TruthX FAQs

TruthX is a quality-first social posting site. It lets users post, reply, repost, and read feeds where AI scoring, source context, and OSINT verification can affect visibility and trust.

TruthX is a social feed that rewards careful, sourced, useful writing instead of raw engagement. For current claims, it adds OSINT-style verification and track-record signals.

Core posting, accounts, feeds, profiles, replies, reposts, AI quality scoring, source URL handling, image uploads, OSINT verification records, OSINT track-record metrics, writing guidance, and this wiki/FAQ are built in the current codebase.

The largest unfinished areas are full Stripe checkout/subscriptions/webhooks, complete subscriber access enforcement, confirmed server cron scheduling, broader notification emails, a full trust-score system, and several product-policy decisions around paid writers and moderation.

It is closer to a working product prototype or controlled beta than a full launch. The main feature paths exist, but payments, cron confirmation, production monitoring, policy decisions, and signed-in browser testing should be completed before broad public traffic.

The product focuses on scoreable writing quality, evidence, source context, and current-claim verification. Popularity signals exist, but the feed is designed to favor quality scores and verification signals over pure engagement.

A good demo should show the feed, a scored post with the score details panel, the small API metadata dollar sign, an OSINT post with verification details, the writing guide, a repost, and the wiki/FAQ.

The wiki lives under public/wiki/ as static HTML, CSS, and JavaScript. That makes it easy to delete or replace later without touching the application router, database, or view system.

Users register through the public registration route, then sign in with the normal login flow. Password reset and email verification support are present in the auth layer.

The visible username is case-sensitive alphanumeric, such as TomJones33. Profile URL lookup can resolve case-insensitively so users do not have to type the exact capitalization in the address bar.

No. The project combined those concepts and uses the term username. The visible identity and profile identity are the username.

Email verification fields and routes exist, and the UI can show verified status. Broader production email verification delivery still needs final SMTP/live-flow confirmation.

Admins can use protected actions such as manual AI rescore and manual OSINT assessment. Admins can also publish OSINT posts.

An OSINT publisher is a user with permission to mark top-level posts as current OSINT claims. This is controlled by the is_breaking_news_publisher flag or admin status.

Writer quality tiers are based on scored posts. New users start at 280 characters. Emerging, Established, and Elite tiers unlock higher character limits and, for higher tiers, essays and paywall eligibility.

Self-service account deletion is not a complete documented feature in the current product surface. Post deletion exists for a user's own posts, but account lifecycle policy needs a separate decision.

A signed-in user submits the composer. The post service validates text, images, writer privileges, paywall settings, and OSINT eligibility before saving the post and queueing scoring.

New users start at 280 characters. Emerging users can reach 1000, Established users 2500, and Elite users 5000, subject to the configured tier thresholds.

Essay mode creates a long-form top-level post. It is available to Established and Elite writers and appears in feeds with a summary and a read link.

Replies are posts linked to a parent post. Thread pages show the parent chain and replies, and reply quality can be scored like other post types.

Pressing repost records the repost, increments the original post's repost count, and creates a feed-visible repost shell. Pressing it again removes the repost record and soft-deletes that repost shell.

Pure reposts are intended to inherit and rank by the original post's score. They should not create a misleading new content-quality score for the person who merely reposted.

No. Quote reposts are disabled. The button and prompt flow were removed, server-side quote creation is rejected, and quote attachments are not rendered in feeds or threads.

Posts are soft-deleted. In thread context, deleted posts can remain as placeholders so conversations do not collapse or lose reply context.

Yes. The media service stores uploads, creates media records, attaches them to posts, and displays image variants in post cards.

Source URLs give the scoring and OSINT systems external evidence to inspect. For normal scoring, the system fetches and summarizes a capped number of URLs. For OSINT posts, initial source URLs can become manual-source evidence.

The primary score uses weighted rubric fields: claim clarity 15%, evidence support 25%, reasoning quality 20%, factual responsibility 15%, discussion value 15%, and tone/conduct 10%. Spam, toxicity, and low confidence can reduce the final number.

The scoring prompt asks for clarity, evidence, reasoning, factual responsibility, discussion value, tone, spam risk, toxicity risk, and confidence. OSINT posts use a separate publish-time quality rubric.

The code supports OpenAI, Anthropic, image scorers, and heuristic fallback scoring. When multiple model rows exist, the latest score from each model is averaged for display; public/home ranking then adjusts that score by post age.

The scoring service has a heuristic fallback. Depending on configuration, regular scoring may require all configured text providers to succeed before saving partial results.

The visible score panel shows the score, score range when multiple model scores exist, model chips, reasons, and a link to the What We Measure explainer.

The small $ details control is a temporary testing/debug aid. It reveals API metadata such as model, tokens, latency, and estimated cost for scoring and verification calls.

Public and home feeds rank by effective score divided by max(days since posted, 1). Reposts use the original post's score for ranking, then apply the repost shell's age.

The quality filter lets viewers choose floors such as Show all, 25+, 50+, 70+, and 85+. The default is currently 50 unless changed by configuration or saved user preference.

Source links can improve evidence context when they are fetchable and relevant. The source verifier captures metadata, summaries, and fetch status for the scoring service.

Admins can trigger manual rescoring from the post card. Normal users do not currently have a public self-service appeal or rescore workflow.

The How To page gives a brief guide to journalistic writing habits such as sourcing, clarity, proportional claims, and avoiding unsupported certainty.

OSINT is the fast current-claim workflow. It is for near-current factual claims that can be checked against trusted sources and manual evidence.

Admins and users with the OSINT publisher flag can create OSINT posts. Normal users post through the regular workflow unless granted that permission.

