Changelog

Content freshness automation is now live across the full database. Compounds, countries, subnational states, reimbursement policies, stakeholders, and topic pages each run on their own rolling schedule — ranging from weekly to quarterly — with all proposed updates staged as drafts for review before publishing. A Friday digest aggregates all freshness activity so nothing slips through.

Paper pages now offer BibTeX and RIS citation downloads for direct import into reference managers. Secondary sources are surfaced alongside primary references, and links to related trials, manual entries, and citation relationships are made clearer throughout.

Improved the ordering of search results and how publications are attributed to their authors and institutions, making the most relevant entries easier to find.

Added a demo profile for Heroic Hearts Project as a training provider. The PharmAla demo entry was also corrected with the right name and URL.

Researchers and readers can now submit papers for consideration through a new external suggestion workflow, feeding directly into the curation queue.

Added a key-figure generation pipeline to research paper pages — extracted figures now appear directly after Blossom's Take for better reading flow. The generator enforces topic gating and a biology guardrail to keep figures focused on relevant content, with a follow-up sweep that cleaned up any non-conforming entries already in the database.

Tightened course discovery to require content-based psychedelic signals before surfacing a result, blocking generic aggregator domains and adding an LLM scope gate to reduce noise. Ran a batch improvement wave across course provider profiles and introduced Tuesday automation for ongoing refreshes. Added a monthly accreditation spot-check report for course providers to surface credentialing gaps.

Added an FDA guidance crosswalk to the clinical guidelines section, mapping treatment protocols to regulatory standards. Published a new episode cost guide for psychedelic therapy to help readers understand typical pricing structures and what drives them.

Performed investor hygiene checks, verification sweeps, and ecosystem link additions across stakeholder records as part of an ongoing funding follow-up cycle.

Restored compounds data and developer API contracts, fixed forecast pagination to be cursor-safe, and enforced distributed rate limits on the research API. Fixed an ID collision bug in the clinical-guidelines generator, and added cross-country TRD evidence modules to the Norrsken content section.

Expanded the funder and investor dataset significantly: the investor roster grew from 79 to 228 entries, investment records from 93 to 319, and active funders from 7 to 57 — all backed by sourced evidence harvested from company profiles. Philanthropy activity records also grew from 8 to 145.

Added an Ecosystem Snapshot panel to organization and stakeholder detail pages, showing reverse-linked investors, funders, government bodies, and consultancies at a glance. A new Philanthropy Funder Finder on the philanthropy category page makes it easier to discover active philanthropic funders and their documented giving.

Fixed a routing issue introduced on July 10 that caused 255 previously-indexed pages — 203 papers, 50 trials, and 2 person profiles — to return 404 errors. The pages had slugs with trailing hyphens (a natural artefact of 80-character truncation at ingestion) that were incorrectly rejected by the URL guard. All affected pages are restored at their existing URLs with no redirects needed.

Hardened external literature discovery to improve reliability when searching for research outside the core database.

Stakeholder profiles, editorial, and category pages enriched. All 30 stakeholder category pages now carry Blossom-voice editorial summaries and FAQ sections (with FAQPage structured data). Category pages for 10 disciplines — including RWD, technology, legal, manufacturing, and harm reduction — gained facet filter controls that narrow the stakeholder list in real time. Consultancy advisory and media communication profiles were researched and published for 51 organizations.

Research Landscape section launched on compound, topic, and country pages. Each page now shows a clickable trial-growth strip by year, a registry status snapshot (including stopped-early segments), and ranked bars of what's being studied — compounds, indications, or countries depending on context. The section is hidden on pages with fewer than 4 trials and scales automatically with the data.

Research Landscape UX refinements applied. Growth-strip bars now display their trial counts inline — white numbers inside tall bars, labels above short ones. Topic pages gained a subtopic breakdown panel showing trials per child topic. The section intro is now full-width, and compound ranked panels exclude placebo comparator arms, noted in the footnote.

Enterprise admin access preview added. An access-level preview switcher lets admins see what each membership tier can access without switching accounts. Fixed a layout issue that caused the enterprise signal matrix to shift unexpectedly.

SEO monitoring automated and sitemap cleaned up. Weekly Google Search Console reporting moved to unattended service-account authentication. The sitemap now excludes operator-approved noindex pages (journals, measures, books) so crawl signals stay focused on indexable content. Canonical validation findings from the weekly audit were resolved.

Patent Library: Launched a full patent section across the platform. Users can now browse a searchable patent index, open detail pages with prior art and evidence commentary, and see which patents are held by specific companies or relate to specific compounds. An overview dashboard with trend charts provides a high-level picture of the patent landscape. Patent numbers and filing dates were verified via web research, and company assignees are resolved through stakeholder alias matching.

Enterprise Intelligence: Q2 quarterly intelligence is now live for active enterprise accounts. Subscribers can access quarterly intelligence deltas, a structured briefing archive, and curated discovery paths for navigating recent developments. Catalyst timelines have been updated to surface only quarter-relevant entries.

Demo & Team Management: A public product tour is now available at /demo, featuring static visual panels for each section and tailored previews for five stakeholder types. The team management page received a full overhaul with membership status, renewal dates, and seat actions including invite, join, and remove. Enterprise accounts now resolve team members through a domain-based directory instead of a full user scan.

Access & Subscriptions: Monthly subscription tiers are now available — Pro Individual at $25/mo and Pro Team at $99/mo. A daily automated job downgrades lapsed Pro passes for both solo and team plans, and team seats are automatically restored when a subscription renews.

Search & Performance: Users can now opt in to include external academic literature alongside core database results. Under the hood, citation graph cache payloads have been bounded and static page data trimmed, reducing Vercel request and bandwidth costs.

Added retraction screening for all 2,524 papers against the Retraction Watch database. Three retracted MDMA-assisted psychotherapy studies were identified — one of which feeds the /results database — and paper pages now show a prominent red retraction banner with a link to the official notice.

Retracted studies are now automatically excluded from pooled effect estimates in forest plots. A disclosure note ('N retracted studies excluded') appears above pooled panels so readers can see exactly how many papers were removed from each analysis.

Built an AI-assisted Risk of Bias (RoB 2) assessment pipeline for trial papers. The assessor drafts Cochrane RoB 2 ratings from stored full-text, validated at 67% exact overall agreement and 97% within-one-level with expert ratings. Because RoB judgement is inherently subjective, AI drafts require human confirmation before reaching the database.

Added a self-contained HTML review interface for RoB decisions: editors can inspect AI-drafted domain ratings and rationale, edit individual judgements, and confirm or reject each paper's assessment with a single keystroke before anything is written to Sanity.

Applied a comprehensive reliability and security hardening pass across the retraction-check and RoB tooling — atomic file writes, download retries with timeouts, fail-loud format guards, server-side ID validation for decision files, and HTML output sanitisation.

Launched the Psychedelic Therapist Competency Map. The clinical-guidelines section now includes twelve competency category pages, a Training-Evidence Gap Index measuring what trials require versus what training programs teach, sixteen role- and care-stage hub pages, standards crosswalks for the Oregon OHA facilitator curriculum, Colorado DORA outcome competencies, and the RANZCP 2025 training framework, and a free self-assessment tool with a 40-item checklist. Convergence notes and in-practice summaries are now visible on every competency page.

Expanded competency library coverage with fourteen newly ingested Deep Research documents, adding protocols for mescaline (Basel), ibogaine (GITA guidelines, Conclave), 5-MeO-DMT (Beckley BPL-003-203), ayahuasca (UFRN RCT, ICEERS guide), and psilocybin (Yale manual). Standards mappings for Oregon, Colorado, and RANZCP were editorially reviewed and promoted from draft to reviewed status.

Brought the clinical results layer to row-level parity with the Metapsy/SYPRES datasets. Every result row now carries canonical Hedges' g effect sizes, Cochrane Risk of Bias 2 ratings, and both baseline-relative and dose-relative timing. Responder rows are tagged to distinguish source-observed counts from model-adjusted percentages converted to counts. CSV exports include all new fields. Risk of Bias and dose metadata were backfilled across seventeen additional papers.

Fixed extraction errors in the Mithoefer 2011 and 2018 MDMA-PTSD papers. Standard deviations stored as standard errors in the 2011 paper were corrected, and arm miscapture and a cross-paper contamination in the 2018 paper were resolved. CAPS effect sizes now match the Metapsy reference values exactly.

Added a business picture page to Sabba and an event upsert script for the SCALE Prague conference.

Design experiments: Extended /design-experiments to over 30 chart treatments across nine installments. New additions include marimekko bars, alternating-sides timelines, pictogram rows, hexbin density surfaces, parallel coordinates, magnetic-hover scatter plots, waterfall charts, diverging bars, sortable heat matrices, Gantt charts with a today-line, drag-to-brush selection windows, ridgelines, arc diagrams, dumbbell range plots, bullet charts, and treemaps. Each card states its judging criteria; the footer records verdicts on likely keepers versus on-trial treatments.

Interactive explainers: Launched /design-lab/explainers with 14 scientific figures covering field biology and landscape topics. Biology figures include an MDMA synapse mechanism step-through, a population eligibility dot grid, a dose-response therapeutic window, and a clinical-trial attrition chart. Landscape figures cover cost of care, therapist workforce supply vs demand, regulatory approval timelines (esketamine, MDMA, psilocybin), research geography, and the long-tail evidence base. Five evidence-reading explainers were added alongside — a field growth bar chart, compound×indication heatmap, trial-size histogram with a sample-size threshold, a functional-unblinding illustration, and an icon-array number-needed-to-treat. Growth, heatmap, and trial-size charts run on real Blossom trial counts.

Literature figures: Rebuilt figures from nine top-cited clinical trials carrying exact tabulated numbers from the source papers, including Carhart-Harris et al. 2016 (open-label psilocybin for depression — QIDS-16 trajectories, 67% remission at one week), Griffiths 2006, Nagele 2021, and Lijffijt 2022.

