Public evidence · Reviewed as of Nov 1, 2026

The work, including what isn't fixed yet.

A report is not a correction. An agency acknowledgment is not a measurable benefit. These are distinct stages, and we show the difference.

3independently observed corrections
13tracked outcome cases
0accepted outside node submissions

Independently verified changes · S.T.A.R. case files

What we identified, did, and verified.

Each result separates our reporting from what the responsible institution changed. No imaginary beneficiaries or inferred volunteer hours.

WIN-001 · VERIFIED

Texas Medicaid Buy-In guidance

Situation: An impossible income range appeared in a Texas Medicaid Buy-In premium chart.

Task: Identify the correct published range and alert the responsible handbook unit.

Action: Humanity Loop reported the documented contradiction September 25.

Result: Texas HHS corrected the public chart by September 30. Agency action and updated text were independently observed.

Reach: Relevant to people consulting that chart; verified individual readership and benefit not measured.

Official source and evidence ledger →
WIN-002 · VERIFIED

VA Spanish-index rate links

Situation: Four links on a Spanish-language VA rate index led to outdated or unrelated benefit tables.

Task: Document the affected paths and current official rate destinations.

Action: We reported the link discrepancies September 25 and provided an evidence update October 20.

Result: On October 27, all four affected links were independently verified pointing to current 2026 VA rate tables. The VA made the changes; our exclusive causal role is unknown.

Reach: Four corrected navigation paths, not four people. Audience size and actual exposure are unmeasured; linked destinations remain in English.

Before / report / after receipts →
WIN-003 · VERIFIED

Medicare prescription-plan helper

Situation: A Medicare.gov helper showed outdated Extra Help income figures.

Task: Identify the 2026 thresholds and reach the responsible CMS team.

Action: We reported the discrepancy with official comparison sources September 19.

Result: CMS confirmed a September 30 update and explicitly acknowledged our reporting. The helper now displays $23,940 for an individual and $32,460 for a married couple.

Reach: Two corrected thresholds; the number of actual users helped is not measured. The CMS response is attribution evidence, not beneficiary analytics.

Check Medicare's current helper →

These are three verified corrections, with distinct evidence and attribution strengths. Modest counts such as four corrected links are defensible; we will not turn eligible population estimates into invented counts of people helped.

Real engagement, no claimed resolution

Feedback received. Work still open.

ACKNOWLEDGED · OPEN

Texas WIC fair-hearing wording

An HHSC Appeals Division deputy director agreed the handout's lawyer-assignment language was misleading and contacted publications staff. The public PDF was still unresolved at the last documented check.

SOURCE ACCESS · UNVERIFIED

Texas WIC Spanish limits

Current application pages show correct 2026–27 figures. An older page appeared in search results with outdated figures, but direct access returned HTTP 403. That does not establish what an ordinary visitor sees.

AGENCY FEEDBACK · SOURCE-LINKED STATUS

Alameda County Spanish CalFresh information

Acknowledged · correction pending. Alameda County Social Services replied October 30 that it would forward our outdated Spanish CalFresh figures report to the appropriate team. The revised page has not been independently confirmed.

Read the public case receipt and latest replies →

Last independently documented status: October 30, 2026; live refresh not confirmed. New case notes can appear here without a website build. Status changes still require independently documented evidence.

PUBLIC RECORDS · SOURCE-LINKED FEED

What's changed since the last website update?

The public register is read from Humanity Loop's reviewed GitHub records when you open this page. A new verified win or tracker status can appear here without editing the website's code. This is a record of documentation, not a real-time agency monitor.

The latest public win ledger could not be refreshed in this visit. The verified S.T.A.R. case examples above remain available.

The public tracker could not be refreshed. Agency acknowledgments and changes must still be independently corroborated.

Only previously published public source files are read. No visitor account, private beneficiary data or GitHub login is needed. Updates to public GitHub records may take time to appear due to caching.

See the full case and evidence register →

This detailed evidence record is stored in GitHub for auditability; you do not need GitHub to volunteer. Start with an email or small task instead.

Impact measurement

How many people did this actually help?

That is the question we want to answer for every intervention. But a potential audience is not a count of people reached, and an agency correction is not proof of better real-world outcomes. Each win will include a documented S.T.A.R. account (Situation, Task, Action, Result) plus a separate reach and benefit assessment.

POTENTIAL AUDIENCE

Who could benefit?

Dated population estimates with explicit denominator, geographic and program relevance, uncertainty and source links. Never labeled as people helped.

OBSERVED EXPOSURE

Who actually encountered it?

Page views, publication downloads or confirmed tool usage when responsible institutions can provide valid figures. No assumption all visitors changed behavior.

VERIFIED BENEFIT

What changed for someone?

Evidence of improved comprehension, access or outcomes, with permissions and privacy protected. If unavailable, the count is not yet measured.

There are three verified public corrections, but no defensible aggregate person-level beneficiary count. Medicare and Veterans Affairs are different populations and are never interchangeable.

Measurement rules and limitations →

What we will not claim

We cannot infer that Humanity Loop caused an agency change merely because our report came before it. We cannot infer how many people benefited without evidence. Internally completed AI tasks do not equal independent adoption or human volunteer hours.

Help with the next useful check →