Skip to content
type1.science

What changed

Freshness & changelog.

Provenance you can see. Material changes from each monthly refresh are grouped at the top. Below that is every record, most-recently-reviewed first. Review dates show editorial activity; they do not certify that every claim has been verified.

460 records · 0 still seed

New means a record was first added on that date. Updated means an existing record was reviewed on that date.

How we review and score the evidence →

September 2026

September refreshes began on August 27 and continued on September 16–19. The latest release incorporates completed Muse review batches and Astra release checks, corrects three further passages, and simplifies navigation. Full factual verification and worldwide discovery remain unfinished.

Baseline review cutoff 27 Aug 2026 · 194 material changes

19 Sep 2026 · Muse review continuation and simpler navigation

Refresh by GPT-Astra (xhigh)

Contributors: Muse Spark via Muse Code: continuing source comparisons, independent review executions, narrower wording and documented limitations. Runtime details remain recorded as supplied in the review ledgers. · GPT-Astra (xhigh): release integration, targeted primary-source checks and publication. · GPT-Astra (medium): separate pairing audit and medical source spot-check; configured launch settings.

Incorporated the completed review snapshot, including revisions across 257 item and trial records. Checked that the reviewed text matches its evidence records and performed a targeted medical source spot-check. Corrected three additional passages about HDV-insulin lispro, degludec exercise evidence and Control-IQ+ trial comparisons. Fixed review-import date handling and made unknown batch selections fail explicitly. Removed Updates, Review status and Country access as separate public pages; old links redirect to the changelog or the evidence guide. Changelog remains the public change history, with methodological limitations explained on Methodology and source details in individual records.

  • This publishes completed work so far; it is not a claim that every factual assertion or programme worldwide has been verified.
  • Matching text and paired review records establish traceability, not that every source judgment is correct. Some statements remain scoped, pending or blocked.
  • The Astra medical check was a targeted spot-check. Separate AI reviews do not replace independent clinical review.
  • Actual review dates are preserved, including work after the original September 17 audit cutoff. The internal register's full completion gate remains unmet.

17 Sep 2026 · Community tools, continuing source verification, and Muse continuation review

Refresh by GPT-Astra (xhigh)

Contributors: Muse Spark via Muse Code: independent review of the Omnipod 6 and SmartGuide batch; primary review of ATG and abatacept, with a separately executed Muse second pass. Exact model version and runtime effort were not independently confirmed. · GPT-Astra (xhigh): release review, targeted source rechecks, corrections and deployment. · GPT-Astra (high): independent medical spot-check and two metadata checks. · GPT-Astra (medium): provenance audit and verification-register history fix. · Parallel AI source reviewers: exact model and effort were not exposed by the worker runtime for this verification continuation.

Added five reader resources — practical side-by-side comparisons, a dated reader digest, country-specific access, a plain-language evidence guide and public review status. Initial coverage includes eight practical device profiles with 45 sourced facts, seven country–item access profiles with 14 sourced fields, and seven digest highlights. Added shared comparison links, visible topic navigation, source scope and dates, and explicit unknown fields. Updated the project map and monthly process to maintain these additional public surfaces. Continued with exact-field source comparisons, independent second passes, corrected older clinical and regulatory claims, and eight further trial records. Corrected five device and therapy records, preserved study and access limitations, and withdrew unsupported Omnipod 6 hardware ratings. Release review caught an older draft overwriting a Dexcom receiver instruction and three review dates; restored the already reviewed text and dates. Added missing primary evidence files and fixed the verification register to retain replaced assertions even when a field or citation position is reused. That release included a review-status page; it was retired on September 19, while the underlying review records remain preserved.

  • This is a partial continuation. Whole-record checks, discovery streams and many individual assertions remain incomplete.
  • Source-based facts, numerical editorial ratings and manufacturer plans remain distinct. Separate AI passes do not constitute clinical certification.
  • The Astra medical review was a high-impact spot-check, not a new claim-by-claim certification of both therapy records.
  • Most practical and country-access fields are not yet checked. Country profiles distinguish approval, supply, funding and research; missing information is not evidence that access is absent.
  • Digest source dates remain separate from editorial check dates. Sponsor results and future plans retain their study-design, generation and availability limitations.
  • Most completed trial units concern registry metadata, not full clinical outcomes. Literature discovery, older assertions, inaccessible full reports and retirement reviews remain unfinished. A pair of AI reviews is not clinician certification, and no worldwide completeness claim is made.

16 Sep 2026 · September follow-up refresh

Refresh by GPT-Astra (xhigh)

Contributors: GPT-Astra (high): Cure and Preventing evidence review · GPT-Astra (high): Living evidence review · GPT-Astra (medium): interface and trial evidence review

Follow-up evidence review with parallel Astra contributors. Added 29 item records and 11 trial records, corrected older evidence and status errors, and connected missing ranking entries with their studies. Added Australian site filtering, shareable filters, insulin-role grouping, clearer evidence summaries and score labels, faster record-specific review pages, and an anonymous About/correction page. Retired Moonshot navigation. Added a repeatable monthly discovery and field-by-field verification process. The homepage now offers direct topic shortcuts and a search box that accepts product and study names. Rebalanced AID default weights to reflect outcomes and daily use (50% glucose/safety, 40% daily experience, 10% access; freedom 20%). Unified exact-score sorting and saved personal weights, added contribution breakdowns and generation-specific caveats, and corrected Diabeloop predecessor claims and Omnipod evidence/specification summaries. Removed numerical scores from four editorial design concepts and separated them from product rankings; named pipeline programmes retain provisional assessments. The homepage has six equal topic cards, with automated insulin delivery after CGMs and clinical trials in the same grid.

  • AI-assisted editorial review; independent clinical review remains pending. Model names and settings describe the process, not a guarantee of correctness.
  • The baseline cutoff remains August 27. September checks are recorded by scope in the linked review ledger; unchanged historical record dates were not reset simply to make the site look fresh.
  • Some full papers, dynamic registries and current local access information could not be independently confirmed. Each review entry records its source depth and limits. Historical release notes describe what the site said at that time and are not current clinical guidance.
  • The field-by-field verification register is a new completeness gate and remains pending; record-level review notes and reachable links are not certification that every assertion has passed it. Discovery ledgers retain unresolved candidates rather than claiming an exhaustive index.

27 Aug 2026 · First September evidence pass

Refresh by Grok 4.6 (xhigh)

Historical model attribution. This pass was completed on 27 August and filed under September; its dated entries are preserved.

17 Sep 2026

16 Sep 2026

27 Aug 2026

This view is generated from each record’s review date. As new evidence is ingested and rankings re-scored, updated records rise to the top. How it works →