Visualization opportunity map: Added /design-lab/opportunities, a working drafts page with seven entity-page visualizations running on real Blossom trial data — a growth strip, phase-mix chart, status snapshot, ranked compound/indication bars, trial-size histogram, evidence-matrix heat-strip, and country access-readiness scorecard. A companion document maps 21 visualization proposals across the /countries, /compounds, and /topics evergreen pages, each with a concrete Sanity data source and priority/effort rating.

Sabba handbook and stakeholder profiles: Published the Sabba MOU handbook and improved its navigation — Previous and Next now replace the Overview tab in the detail nav. Replaced hand-tuned negative offsets across five card rows with a reusable FlowConnector component that spans the exact grid gap. Landing stats now display live Sanity counts (2,499 papers, 1,185 trials, 1,928 stakeholders, 673 courses), and several sections gained 'See it live' panels with verified links to real pages. Separately, completed an academic research profile audit and applied provenance data to stakeholder records across the database.

Dithered image headers across the site. Category, entity, and blog page headers now feature a film-grain dither treatment, giving them a textured, editorial look instead of flat images. Design system documentation has been updated with the new visual guidelines.

Navigation upgraded with a live third-level menu. The site header now serves dynamic pane content via an ISR-cached API, and the Analysis section has been reordered for better discoverability. Header copy now uses consistent UK English spelling throughout.

New health economics blog and email capture. A health economics blog post has been published, and readers can now sign up to follow the economics series via a new email capture flow.

Country scorecard narratives moved to Sanity. Scorecard narrative text for country pages is now managed directly in Sanity Studio rather than hardcoded, making content updates possible without code changes. Source-label integrity is now enforced in the Studio editor.

Stakeholder capability snapshots consolidated. Stakeholder detail pages now render capability snapshots from a centralized data-driven registry, replacing 21 individual components with 14 shared configurations. Schema constants and label dictionaries have been extracted into shared modules for easier maintenance.

Fixed two bugs affecting data integrity. The voting API now validates that a document is eligible to receive votes before applying any changes, preventing votes from being written to unintended content types. Team invitations now correctly merge into a user's existing account metadata rather than overwriting it, so notification preferences and other settings are no longer silently lost when a user is re-invited.

Several document types that were previously unreachable in Sanity Studio are now accessible from the sidebar. Pages, Videos, ICPR Meeting Slots, Clinical Competencies, and Framework Assessments all have dedicated entries. A new Website and System section groups site-management content together, and Bloom Intros have a deliberate top-level slot.

Added a policy advocacy deep research import report and a stakeholder matrix, with layout refinements applied to the matrix view.

Fixed normalization of Bloom intro items in outgoing emails. The monthly recap draft workflow is now scheduled to run on Mondays.

Grouped CE and non-CE course variants into a unified view across course listings, detail pages, provider pages, and related-course surfaces. Each course now presents its available format options in one place rather than appearing as separate entries.

Added a nested category landscape view to stakeholder browsing, making it easier to explore organizations by their subcategories. Fixed a server error on stakeholder category pages that occurred when reading search parameters during server-side rendering — these pages now render dynamically while keeping Sanity content cached.

Improved the weekly ASK feature with better relaxed search behavior and clearer submission guidance, making it easier to find relevant content when exact matches are unavailable.

Fixed a citation lookup script that was incorrectly resolving matches through draft paper IDs during Phase B imports — lookups now skip drafts and only match against published records.

Enriched investor profiles across the stakeholder database with deeper category data and research outputs from the investor deep-research pipeline. A new idempotent script now upserts approved investor leads from the review workflow directly into Sanity, making it easier to promote vetted investors without manual data entry. Public-facing investor copy was also cleaned up for clarity and consistency.

Fixed two issues with the PostHog analytics and error-tracking setup. Source map uploads are now gated behind an explicit flag so production builds no longer fail when the configured API key lacks the required scope. PostHog initialisation was centralised across the app, known browser-extension errors are filtered from traces, and Clerk authentication failures now degrade gracefully to sign-in links instead of blocking the UI or checkout flow.

Enriched course provider profiles are now live on stakeholder pages. Each course provider page shows an expanded profile — covering background, who the training is for, prerequisites, accreditation, costs, and outcomes — alongside a new 'Similar providers' module so visitors can discover related educators. Three care-delivery snapshot types (Clinical, Practitioner-Led, and Non-Clinical) were added to round out the picture.

Stakeholder category pages have been redesigned as full discovery surfaces. Pages now include an editorial summary, a stat strip, by-country breakdowns linking to country pages, URL-based filters (modality, status, country), an FAQ block, and structured data (CollectionPage, ItemList, FAQPage) for better search visibility. A content-enrichment script and a Google Search Console geo-demand monitor were also added to keep category content fresh and to surface which regions warrant dedicated pages.

A drug developer news workbench is now in place to support ongoing tracking of pipeline and regulatory updates. This round added several focused metadata waves covering the developer roster, plus an exact-patch decision builder to make targeted record updates safer and more auditable.

Course listings on psychedelic-training pages received several quality fixes: duplicate courses are now deduplicated, badges display in a consistent gray style, and HTML entity artifacts (e.g. &) are decoded properly. AI-generated inline citation links that were leaking into prose fields have been stripped from both new enrichment output and existing stored records.

Fixed a query bug that caused the Similar course providers module to render no candidates under the live Sanity client — it now reliably surfaces matches. Spacing above the module was also corrected to match the rest of the page layout. Developer docs were updated to reflect that Next.js 16.2 defaults to Turbopack, requiring an explicit --webpack flag to opt out.

Added a new Drug Development Catalysts section with a filterable calendar of upcoming milestone events across drug programs. Users can filter by drug type, development stage, and timeframe to track expected regulatory submissions, trial readouts, and other key milestones.

The Drug Developers overview page and Development Stages page now include a pipeline landscape matrix mapping developers across development stages. This provides a high-level view of who is active in which parts of the development pipeline.

Drug developer profile pages now feature a momentum panel summarising recent development activity. The panel highlights active programs, trial additions, and publication patterns to surface developers gaining traction.

Developer profile pages now show a Similar Developers section with peer comparisons based on pipeline composition and stage distribution. Each suggestion includes an explanation of why two developers are considered peers.

Fixed an issue where the momentum panel included drug candidates outside the relevant development stage range. Corrected TypeScript types for catalyst milestone windows.

Hardened weekly digest email delivery with built-in rate limiting and automatic retry logic. The Resend API integration now queues requests to stay within sending limits and retries gracefully on rate-limit errors using exponential backoff, reducing failed or dropped digest emails.

Improved how the digest intro is resolved: instead of requiring an exact date match, the system now looks back up to five days when fetching intro content. This means the weekly digest is far less likely to go out without an intro if content is published slightly ahead of schedule.

Reduced unnecessary contact-sync calls triggered by Clerk webhook events. When a digest send updates user metadata, any user.updated webhook arriving within 15 minutes is now skipped for Resend contact syncs — cutting down redundant API traffic generated by the digest delivery itself.

Added an Advocacy and Nonprofit Intelligence panel to country scorecard pages. Each country now surfaces the most relevant advocacy organizations and nonprofits working in that jurisdiction, ranked by geographic fit — from organizations explicitly covering the country to those with physical offices or global mandates there.

Each organization listing shows its type, jurisdiction scope, engagement status, and focus areas, with links through to full stakeholder profiles. Up to eight top-ranked organizations are shown per country, filtering out purely global profiles to keep results locally relevant.

Added a data enrichment workflow for policy advocacy stakeholders. New scripts automate web research using AI and Exa search to propose updates to stakeholder jurisdiction coverage, policy URLs, engagement modes, and verification status, which can be reviewed in a structured HTML report before being applied to the database.

Added automatic generation of Blossom Takes during the paper ingestion pipeline. New research papers added to the database now include an editorial perspective without any manual effort.

Researcher profiles now automatically pull in coverage from up to four recent papers, giving visitors a richer snapshot of each researcher's work at a glance. A bug that could produce duplicate aligned profiles for the same researcher has also been fixed.

Automated the sync of trial links in the clinical trial infrastructure (CTI), keeping trial references current without manual intervention. Improved how CTI evidence is displayed on stakeholder pages, making the supporting data behind each trial easier to review.

Restored the Therapsil data feed after a broken connection path was detected, ensuring Therapsil research content flows correctly into the database again.

Paper discovery pipeline hardened. The automated system that surfaces new research now has improved classification, smarter deduplication, and a new keyword-based discovery channel via OpenAlex. Papers are linked to related work more reliably, the pipeline handles network timeouts gracefully, and abstract formatting is now consistent across newly ingested papers.

Free account experience improved. Free account holders now see clearer messaging about what's available versus Pro-only — across search filters, topic reports, results pages, and sign-in prompts. Search filters inherited from Pro sessions are preserved as read-only, and gated topic report fields are correctly redacted for non-Pro API callers.

Animated CTA button introduced sitewide. A new shared motion call-to-action button component was rolled out across pricing, paywall banners, newsletter sign-ups, course pages, and public conversion touchpoints for a more cohesive upgrade flow.

Design lab updated with new prototypes. New layouts were added exploring full-card hover treatments, Blossom testimonial sections, and a pricing table variant — available in the internal design lab for review.

Researcher profiles received a significant upgrade. Affiliation cards now display role titles alongside Current and Past tenure badges, making it easier to understand each researcher's position at a glance. Bios fill the full page width for better readability, and the most-cited publications section now correctly filters to papers that have actual citations.

Clinical trial infrastructure stakeholder pages were enriched with deep research findings, adding richer context and supporting evidence across the CTI category. Compound pages also received stronger narrative evidence links, improving how evidence is connected and surfaced within reports.

Five new researcher leaderboard ranking pages launched along with a hub page, each featuring structured ItemList data for search engines. Compound and topic reports now carry honest author and reviewedBy bylines — both visible on the page and embedded as JSON-LD — making editorial attribution clear to readers and crawlers alike.

