usually asks one of five questions: What changed? When could it matter? What field interests me? Who is involved? What supports the claim?

The interface maps directly to those questions:

- **The Feed:** important developments with clear labels and event grouping.
- **Impact timeline:** persistent, simple year choices; horizontal scrolling on narrow screens.
- **Topics:** a secondary filter, not a competing mega-menu.
- **People & organizations:** one directory with role and organization-type filters.
- **Search:** one entry point returning grouped stories, people, organizations, topics, and forecast records.
- **Watch:** short captioned explainers, with a read-first equivalent.
- **Our mission:** a clearly identified advocacy destination.
- **My brief / Saved:** reader utility and return visits.

Keep an “All” view and visible reset control. Filters combine across axes, remain in the URL, and show the result count. At scale, use cursor pagination or a deliberate “Load more” interaction. The completed daily brief gets a caught-up state.

Initial topics: AI & intelligence; Robotics; Biotech & health; Energy & climate; Space; Quantum & computing; Work & society; Politics & Power; Data & Numbers; Technocracy. Add finance, materials, and mobility as subtopics initially, then promote only when there is enough distinct reporting. Region and language are independent facets. A story can have one principal topic and several secondary tags.

## People and organizations

Use entities with stable IDs and aliases. Never use a name string as the primary key.

| People roles — multiple allowed | Organization types |
|---|---|
| Founders, executives, investors | Companies and startups |
| Scientists, researchers, engineers | Research labs, universities, institutes |
| Policymakers and public officials | Public agencies and governments |
| Creators, writers, and public thinkers | Nonprofits, standards groups, professional communities |

Allow overlap: a scientist can also be a founder. Roles and affiliations need effective dates and sources. Organization relationships should distinguish parent, subsidiary, partnership, funder, and employer rather than treating every association as endorsement.

A profile contains a concise sourced biography, public roles, related stories, areas of work, a media gallery with rights/provenance, and a prediction record when there are actual attributable claims. Cover professional milestones, arguments, research, and publicly relevant events. The directory should not drift into unsourced private-life gossip.

Forecast records need: claimant; exact quotation or carefully attributed paraphrase; original source; claim date; target date or range; resolution criteria; current status; revisions; outcome evidence; reviewer and review date. Use statuses such as open, achieved, missed, withdrawn, changed, ambiguous, and unverifiable. A revised deadline does not erase a missed original one. Do not invent accuracy percentages from a tiny, selectively collected sample.

## Images, video, and reactions

Each story can use one lead visual. Label **photograph, licensed footage, illustration, AI-generated image, or chart**; store credit, rights, caption, and alt text. A synthetic scene must not masquerade as evidence of an event. Where available, preserve content credentials; C2PA documents a standard for recording media provenance. Provenance alone is not a guarantee that a claim is true. [9]

The distribution format should be a 20–60 second captioned vertical explainer, alongside the same three-line text and source link. The website can also host longer interviews and analysis. Keep the first frame informative, avoid sound autoplay, and load video after an explicit tap. Do not promise Wi-Fi-only playback using unreliable connection detection. Provide a reliable manual data-saver choice instead.

Proposed reaction vocabulary:

| Reaction | Meaning | Optional sound |
|---|---|---|
| 🔥 Fire | Exciting progress | Short rising spark tones |
| 🧊 Ice | Overhyped or unconvincing | High crystalline tones |
| 💸 Cash | Economic significance | Brief cash-register-style chime |
| 👎 Boo | Disagreement | Low descending cue |
| 🤯 Wow | Surprising or mind-expanding | Rising sparkle |
| 🚀 Sooner | Could arrive earlier | Ascending cue |
| 🐢 Later | Could take longer | Two slower low tones |

Default sound off. One changeable reaction per authenticated reader per story in production. Count unique accounts, rate-limit writes, detect bursts, and review suspicious patterns. The prototype stores one local reaction, not a public count. Reactions should not change evidence labels or resolve forecasts. “Sooner / Later” is reader sentiment, not a statistically valid probability estimate.

## Habit and distribution

Build repeat use around utility: a finite morning brief; “since your last visit” updates; followed topics and people; weekly **Horizon Shift**; monthly forecast reviews; and an annual **State of the Future**. Let the reader control alerts and sound.

Use a consistent presenter or small host team for video, with human judgment visible. Distribute extracts through short-video platforms and a newsletter, with source trails and deeper context on ON2100. Build regionally diverse contributor relationships and invite scientists to explain primary work. Treat distribution channels as replaceable; the archive and direct reader relationship are the durable assets.

Test a small number of recurring formats before increasing output. Suggested initial editorial cadence, contingent on staffing: 5–8 checked briefs on weekdays, three short explainers each week, and one deeper weekly analysis. These are proposed operating targets, not forecasts of audience growth.

