1
CronkiteNews You Can Trust
A set of AI agents turns scattered posts, sources, and claims into one clear view of an event: what happened, who is saying what, what each side believes, and what evidence supports it.
One news event
2
The trust problem
We do not know who to trust anymore.
A single story now arrives as posts, clips, screenshots, creator takes, official statements, and instant rebuttals. Everything looks equally urgent. Very little arrives with enough context.
The problem is not access to information. The problem is knowing what to trust.
Every fragment competes before the story is understood.
U.S. trust in mass media.
GallupU.S. concern about what is real and fake online news.
Reuters InstituteThe product opportunity is trust built around the story, not another headline stream.
3
How news is consumed now
The front page is now a crowd of people.
News reaches us through accounts, creators, journalists, experts, institutions, eyewitnesses, and anonymous aggregators. Some are useful. Some are early. Some are wrong. Some are pushing a viewpoint.
If people are the new front page, the product has to understand their history.
61%
third-party platforms
29%
publisher apps/sites
4
Core insight
A post is not the story.
Every event has a past. Every source has a history. Every viewpoint comes from somewhere.
Cronkite starts from the belief that history is the clue to trust.
5
What Cronkite builds
The full idea in one look.
Agents research the topic, search the web for historical context, inspect source histories, validate claims, group viewpoints, and present the full shape of the story.
The output is not a verdict. It is an inspectable map of the story.
One evolving topic
Built through web research and source links.
Grouped by claims, emphasis, and disputed facts.
Previous statements, evidence habits, and topic history.
Linked back to sources where possible.
6
Consumer experience
Before you react, inspect the story.
Cronkite turns a chaotic topic into one readable topic page: what happened, how we got here, what each side believes, who is pushing each perspective, and what evidence exists.
The user sees an evolving timeline with viewpoints and trust profiles attached to the voices inside each perspective.
Agents add new events, source links, documents, and prior coverage as the topic develops.
What each side believes happened, what they emphasize, and where the dispute sits.
Previous statements, evidence habits, corrections, proximity to the event, and topic-specific history.
7
Why now
Agents do the work people cannot.
Cronkite is built around agents. Not one summary bot. A research system that watches X for live signals, builds source material from the web, validates claims, and appends new context to a living topic timeline.
The timeline is not rebuilt from scratch every time. It becomes the memory layer that agents keep updating as the topic evolves.
Collect live signals from X, where claims, clips, and reactions first land.
Find source material, documents, prior coverage, and historical events.
Search relevant X profiles for previous statements on the topic.
Check claims against source material and mark what is supported, contradicted, or unresolved.
Group competing perspectives and the voices pushing each one.
Append new events to the living topic timeline and keep the UI current.
8
First build
MVP: one evolving topic, one living timeline.
The MVP follows one topic as it develops and keeps adding context over time instead of treating news as disconnected updates.
If Cronkite can make one evolving topic understandable over time, it has the foundation for a consumer news trust product.
Posts, clips, reactions, official statements, and arguments arriving as disconnected updates.
New events appended over time, with major viewpoints, source background, trust profiles, and validated claims.
9
Business model
Make trust a weekly subscription.
Cronkite starts as a consumer subscription. A new subscriber chooses topics, sources, and delivery cadence; Cronkite maintains the living context map and delivers briefings when the story changes.
Free public topic pages acquire users. Paid personalization and alerts retain them. Reusable topic memory is the path to margin.
Premium unlocks watched topics, deeper source profiles, archive search, and alerts. One living topic memory can serve many subscribers who follow the same story.
Major stories become public context maps people can search, cite, and share.
Subscribers choose topics, voices, regions, and delivery rhythm.
Daily digest, instant update, or watchlist alert when claims change.
Measure
API cost / living topic
AI cost / subscriber
retention by watched topic
10
Shipyard ask
Credits, Credits, Credits.
Cronkite needs two credit pools: xAI credits for the X Search and Grok calls used by X research agents, plus ChatGPT, Claude, and other AI credits for the Hermes agents themselves.
The ask is API credits to search X, build and run the Hermes agents, and measure cost per living topic.
| Credit bucket | What it funds | Current estimate |
|---|---|---|
| xAI credits | X research agentsUse X Search and Grok to inspect profile history, pull live signals, and find previous statements from relevant voices. | $5 / 1,000 X Search callsUp to 20 handles per call. Local blended estimate with Grok synthesis is about $0.0225 per call. |
| X source graph | Profile monitoringFollow useful X accounts as topics evolve: reporters, experts, institutions, outlets, and primary sources. | 100 profiles: $18-$72 / month500 profiles: $90-$360. 1,000 profiles: $180-$720. Token costs extra. |
| ChatGPT / Claude / other AI credits | Hermes sensemaking agentsBuild the timeline, group viewpoints, validate claims, create trust profiles, summarize evidence, and prepare UI-ready context. | TBD per agent runPrototype must measure cost by Hermes agent type: timeline, validation, viewpoint, profile, and presentation. |
| Raw X API fallback | Deterministic capture if neededHydrated post objects and exact capture when X Search is not enough. | $3k-$7.5k / monthFor 1,000 accounts at 20-50 posts per account per day. This is why xAI X Search is the first path. |