Person pages were enriched with expanded structured data, including areas of expertise, institutional affiliations, and social profile links, improving how researcher profiles appear in search results.

Fixed an issue where clinical trial pages would fail to load when a country reference could not be resolved. Tenure badges on researcher profiles now use the correct on-brand greyscale style instead of an off-palette green.

Enriched person and journal pages with computed research footprints. Researcher profiles now display total papers, active years, trials, top studied compounds and topics, most-cited work, and frequent co-authors — all derived server-side so the information is visible to search engines as well as users. Journal pages received a matching footprint alongside new structured data (JSON-LD) that identifies each journal as a Periodical entity with its ISSN and publication history.

Added clinical guidelines discovery and cross-linking on research pages. Compound research combination pages now include a discovery card surfacing related clinical guidelines, with a direct link to explore them on the guidelines index. The guidelines index gained reciprocal links back to each compound's research hub. Measure pages gained a 'Most Studied With' section listing the top compounds and topics that co-appear with each measure across the literature.

Applied a noindex directive to 313 stub pages identified as too thin to serve in search results — 277 empty journal entries, 35 measure instruments, and 1 book. Each page retains its links (noindex, follow), preserving link equity while steering search engines toward pages with substantive content. The target list was generated by an automated thinness scorer and reviewed before being applied.

Fixed cache-tag subscriptions for measure and journal roster pages. Previously, adding or editing a paper would not immediately refresh the roster on the measure or journal it referenced — the page would stay stale until the host document was re-saved. Rosters now subscribe to the paper-level cache tag and update instantly when any referencing paper changes.

Refreshed compound data and updated the visual lab workbench. Also enriched clinical trial infrastructure stakeholder records with market-map categories, source-backed location data, and additional detail gathered from web sources.

Launched SEO Wave 2 answer capsules across the platform. Compound, topic, country, book, and research-combination pages now display structured AnswerCapsule, ByTheNumbers, and FaqSection components that surface key stats and frequently asked questions at a glance. Capsule styling was refined to a clean, plain design.

Published a comprehensive Wave 2 content corpus: 11 compound capsules, 9 topic capsules, 15 additional clinical-topic capsules, 12 country legal-status capsules, 10 book capsules, and 40 FAQ question-and-answer pairs across 20 compound/topic drafts — all written in Blossom's research-grounded voice.

Added a verified external-ID allowlist and automated applier that backfills valid identifiers (DOIs, ClinicalTrials IDs, and similar) into structured JSON-LD across research pages, improving citation accuracy in search-engine results.

Completed enrichment of the full Tier-3 methodological topics queue. Ten topics received detailed research overviews: set-and-setting, microdosing, personality and trait factors, interpersonal functioning and social connectedness, immunology and inflammation, creativity, public health and behaviour change, equity and ethics, implementation and service delivery, and health economics and reimbursement.

Added in-depth research pages for 20 conditions and demographic groups — the largest single-day expansion of the topic library to date. New pages span mental health (major depression, depressive disorders, bipolar disorder, OCD, eating disorders, BPD, personality disorders, schizophrenia), neurological injury (TBI, neurological injury umbrella), pain (chronic pain, fibromyalgia, headache disorders including cluster and migraine), addiction (alcohol use disorder, tobacco use disorder, opioid use disorder), and special populations (autism, older adults, peripartum, palliative and end-of-life distress).

Each new page carries honest, calibrated evidence ratings. Where prior boilerplate overstated efficacy — sometimes rating compounds 'High/Large' based on extrapolation or preclinical data alone — the updated pages reflect the actual clinical evidence: mostly Small/Low or Very Low ratings, with clear explanations of what that means for patients and clinicians. Pages for schizophrenia, BPD, and Cluster A personality disorders now explicitly frame psychedelics as a contraindication or safety concern rather than a treatment opportunity.

All 471 citation links across 22 enriched topic pages now point to Blossom's own /papers/ pages instead of raw DOI links to external publishers. Blossom paper pages are public and provide richer context — abstracts, research summaries, and curated references — before linking out to the publisher. Of 239 unique citation DOIs, 215 matched papers already in the Blossom database; the remaining 24 (plus non-paper sources like WHO and FDA documents) continue to link externally.

Research content — Significantly enriched the suicidality topic page (/topics/suicidality) with evidence-grounded content. The page now leads with crisis resources and a medical disclaimer before presenting the research, and includes a ~1,480-word public research report backed by 39 citations.

Evidence accuracy — Updated the evidence hierarchy to reflect what the research actually shows: ketamine and esketamine have the strongest rapid anti-ideation data, with esketamine holding a 2020 FDA indication for major depressive disorder with acute suicidality risk. Classic psychedelics are now framed as unproven for suicidality, with dosing-day risk signals clearly noted.

Content corrections — Fixed several inaccuracies from the previous version of the page, including invalid data-consistency labels and overclaims about ayahuasca and esketamine. The page now presents a more honest picture, including the important caveat that reducing ideation in a clinical trial is not the same as proven suicide prevention.

Launched a new Top 10 section with curated lists of the most-cited research papers organized by compound and topic. Lists now feature topic and compound chips, a sticky table of contents, FAQ sections, a methodology note, paper-level study-type and citation signals, and an Ask CTA. Compound and topic sidebar pages also gained "Start here" cards that link visitors directly to the relevant top-10 list.

Top-10 pages received full SEO coverage: ItemList and CollectionPage JSON-LD structured data, Open Graph images with paper-count metadata, canonical URL routing (shadow URLs return 308 redirects, the legacy /research/top-10 path retired), and sitemap inclusion with accurate lastmod timestamps.

Five research topic pages were rewritten with deeper, honest, evidence-grounded content: Veterans, Adolescents, Treatment-Resistant Depression, PTSD, and Anxiety Disorders. Each page now leads with the strongest available clinical evidence, includes explicit caveats around trial limitations and unblinding concerns (such as the 2024 FDA rejection for MDMA-assisted PTSD therapy), updated effect matrices, and expanded research reports with web-verified citations.

Fixed a bug where page titles displayed the site name twice. Corrected a link-color issue on top-10 kicker elements — gray is now used consistently for informational links, with orange reserved for action buttons. Applied additional code-review polish across the new top-10 pages.

Upgraded the CMS backend from Sanity Studio v5 to v6, moving to a Vite 8 build system and updating all plugins to v6-compatible versions. Schema extract and typegen output are unchanged.

Redesigned regional report pages with richer data displays. Each region now shows a country leaderboard ranked by trial activity, alongside breakdowns by trial phase, compound, and topic. Trend charts and hub-and-spoke navigation links to related countries, compounds, and topics complete the overhaul.

Upgraded SEO and discoverability for all regional pages. Pages now emit MedicalWebPage, Dataset, and ItemList structured data (JSON-LD), carry proper canonical URLs and Open Graph tags, and display a 'Data as of [Month Year]' freshness indicator. The /regions index also publishes a sitemap entry tied to real content freshness.

Added curated Q&A content to all seven regional reports — six questions per region covering regulatory landscape, access programmes, and comparative standing. This content is rendered inline on each region page and is also surfaced through region-scoped ASK chat queries.

Fixed a production error on all region detail pages caused by trial phase data being stored as an array rather than a plain string. Also fixed top-10 lists that were mistakenly shipping with draft placeholder text, restored correct cache routing for trial category pages, and re-synced the Bloom weekly digest cron schedule.

Added a psychedelic therapy economics model to the resources library, providing structured data on the financial dimensions of psychedelic-assisted treatment.

Fixed the AI-powered regional roundup script end-to-end, resolving a JSON schema validation error, an unsupported API parameter, and a seed/apply document mismatch. All 7 global regions — Europe, Nordics, North America, Latin America, Asia-Pacific, MENA, and Sub-Saharan Africa — now generate clean editorial drafts ready for review.

Fixed weekly digest delivery reliability to ensure members consistently receive their scheduled digests. Sent a Bloom apology email styled to match the platform welcome design, and scheduled a follow-up make-good email.

Fixed an issue where citations could go dead when continuing an anchored follow-up conversation in ASK chat. Citations are now validated before new assistant turns are appended, ensuring references remain accurate in both private and shared thread views.

Added detailed funding context for psychedelic research initiatives in the Netherlands, Germany, and France country reports. Each country now includes source-backed information on named funders, funding amounts, and access or reimbursement caveats.

Published a new evergreen resource page for the ICPR 2026 conference talk 'Are reimbursement pathways ready for licensed psychedelic therapies in Europe?' by Floris Wolswijk. The page includes a full slide-by-slide viewer with transcripts, related links per slide, and a downloadable PDF of the slide deck.

Added a video entry for the ICPR 2026 talk on the videos page, and registered the new resource in the implementation resources registry so it appears in the Road to Access section alongside other implementation resources.

Updated the sitemap to include the new talk resource and video pages, ensuring they are discoverable by search engines.

Fixed a reliability issue in the monthly research recap where important papers were being dropped from their correct sections. The root cause was the LLM mangling long Sanity UUID references — papers are now referenced by short stable tokens (p1, p2, …) instead. Dropped papers are re-inserted into the section matching their research group rather than falling into a catch-all at the end.

Improved the monthly recap scheduler to prevent accidental double-runs when a delayed cron job fires after midnight. The scheduler now skips regeneration if a recap for the month already exists, handles cron timing delays more gracefully, and deduplicates internal evidence ordering logic for cleaner output.

Restored the ability to save blog posts in Sanity Studio that contain site-relative links (e.g. /papers/some-slug). The Studio's href validator was too strict — rejecting valid internal paths — which blocked editors from saving 289 older documents containing ~3,700 relative links. The validator now accepts site-relative paths and in-page anchors alongside standard URLs.

Closed a security vulnerability where protocol-relative URLs (e.g. //evil.com or /\host) in link fields could act as open redirects or phishing vectors. The fix applies to all Studio content editors and adds a render-layer guard in the site's link resolver, which now neutralizes unsafe hrefs before they reach the browser.