## Advocacy and editorial credibility

The advocacy platform can support industry-led self-regulation through public evaluations, incident disclosure, independent review, open standards, and accountable commitments. This is a proposed position, not a claim that self-regulation alone resolves every policy problem.

Keep bylines, funding disclosures, sponsored labels, sources, and a visible correction history. News judgment remains independent of campaigns and sponsors. Do not treat criticism as anti-technology by default; accurate criticism can expose what needs to improve for adoption to succeed.

For launch, publish: mission; editorial standards; AI-use policy; funding and sponsorship disclosures; corrections process; contact; and reader data practices. AI can help discover, group, translate, and draft. An editor verifies facts, attribution, dates, permissions, and labels before publication.

## Scalable content system

Model the editorial data before choosing a large stack. A relational database and a search index are sufficient for an initial production system; a graph database is not necessary just because people and organizations are connected.

| Record | Essential fields / relationships |
|---|---|
| Event | Stable ID, normalized event date, headline, related stories |
| Story | Event ID, three lines, type, publication/update time, primary horizon, date basis, topics, regions, status, authors |
| Source | URL, publisher, publication date, retrieval date, source type, rights, supported claim IDs |
| Entity | Stable ID, kind, names/aliases, sourced roles and dated affiliations |
| Claim | Story/entity/source links, exact assertion, target range, criteria, status |
| Revision / outcome | Original claim ID, changed fields, reason, evidence, reviewer, timestamp |
| Media | Rights, credit, provenance, captions, transcript, alt text, derivatives |
| Reader preferences | Follows, saves, notification settings, consent status |
| Reaction | Reader ID, story ID, reaction type, timestamp; unique reader/story pair |

An ingestion queue accepts permitted feeds, official announcements, research sources, and editorial submissions. Normalize URLs; deduplicate first by URL and event candidates; propose tags; then send an evidence-linked draft to a human review queue. Do not re-host complete publisher articles without the relevant rights.

Publish approved records to the website, search index, and distribution queue. Preserve provenance at the claim level so a summary can be audited. Changes to sources, names, and targets should trigger review rather than silent regeneration.

For production search, configure facets for horizon, date, story type, topic, entity ID, region, language, and media format. Meilisearch’s official documentation describes filterable attributes and combined filters. [7] Select a service after benchmarking a representative corpus, editorial update volume, typo handling, and operating cost. The current prototype is a local in-memory search, not a production search deployment.

Ranking should favor exact entity matches, query relevance, editorial importance, originality, source quality, and appropriate freshness. Keep reader-reaction sorting separate. Provide distinct Latest and Recommended modes and explain personalization. Add semantic retrieval only when it improves tested retrieval quality; generated answers must cite the archive and indicate when it has no answer.

Use durable article URLs, canonical URLs, sitemaps, RSS, and structured metadata on the public launch. Schema.org’s NewsArticle supports describing news and its context. [8] Do not mark imagined scenarios as verified reporting. Cache public reading pages; keep reader-specific data separate. Protect publishing roles and reaction writes on the server, not only in the interface.

## Revenue and measures

Keep the core brief and source access free initially. Potential revenue experiments include clearly labeled sponsorship, paid research/forecast tools, team subscriptions, and editorially independent events. Avoid paid prominence in scientific credibility rankings. Determine pricing after reader interviews and observed retention; this concept does not establish a viable price point or revenue forecast.

Measure returning readers, brief completion, source opens, saves, successful searches, forecast follow-up completion, correction turnaround, newsletter retention, and paid conversion if introduced. Measure article comprehension and trust through reader feedback. Do not make time spent scrolling the primary success measure.

## Proposed 90-day sequence

| Period | Work | Exit condition |
|---|---|---|
| Days 1–30 | Identity, editorial standards, publishing workflow, verified story library, newsletter signup/delivery, mobile reading, sources and corrections | A complete daily edition can be produced, checked, delivered, and corrected |
| Days 31–60 | Accounts, synchronized follows/saves, full-text facets, entity profiles, captioned video production, forecast records | Readers can follow an entity or horizon and reliably receive relevant updates |
| Days 61–90 | Event clustering, search quality review, newsletter/video experiments, contributor onboarding, first Horizon Shift report | Retention and quality evidence supports the next investment decision |

Staffing assumption: an accountable editor plus research/reporting and multimedia capacity. Adjust scope to actual team availability. The exact schedule is a planning proposal, not a delivery commitment.

## What this working prototype includes

- Responsive feed and eight impact horizons.
- Fourteen reference/scenario stories with type labels and context.
- A searchable directory of seven public people/organizations.
- Local topic, type, horizon, and keyword filters; story links.
- Source links, a sample original forecast record, and related entities.
- Browser-local bookmarks, follows, personal brief, and changeable reactions.
- Optional synthesized reaction sounds; no sound autoplay.
- Original AI-generated cover illustration and a 24-second captioned concept film.
- Advocacy page, research page, and this downloadable blueprint.

