Corrections

What was wrong — While fixing genuine defects in the machine-readable project file on 2026-08-30 we also stripped its genesis date, on the assumption that the June 2026 chain reset had changed it. That assumption was wrong. The reset wiped chain data and resynced from the same hardcoded genesis block, so the date never moved. We replaced a correct fact with an instruction to go and look it up elsewhere, which is a small loss of accuracy in the file answer engines rely on most.

What we did — Restored the genesis date, and improved on what was there before: it now carries the exact UTC timestamp and the block hash, so a reader can verify it rather than take our word for it. Checked against malairte-cli getblock on the production node. The earlier correction entry has been amended to drop the incorrect reason rather than silently left standing.

Guard — A test asserts the published genesis hash matches the constant the node reports.

What was wrong — The about page described the project as having a "named team", and the fair-launch page attributed treasury control to the Malairte Project "(Irish-registered entity, named team lead)". No individual is named anywhere on the site, so both claimed a form of accountability we do not currently demonstrate. An unverifiable authority claim is worse than no claim, because a reader can check it and find nothing.

What we did — Removed both claims. The about page now describes the project by what is verifiable — EU jurisdiction, no premine, no ICO, no founder allocation — and the fair-launch page attributes the treasury to the registered entity without asserting a named lead. The press boilerplate's NVIDIA-only mining framing was corrected at the same time for consistency with the GPU correction below.

Guard — A test fails if any public page reasserts a named team or leadership claim while no author entity exists.

What was wrong — Our machine-readable project file told AI assistants that Malairte GPU mining required an NVIDIA card with CUDA and compute capability 7.5 or newer. The standard node build actually mines via OpenCL and is not vendor-locked; the CUDA build is a separate, optional download. The effect was to tell anyone with an AMD or Intel GPU that they could not participate, which is both wrong and contrary to the commodity-hardware design goal.

What we did — Corrected every GPU claim in /llms.txt to describe OpenCL as the standard path with CUDA as an optional NVIDIA build, and removed the unverified compute-capability figure. Verified against malairte-node --help on the production node.

Guard — A test asserts the file describes OpenCL GPU support and does not reassert an NVIDIA-only requirement.

What was wrong — The disclosures page listed our commercial relationships but omitted a material one: MLRT trades on a single venue, CoinDock, which is operated by the same people who maintain Malairte. The mining profitability calculator also defaults its price field to the last trade on that exchange. Neither fact was stated, and the page implied its list of relationships was complete.

What we did — Added a related-party section to the disclosures page naming CoinDock, explaining that a price set on a venue the project operates is not an independent market price, and pointing to where that price is used on the site. The calculator now names its price source in its published method.

Guard — A test asserts the disclosures page names CoinDock as a related party and that the calculator discloses its price source.

What was wrong — Every pillar sitemap listed only its own home page. 404 published items — guides, how-tos, FAQs, glossary terms and articles — appeared in no sitemap at all, because the command that builds the sitemap table had never been run in production and nothing scheduled it.

What we did — Generated the entries (0 to 416) and scheduled the rebuild to run daily.

Guard — A test asserts sitemap:generate is registered in the schedule.

What was wrong — The machine-readable file that AI assistants read as the canonical source of project facts advertised three URLs that returned 404 (/treasury, /press-kit, /technology/mlrthash-spec) and claimed the project lead was publicly identified on /about when no name appears there. (Amended 2026-08-31: this entry originally also cited a stale genesis date. That part was our own error rather than a defect in the file — see the entry above.)

What we did — The file is now generated from live routes and the published pillar list. The dead URLs are gone and the unsupported claim about a named lead was removed.

Guard — A test extracts every URL the file advertises and fails if any of them does not resolve.

What was wrong — A leftover static file shadowed the generated one, so 8 of the 9 pillar sitemaps were never advertised to any crawler. Pillar subdomains served the apex file, pointing them at the wrong sitemap.

What we did — The static file was removed and each subdomain now serves its own robots.txt listing its own sitemap first.

Guard — A test fails if the static file is ever reintroduced.

What was wrong — The "last modified" date published to search engines was hardcoded to 2026-04-18 on the technology, security, wallet, roadmap, whitepaper, quantum-security and fair-launch pages — telling crawlers nothing had changed in over four months.

What we did — Dates now come from a single freshness register, and a scheduled report flags pages whose review has gone stale.

Guard — A test scans every template and fails if a hardcoded date reappears.