Fixed a caching bug on clinical trial detail pages that caused stale data to be served in certain request contexts. Trial details now render correctly with fresh, request-aware data.

Added a new Group Therapy Economics Model to the Road to Access implementation resources. The interactive tool models staff costs, session capacity, and revenue projections for group-format psychedelic therapy programs, with country-aware assumptions covering salary benchmarks across multiple regions.

Expanded the personnel-hours resource with refreshed regional cost data and updated links to related implementation models. Country-specific staff cost profiles are now shared across resources so estimates stay consistent as data is updated.

Added a geo-location API endpoint to support country-aware defaults in interactive tools. Simplified the Europe Access Clock table layout and streamlined the Spravato Access Precedent visual for cleaner presentation.

Fixed an issue in course FAQ copy where a provider object was incorrectly passed instead of a plain string, preventing some FAQ entries from rendering correctly.

Added a new Road to Access Resources library at /implementation/resources/, a structured collection of interactive tools and reference materials organized around the psychedelic therapy access journey. Each resource maps to a specific step — from regulatory approval through reimbursement and workforce readiness — and is designed for different audiences including researchers, investors, drug developers, operators, and policy readers.

Launched six new interactive data tools covering key access bottlenecks: Europe Access Clocks (visualizing the gap between approval and reimbursement across European countries), the Netherlands Psilocybin Access Sites map (showing where therapy could realistically be delivered), the Spravato Access Precedent visualization (exploring esketamine's reimbursement trajectory as a comparator), the Payment Bundle Model (breaking down cost components for psilocybin therapy sessions), the Netherlands Workforce Capacity Model (estimating therapist availability against projected patient demand), and a Trial Follow-Up vs HTA Horizon tool highlighting evidence gaps between trial endpoints and what payers require.

Significantly updated the Netherlands country and reimbursement pages, incorporating deep research on the ZIN HTA procedure and workforce capacity constraints. Added a dedicated ICPR Netherlands Access Map page surfacing site-level data on potential treatment locations.

Added all new resource pages to the sitemap, ensuring they are discoverable and indexed going forward.

Launched Enterprise Intelligence, a new access tier that delivers curated competitive-intelligence briefings to enterprise subscribers. Briefings cover indication-level pipeline data, clinical-trial landscapes, and regulatory catalysts, and are gated to accounts provisioned in Sanity with matching Clerk metadata. Public visitors see a sample indication with a locked panel inviting them to enquire.

Added an ICPR 2026 meeting booking page at /icpr-meeting. Researchers attending the International Conference on Psychedelic Research in June can reserve a 20-minute slot with Floris directly on the site. The page shows live availability, handles one booking per user, and gives admins a full slot overview.

The Ask AI assistant can now answer questions about the Research API, account API keys, and pricing. A new internal site-docs index feeds the assistant structured summaries and keywords so it routes API and access questions accurately instead of hallucinating.

Added a Blog index page at /blog listing all articles with publication dates, article types, and series tags. The page is SEO-optimised with breadcrumb JSON-LD and a long cache TTL.

Enriched the psilocybin-for-depression evidence base with source-level reviews of the Davis, Goodwin NEJM, and Goodwin JAD trials against Metapsy meta-analytic data, surfacing responder rates and effect-size comparisons used to validate the database entries.

Automation reliability: The weekly Sanity content backup is now running correctly again — a misconfigured authentication token that had been causing export failures has been fixed. Sanity Studio deployments have also been consolidated into a dedicated script (npm run deploy:studio) that cleanly manages credentials, replacing a more fragile manual process.

Clinical guideline data quality: The clinical guideline pipeline now classifies source documents into five quality tiers — trial supplements, protocol papers, lab manuals, SOPs, and historical handbooks. This lays the groundwork for finer-grained filtering and ranking of research material by source quality in future views.

Added dose summary report pages for compounds and research topics. A new /dose-summaries index lists all available reports, and each detail page at /research/[compound]/[topic]/dose-summaries presents structured dosing data drawn from clinical trial results, including dose ranges, schedules, and administration routes.

Added adverse event summary pages for compounds and research topics. A new /adverse-events index lists all reports, and each detail page at /research/[compound]/[topic]/adverse-events surfaces safety and tolerability data extracted from clinical studies, covering incidence rates and severity breakdowns for named adverse events.

The AI Ask assistant can now retrieve content from both dose summary and adverse event reports when answering research questions, giving users richer, source-backed answers on dosing and safety topics.

Added structured dose and adverse event data to trial records. Dosing information and adverse events from source documents are now normalized into canonical schema fields, making it easier to compare treatment protocols and safety profiles across studies.

Added a new Curation page explaining how Blossom decides what research to include and how data is reviewed. The page covers editorial standards, source criteria, and the data enrichment process.

Fixed a bug in the Ask feature where assistant messages were dropped when saving long conversations. Full chat history is now correctly persisted, and feedback submissions can link back to specific assistant responses.

Added structured data (JSON-LD) markup to course listing and detail pages across both the Blossom Courses and Psychedelics Courses directories. Course pages are now eligible for Google rich results, which can surface course titles, providers, and other details directly in search listings and improve click-through rates.

Consolidated the course schema generation into a shared helper, replacing duplicated inline definitions on each page with a single consistent implementation. This ensures both course directories produce identical, correctly formed Schema.org markup.

Added internal diagnostic tooling that pulls data from Google Search Console and PostHog to identify indexing gaps and low-CTR pages. These scripts feed a prioritised backlog of SEO improvements to work through over the coming weeks.

Improved Ask answer quality and added better citation display so responses trace more clearly back to source papers. Connected Ask to the events database, expanded example prompts, and polished the layout — including the ability to delete individual threads.

Audited and patched clinical results provenance for key trials, including MDMA MAPP1 and the psilocybin cancer anxiety studies, to ensure data points link to their correct source publications. Added a methods explainer page describing how results data is sourced and structured across the database.

MDI blog posts now weave organic inline links to Blossom papers directly into the article body and close with an explicit reference sentence, replacing the old styled CTA boxes and standalone Community sections. Post URLs also switched to clean, readable slugs (e.g. /psilocybin-microdosing-creativity-2024) instead of doc-ID prefixed forms.

Expanded the /latest feed to surface updated Sanity content beyond papers and trials — including pages and other public document types — with proper cache invalidation so the feed stays current after edits.

Fixed sitemap routing by introducing a sitemap index and correcting canonical metadata across the site. Added redirects for legacy GSC URLs that were renamed, and tightened PostHog search event tracking to distinguish real keyword searches from browsing and pagination activity.

Added a weekly automated cron that syncs MDI research papers to WordPress every Tuesday at 08:00 UTC. Each generated post now includes a standard research call-to-action module and sends requests with a modern browser signature to bypass SiteGround's bot detection. A pre-launch flag keeps founder notification emails off until the sync goes live.

Raised the MDI sync maximum execution time to 800 seconds — enough headroom to reliably handle worst-case batches of three papers, each requiring LLM summarisation, image generation, and a WordPress media upload.

Added overview images to country pages and updated the tooling for managing country and US state page content.

Fixed the global site navigation and chrome appearing on the courses subdomain — it now correctly hides on the client side.

Added US state and territory pages to the Countries section. Each US state and territory now has its own research profile page covering policy status, clinical access, key trials, and regulatory milestones. Pages include an interactive activity map and a policy timeline, and are linked from the main United States country page.

Refreshed the Fluence provider catalog in the psychedelics courses section, updating course listings and teacher profile pages to reflect the latest offerings. Fixed an issue where the Blossom site header and navigation were incorrectly appearing when visiting the courses domain — the courses experience now renders cleanly on its own.

Restored automated notifications for the paper discovery cron job. Research teams will once again receive alerts when new relevant papers are identified by the daily discovery pipeline.

Added richer evidence display to clinical trial outcome pages, including support for responder-only analyses, subgroup breakdowns, pooled treatment arms, and time-elapsed result rows. Forest plots and outcome tables now surface more nuanced data from published papers.

Redesigned the pricing page to sharpen its positioning around research intelligence value. Team is now the recommended self-serve plan with clearer feature differentiation, Enterprise pricing is anchored more concretely, and the team value comparison section is wider for better readability.

Added a guided, category-based journey to the Implementation section that helps users navigate the road-to-access process step by step. Category pages now surface relevant content in a more structured flow.

Refined stakeholder and course detail pages, improving layout and back-navigation behavior across secondary detail views.

Launched the Government Funding Catalogue — a new page at /funding/government listing government grants awarded to psychedelic research organisations. The table is sortable by year, amount, funder, and recipient, and supports multi-select filters for funder agency, compound, and indication. Stakeholder profile pages now include a "Government Funding Received" section showing grants at a glance, with a deep-link into the full catalogue filtered to that recipient.

Seeded the catalogue with 21 real grants sourced from NIH RePORTER and primary agency announcements. Funders represented include NIDA, NIMH, NIAAA, NCCIH, NCI, DoD CDMRP, VA, ARPA-H, the European Commission (Horizon Europe), ERC, UKRI/MRC, SNSF, and Australia's MRFF. Recipients include Johns Hopkins CPCR, NYU Center for Psychedelic Medicine, Yale, Emory, UC San Diego, Imperial College London, Monash University, the University of Sydney, and others. Grant amounts were verified against primary sources; estimates are flagged inline.

Added ?promo= support to the /pricing and /access pages. Visitors arriving via a referral or marketing link with a promo code in the URL now see the discount pre-applied on the Stripe Checkout page — no manual code entry required. The code is preserved through the sign-up redirect for signed-out flows, and both promo attempts and successful applications are tracked in PostHog for funnel analysis.

Fixed a rendering bug in the PE Review table where markdown files that end immediately after the closing frontmatter --- (no trailing newline) rendered the raw YAML keys as visible text. Switched the Next.js dev server from Turbopack to the standard Webpack pipeline to resolve an indefinite compile hang on slow filesystems.