It does **not** claim to be a live newsroom. Automated ingestion, public publishing/CMS, newsletter delivery, multi-user counts, user accounts, synchronized data, real-time alerts, and on2100.org domain routing are not connected. The illustration and concept film are not documentary evidence. Production requires a separately implemented and tested editorial/data system.

## Sources

Sources reviewed 9 November 2026. The design recommendations above are ON2100 proposals and inferences, not promises made by these sources.

1. Reuters Institute, Digital News Report 2026 executive summary: https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/dnr-executive-summary
2. Smart Brevity, method description: https://www.smartbrevity.com/what-is-smart-brevity
3. Techmeme, About: https://techmeme.com/about
4. TLDR: https://tldr.tech/
5. The Verge, “Welcome to the new Verge,” 13 September 2022: https://www.theverge.com/2022/9/13/23349876/the-verge-website-redesign-new-newsfeed-blogs-logo
6. Metaculus, FAQ: https://www.metaculus.com/faq/
7. Meilisearch, filtering documentation: https://www.meilisearch.com/docs/capabilities/filtering_sorting_faceting/getting_started
8. Schema.org, NewsArticle: https://schema.org/NewsArticle
9. C2PA, content provenance: https://c2pa.org/

Reference-story and profile sources:

- Google DeepMind, Gemini Robotics announcement, 12 March 2025: https://deepmind.google/blog/gemini-robotics-brings-ai-into-the-physical-world/
- Google DeepMind, AlphaFold: https://deepmind.google/science/alphafold/
- IBM, Starling roadmap, 10 June 2025: https://www.ibm.com/quantum/blog/large-scale-ftqc
- NASA, Artemis: https://www.nasa.gov/humans-in-space/artemis/
- NVIDIA, Jensen Huang biography: https://www.nvidia.com/en-us/about-nvidia/board-of-directors/jensen-huang/
- Nobel Prize, Demis Hassabis facts: https://www.nobelprize.org/prizes/chemistry/2024/hassabis/facts/
- Stanford, Fei-Fei Li profile: https://profiles.stanford.edu/fei-fei-li


## Expanded editorial ambition — 9 November 2026

The ambition is to become a leading global reference for AI, emerging technology, political change, and the institutions shaping the future. A forceful visual identity attracts attention; an original, dependable reporting record earns authority. This is an ambition, not a claim that ON2100 already has that standing.

The revised visual system uses near-black, acid yellow, hot pink, cyan, violet, and orange. Colored horizons and beat entrances support scanning. Bold typography gives the outlet a contemporary, passionate voice. Distinct labels and legible body text remain central.

| Area | What it covers | Evidence standard |
|---|---|---|
| Politics & Power | Elections, geopolitical competition, technology policy, national capability, and political forecasts | Named sources; distinguish reported events, polling, model probabilities, and market-implied prices |
| Data & Numbers | AI adoption, investment, compute, energy, research, and governance indicators | Show period, geography, unit, denominator, source, methodology, revisions, and uncertainty |
| Technocracy | Expert-led government, AI public services, algorithmic decisions, digital states, and the relationship between expertise and democracy | Examine effectiveness, accountability, contestability, citizen participation, and distribution of benefits |

Political forecasting should be an auditable record. Store the question, jurisdiction, claimant/model, original probability when actually available, timestamp, target date, resolution criteria, update history, and official outcome source. Never manufacture numerical probabilities to decorate a story. A poll measures respondents; a forecast estimates an outcome; a prediction market reflects trading conditions. Label them separately. Consider calibration and Brier scores only when enough comparable resolved probabilistic forecasts exist, and disclose the sample.

Technocracy is a reporting area, not an automatic endorsement of transferring political power to technical experts. ON2100 can advocate pro-AI innovation while investigating the quality and legitimacy of algorithmic governance.

To earn leading-media status, prioritize original interviews, regional reporting, a carefully maintained data desk, public corrections, searchable forecast histories, and attributable primary-source coverage. Start with quality in a few beats, then expand editorial capacity. Track return readership, source engagement, citations by other institutions, successful searches, and correction performance. Do not substitute reaction counts for credibility.

New reference sources:

- OECD, Governing with Artificial Intelligence (2025): https://www.oecd.org/en/publications/governing-with-artificial-intelligence_795de142-en.html
- World Bank, Worldwide Governance Indicators: https://www.worldbank.org/en/publication/worldwide-governance-indicators
- World Bank, WGI methodology and documentation: https://www.worldbank.org/en/publication/worldwide-governance-indicators/documentation
- Metaculus, forecast resolution guidance: https://www.metaculus.com/faq/

The prototype includes reference analyses and an explicitly labeled political future scenario. It does not publish live election forecasts or claim a calibrated political prediction model.