No. Older research-style items should be posted normally, not in fast OSINT. The service now blocks drafts that look older or non-current from the OSINT path.

New OSINT posts enqueue an immediate self-attestation check, then automated corroboration checks at 1 hour, 6 hours, and 24 hours. The code can process due jobs through the OSINT cron worker.

The application creates persistent database jobs and includes a worker script. A server-level cron schedule still needs to be confirmed on the host for reliable automatic wakeups.

The truth factor combines a capped self-attestation baseline with corroboration scoring: key claim accuracy 40%, source reputation 30%, corroboration breadth 20%, and verification speed 10%. Dispute signals can cap the score.

Assessment labels include likely true, developing, uncertain, and disputed. The post-level status is awaiting verification while under 24 hours with score 0, verified when any pathway raises the score above 0, and unverified if it remains 0 after 24 hours.

Owners and admins can submit manual verification sources for OSINT posts. Those sources are merged with trusted news search results during assessment and can raise the score later, even after the first-day automated window.

The track record summarizes a publisher's resolved OSINT posts, including all-time average, last-50 average, verified count, false-positive/noise rate, and signal quality score.

OSINT post quality scores discipline 30%, source tier 25%, information density 20%, confidence signaling 15%, and coherence 10%, then applies a redundancy modifier.

Significant new intelligence keeps a 1.0 modifier. Partial new detail gets 0.8, essentially redundant information gets 0.6, and pure repost/no-new-intelligence gets 0.4.

Only the eligibility logic is present. A complete paid-content product with checkout, subscriber access, webhook handling, and payout policy is not finished.

Paywall eligibility is limited to Elite writers, top-level posts, and posts at least 1000 characters long. Replies cannot be paywalled.

Yes. The current policy keeps posts under the configured paywall minimum public.

No. Stripe configuration fields exist, but full checkout, saved payment method UI, subscription access, webhooks, refunds, and production payment flows are future work.

Users below the minimum free writer-quality threshold after the configured number of scored posts can be required to add a payment method before posting more AI-scored content. The logic exists, but the payment UI to satisfy it is not finished.

No. Paywall eligibility and content quality scoring are separate. Paid posts should still be subject to quality and policy scoring.

The code does not finalize platform percentage, payout handling, or refund policy. Those are product and business decisions still to be made.

That behavior is not finalized. Public/private subscription activity, profile display, and privacy rules need a product decision.

Admin users see a rescore action on post cards. The route is protected by auth, admin middleware, and CSRF middleware.

Admin users can trigger a manual OSINT assessment. The verification service then records a new assessment and updates the post's verification status fields.

No. Adjacent systems exist, such as content quality scores, spam/toxicity flags, email verification, writer limits, and AI cost caps, but the full trust-score event/risk engine is not built.

Spam and toxicity are part of the scoring model and can reduce scores or verdicts. This is not the same as a full moderation queue or enforcement system.

A full moderation dashboard is not currently built. Admin scoring actions exist, but broader moderation workflow needs separate implementation.

There is partial protection through duplicate-link checks, fetch checks, source reputation, and self-owned-source warnings. CAPTCHA, throttling, and penalty systems are not complete.

The app has an AI cost guard with a monthly spend cap and daily per-user scored-post limit. The visible $ metadata also helps inspect token and cost behavior during testing.

There is no public appeal workflow yet. Admins can rescore, but user-facing appeals or moderation review are future work.

The app is a PHP/MySQL site with a custom router/controller/service structure, server-rendered views, static assets in public/, and CLI cron workers.

Routes are defined in routes/web.php. Public routes, auth routes, post actions, follows, admin actions, and profile routes are all registered there.

Controllers live in app/Controllers, service logic lives in app/Services, models and database query helpers live in app/Models, and templates live in views.

SQL migration files live in database/migrations and are applied with database/migrate.php. The server database is tested with temporary PHP scripts when local DB access is unavailable.

Runtime files are uploaded over SFTP using the project's local SFTP configuration. Credentials should remain private and should not be copied into docs or committed.

OpenAI, Anthropic, Stripe, SMTP, app URL, release stamp, and feature flags are read through configuration and environment values. Secrets should not be committed.

The mailer supports SMTP-backed sending. Password reset uses email paths, and low-score warning emails are wired in the writer quality service when notifications are enabled.

The main app layout reads release version and timestamp from app configuration and shows them in the footer.

Delete public/wiki/ and remove the Wiki/FAQs links from the main navigation templates. No database or router cleanup is required for the static wiki itself.

Local checks include PHP syntax checks and PHPUnit. Remote behavior is smoke-tested with HTTP requests and, when database behavior matters, temporary PHP test scripts on the server.

There is not a local database available for this project. Database-dependent tests need to run against the remote server, and temporary test files should be deleted after success.

Smoke tests should check the home page, OSINT page, What We Measure, How To Write Well, the technical wiki, the FAQ page, static assets, and any changed JavaScript.

Signed-in tests should cover posting, replying, reposting and unreposting, profile display, OSINT source submission, admin rescore, and any paywall eligibility UI.

Bubble U game/nonprofit integration is intentionally not part of this current pass. The site now includes a How To writing section instead of a Bubble U route.

Update public/wiki/faq.html for question content, public/wiki/index.html for technical reference sections, and public/wiki/wiki.js only if search behavior changes.

The static wiki and FAQ content still render and the accordion details elements still work in modern browsers. Search, autocomplete, and prompt buttons require JavaScript.

Confirm server cron, SMTP, payment flows, production API keys, cost caps, database migrations, signed-in UI paths, admin paths, backup strategy, monitoring, and the final moderation/payment policies.