Improved readability across the PE Review section. The dashboard table now uses fixed column widths so that long titles no longer cause layout shifts, and course titles wrap to two lines with a tooltip instead of being cut off mid-word. Provider names and parent programme labels also show tooltips on hover.

Reference country pages received the same treatment: the article renderer was updated to handle source-block sections more cleanly, improving visual separation between main content and citation lists.

Fixed a bug where newly registered users were not enrolled in the weekly digest. Signup email preferences are now written to Clerk metadata at account creation time, ensuring fresh accounts receive The Bloom without needing to manually save their preferences first.

Added a new Ibogaine Care Providers directory at /ibogaine-care-providers — a source-backed listing of organizations that directly provide or enable access to ibogaine care. Providers are grouped by country, shown on an interactive map, and each card links through to the organization's Blossom stakeholder profile and official website.

The directory connects to Blossom's Ibogaine compound report and draws from enriched stakeholder records with public location data, care category tags, and descriptions. A disclaimer notes that inclusion is not an endorsement, medical referral, or safety certification.

Added a repair script for monthly recap blog posts that were created before the WordPress importer ran, patching missing SEO descriptions, featured images, and outdated internal links. The script runs in dry-run mode by default and only touches fields that were previously absent.

Applied source-reviewed quality standards to clinical outcome data across six compound cohorts — MDMA, psilocybin, ketamine, esketamine, LSD, and additional small compounds. Trial endpoint schedules have been normalized and backfilled, improving the consistency and accuracy of outcome data visible throughout the database.

Hardened Stripe subscription checkout and webhook provisioning to reduce edge cases where successful payments did not fully activate Pro access. A backfill tool was added to retroactively identify and restore Pro access for any subscribers affected by historical provisioning gaps.

Added nine Pro-gated European country reference pages under /pe-review/europe, covering psychotherapy licensing and supervision requirements for the Netherlands, Germany, UK, Switzerland, Czech Republic, Spain, Portugal, Austria, and France. Each profile maps statutory licensing bodies, professional associations, supervision norms, and credential acceptance paths sourced from PE's latest jurisdiction documents.

Launched a dedicated /pe-review/framework page presenting the PE L1 credential floor criteria, replacing the previous placeholder. A new shared sidebar now links the PE Review dashboard, framework page, and all nine country profiles in one place, and the dashboard's credential links have been updated to point to the new internal pages.

Shipped a credential-floor audit that cross-references approximately 14 PE-named training programmes against the L1 credential floor and auto-derives pass, review, or fail assessments, with results exported as a structured output file.

Improved monthly recap detail pages with previous/next month navigation and a structured list of referenced research papers, making it easier to browse between months and follow citations. Section headers now use title case, inline paper links are more robust, and recap entry grouping and ordering logic has been hardened for consistency across all recap pages.

Launched the new Pink Elephant (PE) Review dashboard at /pe-review, giving Pro subscribers a filterable, searchable table of psychedelic training programmes assessed against Pink Elephant framework criteria. The dashboard surfaces structured fields for training hours, substance self-experience components, real-client practicum details, and professional entry requirements, and includes a one-click Excel export. Fixed a Pro access bug where the auth check was reading the wrong metadata path, causing 403 errors for paying subscribers on the dashboard and export route.

Enriched over 80 psychedelic training programmes with verified, structured data. Each programme was reviewed against live provider pages and patched with contact and practicum hours, required substance experience (substance type, session count, supervisor requirement), real-client practicum status, and professional entry requirements including eligible professions and licence prerequisites. All enriched values carry a verified-by provenance tag for audit traceability.

Added an automated pipeline that generates monthly research recap blog post drafts. At the start of each month the system gathers published papers from the previous month, groups them by topic area, and uses an LLM to draft narrative summaries following the site's editorial style guide. Drafts land in Sanity Studio ready for editorial review and refinement before publishing.

Fixed a data quality issue affecting 471 training courses where missing internal array keys were preventing Studio editors from opening and editing those records. Separately, removed scraper-generated HTML and CSS junk from the descriptions of 18 courses that had been incorrectly imported, allowing operators to re-enter clean descriptions from the source.

Launched /ask, a new AI-powered research assistant built into Blossom. Signed-in users can ask natural-language questions about clinical trials, countries, researchers, and papers — the assistant queries the live database and cites every source it uses. Conversations are saved as persistent threads so users can return to earlier research sessions.

Added citation hover previews that surface source details inline without leaving the conversation, along with a slide-in source panel for full citation verification. Each message now shows a thumbs-up/down feedback widget, users can stop a response mid-generation, and administrators can bypass the feature kill-switch to test changes before public rollout. Share a thread via a public link at any time, with the ability to revoke access.

Refreshed research content for the Netherlands, Belgium, France, and Germany, including improved healthcare access information for Germany. Backfilled internal v1 country reports across all country profiles and introduced a research snapshot to streamline the enrichment workflow. Reimbursement pages across country profiles have been realigned to the design system.

Expanded the clinical trial registry ingestion scripts to capture a broader set of studies, improving data coverage in the trials database.

Reduced hosting costs by enabling CDN-edge caching across the PsychedelicsCourses section of the site — pages are now served directly from Vercel's edge network for up to 24 hours, with stale content revalidated in the background for up to 30 days. A duplicate daily operator pulse cron job was removed, and the email-sequences cron was throttled from every 15 minutes to once per hour.

Fixed the PsychedelicsCourses microsite header and footer so they reliably appear when visitors arrive directly at psychedelicscourses.com, not just when routed through the internal proxy. Also corrected a layout issue on the course-matching quiz where the filter sidebar was incorrectly sticky on large screens.

Improved build stability by skipping Sanity fetch traffic counters during static page generation, preventing the per-request rate-limit guard from triggering false positives at build time.

Redesigned the Blossom Pro success page to help new subscribers get started faster. The page now leads with direct links to the research database, includes a step-by-step getting-started checklist covering research tips, email preferences, and account management, and adapts its guidance based on plan type — team and enterprise admins see workspace setup prominently, while individual subscribers are directed straight to the papers database.

Cleaned up the curated Top 10 Psychedelic Papers lists that were previously migrated from WordPress. Empty legacy placeholder documents were removed so they no longer appear in listings or routes, missing SEO descriptions were filled in for active lists, and ranking gaps were resolved or flagged for manual review where a confident paper match could not be confirmed.

Added a new public REST API at /api/v1/countries that exposes structured country research data for external partner integrations. The Germany country page was also significantly redesigned and refreshed with updated clinical and regulatory information for 2026.

Built a synchronization workflow between Sanity roadmap feedback and GitHub Issues, keeping the public roadmap and internal implementation tracking aligned automatically. The admin dashboard Feedback tab now surfaces linked GitHub issue status and metadata alongside each feedback entry.

Completed a clinical competency data merge to consolidate duplicate entries in the clinical guidelines dataset. The GuidelineCard component was updated to reflect the cleaned data structure.

Added an Excel-based review workflow for the clinical competency merge project. Operators can now export the full merge proposal to a spreadsheet with decision dropdowns, edit canonical titles and summaries in place, and ingest all changes back into the proposal — title and summary edits that were previously discarded are now applied correctly. Switched the spreadsheet parser from ExcelJS to SheetJS to fix read failures on Excel-saved files.

Expanded the clustering pipeline to handle the 457 competencies that did not cluster in the first pass. A new singleton analysis tool identified 262 fold-in candidates and surfaced 11 new sub-clusters, all of which operators reviewed and applied — bringing total coverage to 774 source documents across 53 canonical clusters. A diff tool was also added to preview what a spreadsheet ingest will change before committing.

Broke down the oversized 389-member 'Psychedelic-assisted care model' cluster into focused sub-groups. Nine members were redistributed to better-fitting existing clusters, 28 tight topical sub-clusters were created (covering 97 members), and 7 broad category clusters were formed for the remainder. The merge project now stands at 88 review clusters and 266 canonical concepts — a 72% reduction from the original 952 competency documents.

Fixed a parsing bug where Gemma 4's thinking-output parts were being concatenated into the JSON response, causing silent failures across the LLM confirmation step. Added a per-request 60-second timeout with retry backoff, and restored the --skip-llm flag as an escape hatch for running the pipeline deterministically when the LLM is unavailable.

Fixed the country filter on the courses page, which was positioned outside the filter options panel. Added a comprehensive operator handoff document for the clinical competency merge project, covering the full runbook for applying changes to production, rollback steps, file map, and queued follow-up work.

Added a new academic free-access tier for domain-verified researchers. Users with institutional email addresses (.edu, .ac.uk, and similar) now receive free read access to the full research section — including trial and paper detail pages and CSV exports from the results view — without requiring a Pro subscription. A Sanity-managed domain allowlist and blocklist give admins fine-grained control over which institutions qualify.

Redesigned the Clinical Guidelines competency explorer for clarity and mobile usability. The explorer now loads collapsed by default with a jump-index for quick navigation, and competency details appear inline directly below the clicked row rather than in a separate side panel — keeping your place in the list without any scrolling. Mobile layout and attribution display have also been corrected.

Improved discovery for Psychedelics Courses on the homepage. Source-backed enrichment now runs more reliably, manual enrichment fields are preserved across runs, and the country filter is properly integrated into search so results stay regionally relevant.

Polished the Results page as part of a Phase 1 pass: fixed scoring direction display for outcomes, added loading skeletons while data is fetching, and improved Forest Plot accessibility for screen readers.

Cleaned up the clinical competency taxonomy by running an embedding-based deduplication pipeline that identified near-duplicate entries across 952 competency documents and merged them into canonical records. Users will see more consistent competency names and reduced redundancy across guidelines and courses.

Added a new psychedelics courses microsite, expanding the site's educational resource offerings.

The /results page now displays paper role badges (primary, secondary, follow-up, pooled analysis, and more) alongside trial design type badges, helping researchers avoid double-counting across study types. The measure view defaults to primary and long-term follow-up papers, with a one-click toggle to include secondary analyses. Pooled baseline values and subgroup breakdown indicators now appear inline on each paper row.

