Terms of use

These terms describe what 605Buzz publishes and on what footing you read it. They are written to be understood rather than to be imposing, and they describe only things this site actually does.

1. What 605Buzz publishes

605Buzz publishes South Dakota public records and published listings as news articles. seventeen desks publish today, covering things like state layoff (WARN) filings, municipal building permits, road work, city and county meeting agendas, auctions, sports schedules and community listings. Each desk names its source, and each article links back to the record it was built from.

No page on this site claims to hold everything a source contains. Each tracker counts what 605Buzz published, not what the source publishes.

2. The articles are machine-drafted, and no editor reads them first

Articles from the live desks are assembled by fixed software templates that read structured fields out of a public record and fill in prepared sentences. The same record always produces the same article. They publish automatically, under the byline “605Buzz automated newsroom, from public records,” and each one displays “Published automatically from public-record data — not yet reviewed by an editor.”

Between Aug. 9 and Aug. 18, 2026 a language model edited 37 articles for length and order once the templates had written them. That has been withdrawn. The 37 are listed at /edited and are the only ones of that kind; each carries a note, and each keeps the template's own text in its version history — 10 of the 37 do; the rest were edited before first publication and have no earlier version. While the feature ran, this section said software enforced that the model added no fact and removed none. The check was real but weaker than that: it compared tokens rather than claims. No person reviewed either version.

A language model shortens some articles, by choosing whole sentences to delete from what the template wrote. It replies with the numbers of the sentences to cut and nothing else: it cannot write, rephrase, reorder or add text, and the published article is reassembled from the remaining sentences unchanged. Sentences carrying a caveat, an attribution or a source link are excluded from what may be cut. Each one is requested and approved by a person before it publishes, and the change is filed on the corrections ledger as a rewording, because the words changed and the public record did not.

In the machine-readable feeds the same fact appears as humanReviewed: false and editorialTier: tier1-auto-publish. If you are building on this data, read those fields rather than inferring review from the fact that something was published.

3. No warranty of accuracy

605Buzz is provided as-is, with no warranty that any article is accurate, current or complete. Two distinct things can go wrong and we do not promise to prevent either:

• The underlying record can be wrong. If a public agency publishes an incorrect figure, name or date, this site reproduces it. We report what the record says; we do not independently verify it, and no editor checks it before publication.
• The software can be wrong. A parser can misread a column, a source can change its format, and a template can render a field in a misleading place. That is our error, and it is the kind the corrections process below exists for.

Nothing here is legal, financial, employment or safety advice. Do not rely on a 605Buzz article alone for a decision that matters — follow the source link and read the record. For road conditions in particular, check the official source before you travel.

4. Corrections

If something is wrong, tell us and we will fix it. Write to corrections@605buzz.com naming the article and the specific fact you are disputing. You do not need a lawyer to ask, and you do not need to be the subject of the article.

Every correction we issue is recorded with its reason and timestamp, and each article keeps a complete, append-only version chain showing which fields and which body lines changed, version by version. Articles are never silently edited. The corrections index lists the most recent; every correction an article has ever carried is on that article's own page, which is the complete record. Corrections are stored under one of seven reason codes — that is the entire vocabulary the system can record, not a sample:

• factual — a stated fact was wrong and is now right
• clarification — the facts held, the wording misled
• spelling — a name, place, or figure was misspelled
• source_update — the underlying public record itself changed
• rewording — a language model reworded the article for length after the template wrote it; no figure, date, name, source link or caveat moved, and the diff below each version is how you check that rather than believe it
• expanded_reading — 605Buzz began publishing part of a record it had always linked to but had not published before — most often the numbered item lines of a meeting agenda. The public record did not change; what changed is how much of it we print
• retraction — the reporting is withdrawn; the article serves a 410 Gone tombstone

We do not promise a fixed turnaround. 605Buzz is run by one person, and where a matter carries legal weight a correction is reviewed before it is published, so timing depends on the request. The corrections page describes the workflow in full.

5. Retractions

When an article is withdrawn it is not deleted, and this is deliberate. A retracted article stops serving its text and instead returns an HTTP 410 Gone tombstone that points at its correction record, so inbound links, bookmarks and search-engine caches resolve to an explanation rather than to nothing. The retraction is recorded in the public ledger under the retraction reason code, and the article is removed from the feeds so downstream consumers stop republishing it.

The consequence worth stating plainly: the fact that an article existed, and the fact that it was withdrawn, remain permanently visible. We will not quietly erase the record of having published something.

6. Reusing the articles and the feeds

The underlying material is public records, and facts are not ours to own. The feeds exist to be consumed — by people, by readers, and by automated agents. We ask three things, and they are requests about good practice, not a licence you must accept to read the site:

• Link back. Attribute to the article's canonical URL, https://605buzz.com/articles/<slug>, which is also each feed item's id. That is the address a correction or retraction will propagate to.
• Carry the provenance with the text. Every feed item ships a provenance block containing the source URL, the public-record date, a SHA-256 hash of the version you fetched, the editorial tier, the correction count and the humanReviewed flag. If you republish the text, keep the fact that it was machine-drafted and unedited with it. Values the source does not state are null — never fabricated — so treat null as unknown rather than as zero or false.
• Re-poll, and honor retractions. Articles are corrected by appending a new version, so a copy you fetched once can go stale or be withdrawn. A retracted article disappears from the feeds and answers 410; please stop serving it when it does.

The machine surfaces are /feed.json (JSON Feed 1.1), /feed.xml (RSS 2.0) and a per-article JSON document at /articles/<slug>.json. The JSON Feed extension is _buzz at contract version 605buzz-feeds/3.0; that name is frozen and a published version's meaning never changes, so any breaking change ships under a new version string. Version 2.0 (2026-08-09) renamed one provenance value: ai-drafted became template-assembled, because the first name was inaccurate. Version 3.0 is a breaking change with no key, type or cardinality difference at all: a template still assembles every article, but a language model edited 37 of them for length between Aug. 9 and Aug. 18, 2026 — a trial since withdrawn — so template-assembled means something narrower for those rows than it did at 2.0. An edited article carries metadata.derivedFacts.editedBy; its absence means the template's own text is what published, and a consumer that wants byte-deterministic prose reads that key and skips the rest. Retired values stay declared on the wire at _buzz.retiredValues rather than being swapped in silence. robots.txt allows every user agent — please send a descriptive one, and please crawl at a sane rate.

The site's own source code is published separately under the MIT licence. That licence covers the software, not the article text.

7. If an article is about you

These articles are built from records that public bodies publish, and they can name a company, an employer or an individual because the record does. If an article names you and a fact in it is wrong, the corrections route above is the fastest fix and we would rather hear from you than not. If your concern is about the handling of your personal data rather than about accuracy, the privacy notice sets out what we collect and the rights you can exercise.

8. Availability, and where the site is not served

We do not promise uptime. When a data source fails, the affected page returns an error rather than an empty list — an empty page would falsely assert that there is nothing to report, and we would rather show a fault than a wrong answer.

Visitors from the EU, EEA and UK receive an HTTP 451 on article and tracker pages while the EU AI Act Article 50 transparency question is unresolved for this site. This page, /about, /privacy, /why-verified and /corrections are deliberately kept reachable from everywhere, so that a blocked visitor can still find out why and still request a correction.

9. Changes to these terms

If these terms change materially we will say so on this page rather than swapping the text silently, in the same spirit as the dated changelog on the privacy notice.

Contact

Corrections: corrections@605buzz.com
Everything else, including legal notice: hello@605buzz.com
Privacy and data-rights requests: privacy@605buzz.com