Extended the data schema with several new fields: pooledBaseline on outcome measures captures single-arm pooled baselines reported across all arms, designType on trials tracks study design (parallel, crossover, single-arm, etc.) with crossover-specific washout and period counts, and subgroupBreakdowns records per-measure subgroup results.

Ran an automated Gemini-2.5-Pro backfill to extract baseline values from primary papers, including within-group cascade logic for sequential measures. A heuristic classifier (pass 2) auto-tagged 15 additional papers with their paper roles from titles and abstracts, leaving 14 ambiguous cases flagged for manual review. Added a payer evidence resource to the database.

Fixed a broken production build caused by missing fast-xml-parser and openai package dependencies in the Next.js workspace — both are now properly listed in nextjs-app/package.json.

Wrapped up Phase 5 of the research data quality review: 110 papers republished with 445 individual fixes across baselines, demographics, and outcomes. The full database has now been judged end-to-end.

Tuned the population-misclassified-healthy validator to dramatically cut false positives, and drafted four schema-/workstream-level proposals for residual data issues so we can plan the next pass cleanly.

Phase 5 batch 2 of the data quality review: 26 more papers republished with 119 fixes. Total of 92 papers judged so far in this sweep.

Added a new payer evidence checklist resource to the implementation section, providing structured guidance for building market-access dossiers and supporting reimbursement conversations.

Added a personnel-hours resource for psychedelic therapy implementation, covering staffing requirements and time investments across preparation, session, and integration phases. The resource text was also strengthened with clearer, more actionable language.

Updated the trials forecast screener with regulatory signal indicators and high-signal filtering, making it easier to surface the most relevant clinical trials in the pipeline.

Throttled automated URL recovery and reduced web-discovery frequency for course tracking. Recovery is now skipped if a URL was checked within the last 30 days and abandoned after 90 days of being dead; web discovery runs three days a week instead of daily. This cuts automated web searches from roughly 130 per day down to 4–8 per day.

Completed a full phase-agnostic audit of all 388 trial-linked papers in the database, resolving every flagged issue. Null statistical type labels were repaired automatically from timepoint data across hundreds of arm-timepoint rows, legitimately outcome-free papers (protocols, pharmacokinetic studies, qualitative analyses) were confirmed with written review notes, and three papers with genuine data gaps had their missing outcomes hand-extracted from PDFs — including MDMA MAPP2 self-experience measures, the ESCAPE-TRD esketamine trial, and a late-life TRD ketamine study. The database now has zero unresolved data flags across all trial phases.

Significantly redesigned the trial results page with a cleaner visual hierarchy. A new MeasureHeaderCard consolidates the measure name, full name, score range, scoring direction, clinical thresholds, and observation count into a single card at the top of each measure view, replacing three separate horizontal header bands. The landing grid now features a hero tile for the most-studied measure alongside a properly labeled section heading and directional arrows on secondary tiles.

Improved the compound filter on the results page: Psilocybin, Ketamine, LSD, and MDMA are now pinned as featured compounds at the top of the selector, while comparator-only compounds (Placebo, Ketanserin) and niche substances are hidden. Forest plots now display directional axis labels ("← Better" / "Worse →") on both ends instead of a single caption, and all data table column headers have tooltips explaining their meaning.

Added a paywall teaser that shows one real data row above the access gate for premium measures, giving free and anonymous users a concrete preview of what results data looks like before subscribing.

Fixed a webhook feedback loop that was causing excess Sanity API requests from trial forecast updates, and reduced overall request volume by caching research snapshots daily with normalized fetch cache keys. The trials forecast refresh is now scheduled to run weekly rather than on every content update.

Added the canonical clinical-competency explorer to the Clinical Guidelines section, and prioritized provider-native sources when a recognized organization publishes its own competency framework.

Optimized Sanity history tracking and tagged script requests so we can keep an eye on cost and traffic. Fixed a slug rendering issue on auth and header pages.

Launched the Clinical Guidelines library page and ingestion pipelines, plus an interactive competency skill-tree map so you can browse psychedelic-therapy competencies visually.

Rolled out the broader Sanity optimization and trials-forecast workstream — the database is now leaner, faster, and cheaper to run. Phase D results data also got a quality pass: sub-day timepoints, ghost-row cleanup, and skeleton detection.

Added enrichment for unregistered trials using a local Gemma 4 model — these previously slipped through because they don't have a registry ID, and now they get a forecast and metadata too.

Tuned review severities to skip very old (pre-2010) papers for some checks, and ran a Gemini 2.5 Pro sweep on the remaining warning-level papers.

Hand-mapped demographics for 15 papers from judge PDF quotes, and fixed two systematic bugs (responders > N and cohort-N mismatch) across the database.

Made Vercel installs deterministic with offline preference and added timing-check tooling to keep build cost predictable.

Big day for the research review: Phase 1 data-quality validators, Phase 2 LLM judge (Gemma 4 with Gemini 2.5 Pro escalation), and a safe drafts-only autofix pass for mechanical issues.

Refreshed the courses listings UX with a quality cleanup workflow, and reworked the discovery copy on the papers and trials pages for clearer SEO and a friendlier first impression.

Added report-style figures across the Road to Access chapters, with non-academic citation links so readers can trace claims back to source.

Stakeholder taxonomy: finalized academic core-field normalization, the investor/nonclinical/academic refresh wave, and added Floris's profile and a Blossom upsert script. Renamed the Delphi stakeholder slug with a legacy redirect, and fixed a Research Snapshot crash.

Stakeholder taxonomy overhaul: fixed the hierarchy and descriptions, categorized 253 stakeholders, added breadcrumbs, and rolled out new waves for activists and indigenous voices, government policy, care-delivery practitioners, adjunct consultancy, creative, and wisdom keepers/elders.

Implemented the supplement download pipeline for trial papers, started linking trials to drug developers, and improved Sanity Studio ergonomics. Resolved DOI primary links and sanitized noisy URL parameters.

HeroSearch presets fixed for depressive disorders, pricing preview surfaced, and a Vercel cost optimization rollout to keep build minutes in check. Added admin endpoints for bulk preference enable and a test-Bloom send.

Added care-delivery, creative, and harm-reduction stakeholder profiles with seed data and verification waves. The stakeholder coverage continues to expand toward a comprehensive map of the psychedelic ecosystem.

New Features

Blossom now has RSS feeds. You can subscribe to all new research papers, or to papers within a specific topic, directly in any feed reader — Feedly, Inoreader, or any tool that supports RSS. This means you no longer have to check Blossom manually to stay on top of new publications.

A new Partners page is live, giving partner organisations a single place to find RSS feed URLs, setup instructions, the Blossom brand kit, and API documentation. The page is linked from the footer.

Fixes

The RSS feed had an edge case where missing data could cause the feed to break entirely — that has been fixed and the feeds are now robust against incomplete records. The backend pipeline that pushes new microdosing papers to partner sites also received a security and reliability hardening pass.

New Features

Research Snapshot pages are now live for every topic on Blossom. Each snapshot gives you a visual overview of how much research exists in that area — paper counts, clinical trial numbers, how the field has grown over time, and the key themes emerging from the literature. It is the fastest way to orient yourself when exploring a new topic.

The drug developer database was significantly enriched this week, with detailed pipeline profiles for 19 companies covering 43 drug candidates. Each candidate entry now includes the compound name, target condition, mechanism of action, and current development stage.

A new Foundations & Philanthropy section is now live in the stakeholder directory. This covers the foundations and philanthropic organisations funding research in this space — giving you a fuller picture of who is driving the science, beyond just the researchers and companies.

Search results are now shareable. Whatever filters you have applied, the URL in your browser reflects your exact search state — copy it and send it to your doctor, a friend, or save it for later, and it will open to exactly what you found.

Stakeholder profiles gained a new Manufacturing tab, rounding out coverage of the full pharmaceutical supply chain — from research groups and drug developers through to the companies that manufacture the compounds.

Improvements

Academic institution profiles now surface consortium and network links — so when a university is part of a larger research network or multi-site initiative, you can see those connections directly. Relevant stakeholders also gained real-world data profiles, capturing their experience with observational data outside of controlled trials.

Research article pages now display content in a slightly wider column, making long-form text easier to read. The data cache was also fixed to ensure pages always show the latest content without needing a manual refresh.

Fixes

Several design components that had gone missing from the interface were restored. A bug in the course discovery section where course identifiers could go out of bounds was also resolved, preventing occasional display errors.

New Features

The Bloom newsletter was redesigned from the ground up. It now has distinct sections for different types of updates, separate content for Pro and Free members, and arrives every Monday morning. The layout is cleaner, with a richer intro, highlighted research picks, and a personal note at the end.

Blossom's visibility in search engines improved significantly this week. All major pages now have tailored title tags and meta descriptions, and a Google Search Console integration was added so we can monitor how pages perform in search results. Blossom also now pings Google automatically whenever new content is published, helping new pages get indexed faster.

The drug developer section now contains pipeline data for over 90 companies — covering what compounds they are developing, what conditions they target, and what stage of research or trials each candidate is in. This is some of the most detailed pipeline intelligence we have published to date.

The clinical trial results page gained several new capabilities: you can now view results broken down by time-point (baseline vs. follow-up), compare changes within a trial arm, and export results to CSV for your own analysis. Trial type filters were also added to help narrow down the studies you care about.

Improvements

Over 1,195 research paper abstracts were reformatted with clear section headers (Background, Methods, Results, Conclusions), making it much easier to scan and extract the key findings without reading the full abstract.

The stakeholder directory received a major taxonomy overhaul. All 253 stakeholders were reviewed and recategorised into a cleaner, more logical hierarchy — with breadcrumb navigation added so you can always see where you are and move between related categories with ease.

Blossom now collects VAT for subscribers in the European Union, bringing billing in line with EU tax requirements. Publisher records across the research database were also unified — the same publisher no longer appears under multiple slightly different names.

Fixes

A batch of duplicate blog post redirects were cleaned up. The course directory pipeline was hardened against timeouts that could occur during long-running data updates.

Fixed an issue in the PostHog error analytics pipeline where the majority of tracked errors were being skipped during processing. Error tracking is now more reliable across the platform.

Added an "Experimental" badge to the Results Explorer page and to the Results and Safety tabs on paper detail pages. The badge includes a flask icon and a note that these data points are AI-extracted, helping users understand the confidence level of the information.

Improved dev server startup speed by removing automatic type generation from the predev step (now run manually after schema changes) and adding a 5-minute in-memory cache for the filter options query on the Results page, which performs correlated lookups across papers, measures, and trials.

Clinical trial data enrichment: Ran Phase C & D enrichment pipeline, extracting structured outcome measures, demographics, and adverse event data from trial publications. This powers the new results and safety tabs on paper detail pages.

Browse history: Authenticated users now have automatic browse history tracking. Visit any paper, trial, compound, or topic page and it's recorded. View your recent history from your account page at /account/history.

Bug fixes: Fixed TypeScript error in history action by separating new-profile path from dedup check. Fixed compound filter on /results page.

Research network update: Redesigned the research network page with new filters, search functionality, and UI polish. Upgraded to Next.js 16 with graphology-types dependency fix.

Trial results redesign: Redesigned trial results display with accordion layout, responder rate percentages, and direct measure links. Results and safety data now appear in dedicated tabs on paper pages.

Cross-trial results explorer: Launched /results page — a new cross-trial results explorer that lets you compare outcome data across studies. Filter by measure or dive into a specific trial to see all reported outcomes side by side.

Search UX: Added loading indicator on HeroSearch while results are fetching.

Road to Access: Built the Road to Access implementation hub, with wider layout, populated real content, and dependency upgrades. Library shelf menu click bug fixed.

Researcher profiles: Discovered and created 296 new researcher profiles from paper authorNames, enriched profiles with institutional affiliations and research focus areas, built unified manage-researchers CLI tool, and merged 3 duplicate person records.

Persons search: Launched full persons search page at /persons with HeroSearch, filters by compound and role, and result cards. Fixed score() syntax error and hid date range + smart sort for person results.

Brand & social: Added brand kit page with logo downloads, color palette, and SVG mark. Updated footer social links to BlossomAct profiles.

Stakeholder cleanup: Built stakeholder deduplication & cleanup script, applied review queue (40 deletions, 21 merges, 2 renames), restored 1,047 lost paperPdf links after a draft-merge incident.

Infrastructure: Implemented 3-level automated backup strategy (Sanity + Clerk + GitHub releases). Vercel cost optimization with ISR caching, email improvements, and webhook filtering. Reduced Sanity API/CDN requests with caching and deduplication. Added CGI-S and SHAPS measures with aliases for QIDS-SR-16 and STAI-S/T.

Topic hierarchy enforcement & retroactive tagging: Fixed parent-child topic hierarchy (3 corrections, 2 topic merges), propagated parent topics to 461 papers/trials, keyword-classified 402 papers, published 3 draft-only trials, cleaned 1,240 stale drafts. New Studio enforcement component auto-adds parent topics when editors select child topics. 4 new migration scripts, 2 slug redirects, full code review with fixes.

Paper ingestion improvements: Added PubMed PMID fallback and secondaryUrl field for papers, APA-style citation generation, and abstract fallback for paragraph summaries. Phase B added inline structured citation extraction with citedPapers linking, plus a backfill-citations script for existing papers.

Topic hierarchy & classification: Fixed parent-child topic hierarchy (merged PPD → Peripartum and Neurodegeneration → Neurocognitive), built parent propagation and paper classification scripts (keyword + LLM tiers), and added a Studio auto-enforcement component that ensures parent topics are always applied when child topics are selected.

Auth & security: Added middleware with single-session enforcement to harden Pro accounts. Added /zoom and /calendly shortlink redirects.

UI fixes: Resolved content-visibility:auto shadow clipping on paper and trial cards, fixed SEO meta title input crash, fixed dualPatch crash on draft-only papers, added 'Back to journals' link on journal pages, updated sidebar parent topic pills to white/gray style.

Paper & trial results display — Rebuilt the results sections on /papers/[slug] and /trials/[slug] with structured tabs, bar charts (recharts), and demographics tables. Outcomes, adverse events, trial arms, and responder data now display with visual hierarchy. 8 code review findings resolved post-PR. (PR #64)

Pro search filters — Added Pro-only paper filters (journal, author, stakeholder, study design) and trial filters (stakeholder, investigator, country, generation, randomized, psychotherapy) with ANY/ALL operator toggle for multi-value combinations. Fixed GROQ injection vulnerability in AND operator. Extracted shared types, deduplicated filter UI components, typed API response. Fixed format reset mismatch and missing operator reset on filter clear.

Author data quality — Deduplication pass, fixed broken profile links, and corrected ordering across 2,990 papers.

Research Collections improvements — Added static params and loading skeleton for faster perceived load. Added /research link to Pro onboarding email.

Atlas mobile fix — Map on /atlas now correctly visible on mobile devices.

Studio improvements — Linked papers now shown on trial overview in Studio. Restricted Studio search to title field for papers and trials. Renamed STUDIO_BASE env var to TRIALS/PAPERS_STUDIO_BASE. Added trial compound backfill script. Fixed paper ingestion webhook loop (skip dispatch during active processing). Email notification sent after paper ingestion completes.

Pro custom request board — Pro users can now commission bespoke research and tools from the account dashboard. Requests include budget range (€500–€3,000+) and visibility preference (public good vs private tool). Tagged as custom request type for internal routing.

Vercel cron + QUICK_BUILD — Added vercel.json to repo root for cron job registration. Added QUICK_BUILD=true env toggle for ~4min builds that skip static pre-rendering (pages render on first request instead).

Email & link fixes — Fixed broken anxiety-disorder link in emails. Added slug-level fuzzy redirect for detail pages with mismatched legacy slugs. Fixed Sanity webhook HMAC signature verification (Stripe-style format). Fixed null person name coalesce in filterOptionsQuery.

Phase C enrichment pipeline — Built inventory script, extraction library, and CLI orchestrator for extracting demographics and safety data (age, sex, drop-out rates, adverse events) from research papers. Schema changes: new phaseC fields on paper document. Studio desk filters for enrichment status. PostHog config for extraction tracking.

Phase D outcomes extraction pipeline — Built outcomes-enricher.ts with 2 LLM calls per paper (measures+timepoints, then contrasts+responders). Tested on 8 ibogaine + ayahuasca papers: 66 outcome measures, 197 timepoints, 25 contrasts, 39 responder outcomes extracted. Created 7 new measures (CADSS, HRS, COWS, SARA, D-KEFS, WAIS-IV PSI, CPT-3). All processed papers set to phaseD: needs-review.

Schema hardening — Added top-level compounds[] reference on trial, no_intervention arm category, adjunctMedications[] structured array on arms, personalPsychosisExclusion boolean, Unknown therapy orientation, isRegistered default true. Updated ~30 field descriptions across paper.ts and trial.ts. Added warning validation on relatedTrials. Hidden fullSummary and metaAnalysisSources (when not meta). Fixed stale-mutex bug in Phase C/D needsProcessing(). Stripped hardcoded Sanity write tokens from 4 legacy scripts.

Bug fixes — Mounted SanityLive in root layout to enable live revalidation. Fixed PostHog defaults date in instrumentation-client. Fixed null type in StakeholderLocationMap Location type.

Admin dashboard — Added admin-only dashboard at /admin with 6 service tabs: Papers, Trials, Email, Clerk, Stripe, and Logs. URL-param tab switching with useTransition for a stable tab bar during navigation. Security hardening: admin-email enforcement, no metadata leak, rate limiting on sensitive endpoints. (4 files)

Stakeholder enrichment complete — Published 186 stakeholders in Sanity. Merged Lykos Biosciences into Resilient Pharmaceuticals, cleaned up stale entries. All stakeholder category pages now correctly use numeric L1/L2/L3 level values.

Stakeholder category pages review — Resolved 9 P2s + 8 P3s: fixed leadStage alphabetical sort, meta robots in App Router, DRUG_STAGE_LABELS divergence across 3 files, added category pages to sitemap, SearchScope deduplication, JSON-LD improvements, OG images, type safety.

Research Data API: stakeholder endpoints + sponsor filter — Added /api/v1/stakeholders list and /api/v1/stakeholders/[slug] detail endpoints with cursor pagination and sponsoredTrials/collaboratedTrials projections. Added ?sponsor= filter to /api/v1/trials. (5 files)

Sponsor-linking pipeline: 18 code review fixes — P1s: drafts. prefix stripping on all _ref values, unmatchedMap preserves original rawName, MAPS PBC alias corrected to Resilient Pharmaceuticals, FUZZY_SCORE_THRESHOLD=0.7 gate. P2s: JSON schema validation before Sanity writes, MatchType discriminated union, dualPatch draft errors logged as warn, TrialSponsorPatch typed interface, --dry-run flag, collaboratedTrials GROQ fix. P3s: Phase 4 integration hooks, normalization idempotency.

Stakeholder hub & Drug Developers pipeline page — Rebuilt /stakeholders overview with live full-text search scoped to organisations and a browsable L1 category grid. Added /drug-developers pipeline page showing companies grouped by clinical-development stage (Discovery to Approved) with a visual pipeline overview bar. Visiting the Company-led Drug Development category now redirects to /drug-developers. Fixed critical level bug: stakeholderReport.level is stored as a number. Build: 1,319 static pages. (6 modified, 1 new)

Code review: evening-9mar-stakeholders — Multi-agent review with 9 specialists. Created 22 structured todos: 3 P1 blockers (leadStage GROQ alphabetical sort, meta robots tag ignored in App Router, DRUG_STAGE_LABELS diverged across 3 files), 9 P2 items (category pages missing from sitemap, SearchScope duplication, sort allowlist), 10 P3 items (GROQ optimisations, JSON-LD, OG images, type safety). Resolved all 3 P1s and 8 actionable P3s. Documented 3 GROQ/Sanity patterns in docs/solutions/.

OG image redesign, header/footer polish, person page dedup — Redesigned all 18+ OG image templates with new layout and white text overlay. Updated global header and footer styles. Deduplicated person page component logic. (8 files)

Journal pages redesign — Rebuilt /journals index (card grid with H-index, paper count, country, open-access badge) and /journals/[slug] detail page (published papers list, metadata sidebar, country badge, external link). Design-system compliant.

Measures section — Created 43 measures in Sanity across 9 categories (mood/affect, anxiety, dissociation, mystical experience, cognitive, personality, quality of life, sleep, neuroimaging) with full metadata: mechanism, instruments, scoring guide, reliability (Cronbach alpha, ICC), construct validity, sensitivity to change. Rebuilt /measures index and /measures/[slug] detail pages with sortable OutcomeTable data grid. (4 new files, 6 modified)

Stakeholder detail page upgrade — Added drugCandidates, foundedYear, ticker, website fields to stakeholder schema. Rebuilt detail page with context-sensitive layout for company, academic, hospital, and government stakeholder types.

Code review: type safety & accessibility — Fixed decorative icons (orange to gray) on stakeholder page, removed as-any casts on stakeholder/journals pages, re-extracted schema and regenerated sanity.types.ts (84 types), added useMemo + keyboard navigation + caption to OutcomeTable, aria-hidden on decorative SVGs, widened TrialCard/PaperCardRef/EventCard prop types to accept null. Removed development/plan.md. Build: 1180 pages. (9 files)

Research Data API code review — Deduplicated CSV serializer and 5 shared flattener functions, fixed cursor pagination bug on 3 endpoints (compounds, topics, journals), parallelized all Vercel Blob uploads via Promise.all, added rate limiter memory cleanup, fixed toJsonExport non-array handling, replaced new Blob().size with Buffer.byteLength. (12 files, net −35 lines)

Research Data API — REST API at /api/v1/ with 11 endpoints: papers, trials, compounds, topics, persons, journals, network, evidence-matrix, and downloads. Clerk JWT auth, per-user rate limiting (100 req/min free / 500 Pro), CORS, cursor-based pagination with compound/topic/date filters. Free/Pro data split. CSV format on all endpoints with R/SPSS conventions. Daily cron exports CSV + JSON + Parquet to Vercel Blob. Developer docs at /developers with code examples for curl, Python, and R. (26 new files)

Research Network — Interactive co-authorship graph at /network powered by Sigma.js + ForceAtlas2 layout. Hover highlighting, click-to-detail side panel, search-to-zoom, compound/topic filter pills, cluster-by coloring. Paywall gated: free users see top 50 nodes, Pro gets full graph and advanced filters. Ego network graphs on researcher profiles with Top Collaborators sidebar. PostHog tracking, ARIA labels, mobile-responsive. (10 new files)

Research Collections — Compound x Topic pages at /research/[compound]/[topic] with paginated paper and trial lists, empty-state fallback, and Explore further links. Index at /research lists all combinations grouped by compound. Pro curated combos in account dashboard. Sitemap entries for all compound x topic pairs. OG images per combo. (4 new files)

Person page enhancements & OG images — Removed avatar images from researcher profiles; added Compounds and Topics pill sections in sidebar (auto-derived via GROQ inverse lookups). Added dynamic OG images across 18 route segments: persons, compounds, topics, events, stakeholders, countries, blog, books, courses, courses/category, implementation, journals, measures, news, posts, recap, top-10, and videos.

Newsletter Signup Hardening: Fixed 12 code review findings across the newsletter signup system. Critical fix: Resend contact updates were silently failing because audienceId was omitted from the SDK call. Security: added email validation, rate limiting (5 req/min/IP), and uniform API responses to prevent email enumeration. Performance: replaced return-null loading states with skeleton placeholders to eliminate CLS, lazy-loaded below-fold newsletter bundles via next/dynamic, and added PostHog onFeatureFlags idempotency guard to prevent duplicate analytics events. Code quality: extracted shared useNewsletterSubscribe hook eliminating ~57 lines of duplicated form logic across 3 components. 4 new files, 9 modified.

Countries & persons pages redesign: rebuilt all country pages (USA, UK, Netherlands, etc.) and stakeholder/person profile pages with the grayscale palette, paywall integration, and consistent layout — completing the platform-wide design overhaul.

Topic & compound overview images: added 24 custom overview images across topic and compound pages (depression, anxiety, psilocybin, ketamine, and more). Each page now features a relevant hero image for better visual context.

Roadmap & feedback board audit: updated 98 feedback cards to reflect current development status, created 6 new cards for recently built features (PDF viewer, citation networks, library notifications, multi-currency pricing), and added guidance text to the roadmap page explaining how to submit requests and vote.

Compound Report Content (8 profiles): Populated full editorial content across all 9 sections (history, pharmacology, safety profile, key trials, clinical outlook, regulatory status, commercial outlook, commercial outlook details, comparative context) for DMT, 5-MeO-DMT, Ibogaine, Ketamine, Mescaline, Salvia Divinorum, Nitrous Oxide, and Placebo. Placebo received substantive methodological treatment covering the history of RCTs in psychedelic research, active placebo strategies (niacin, diphenhydramine, low-dose compound), blinding integrity measurement, landmark trials (MAPS MDMA Phase 3, Imperial escitalopram comparison, COMPASS Phase 2b), and the FDA's 2024 MDMA advisory committee decision and its implications for the field.

Overview Images (23 added): Added compound overview images for 2C-x, 5-MeO-DMT, DMT, Esketamine, LSD, Mescaline, Placebo, and country overview images for Jamaica, Japan, Jordan, Kazakhstan, Lithuania, Malaysia, Mexico, Monaco, Netherlands, New Zealand, Pakistan, Singapore, Slovakia, South Africa, South Korea, Sweden, and Taiwan. Removed outdated the-netherlands.png duplicate.

Welcome Email Sequences: Built complete drip email engine with 7 user segments — free signup welcome (15-min delay), legacy user re-engagement (cohorted 250/day rollout), Pro Individual onboarding, Pro Team onboarding, Pro Enterprise onboarding, team/enterprise member welcome, and subscription renewal reminders (30d, 7d, expiry). Includes declarative sequence definitions, Clerk privateMetadata state tracking, hourly cron processor, and Stripe webhook transitions for free-to-pro upgrades with expiresAt computation.

Tiered paywall system: introduced a flexible paywall for the papers search page. Free users see a limited number of results, while subscribers get full access. Paywall filters, a refined search button, and header improvements round out the UI.

Topics pages redesign: completely rebuilt all topic pages (depression, anxiety, PTSD, etc.) with a grayscale palette, cleaner section layouts, paywall integration, and a unified design language across the platform.

Compounds pages redesign: rebuilt all compound pages (psilocybin, ketamine, MDMA, etc.) with the same grayscale palette and layout overhaul as topics — consistent cards, section headers, and paywall treatment.

Paper & trial detail page improvements: added paywall locks and UI polish to individual paper and clinical trial pages, ensuring a consistent reading experience across the platform.

Homepage redesign and search UI refinements: updated the homepage layout, polished search filter appearance, and improved the global search button styling for a more cohesive first impression.

Courses platform enrichment: added richer course metadata (certificates, scholarships, accreditation detection), built a filter bar for the courses listing page, and created automation scripts for semi-automated course imports from external training providers.

Multi-currency course pricing: courses now display prices in the user's local currency (EUR, USD, GBP, and more) using geo-aware detection. Prices are converted automatically so users see costs in a familiar format.

Course detail page redesign: rebuilt /courses/[slug] pages to match the papers page design conventions — cleaner layout, consistent typography, and improved information hierarchy.

Paper Detail Page v2: Rebuilt /papers/[slug] with a full research layout — interactive citation network sidebar, in-page access locks, PDF section summaries (Introduction, Methods, Results, Discussion, Conclusion), and strict gray-palette design compliance.

Codebase health sweep (50+ files, 12 categories): removed console.log statements from webhook and search routes, removed PostHog window.posthog assignment, created client-component lazy wrappers for 3 map components to fix Next.js 16 Turbopack ssr:false restriction, added lucide-react wildcard TypeScript declaration, resolved typedRoutes and baseline-browser-mapping build warnings, slimmed CLAUDE.md with modular .claude/rules/, added stale-worktree auto-cleanup hook. Build passes at 4,107 static pages.

GROQ query optimisation: added slim paperCardFields (9 fields) and trialCardFields (11 fields) projections for category listing pages, reducing Sanity response size from ~37 MB to under 2 MB and enabling Next.js fetch-cache for faster ISR revalidation.

Component composition refactor: split SearchResultCard into 3 explicit named variant components — EventSearchCard (calendar-widget layout), EntitySearchCard (profile card for stakeholders/compounds/topics), and ResearchSearchCard (standard paper/trial card). The exported dispatcher is now ~15 lines.

Atlas of Psychedelic Research enhancements (Feb 26): added recruitment-status filter, deduplicated overlapping trial map markers, added acknowledgements section for data sources, and refined research-centre groupings.

We've integrated a sophisticated research pipeline that automatically processes thousands of scientific PDFs. Blossom now extracts and presents key findings, experimental methods, and core conclusions directly on the paper pages. This AI-powered layer helps you grasp the essence of complex research at a glance, starting with thousands of open-access papers in our database.

SEO & AEO Infrastructure: Deployed unified SEO infrastructure across all content types. Added a reusable seo object type (metaTitle, metaDescription, ogImage with live character counter) to all 20 public-facing Sanity document types, replacing fragmented ad-hoc fields. Created shared buildMetadata() utility and JSON-LD structured data builders for 10 Schema.org types (ScholarlyArticle, MedicalStudy, Person, Event, Organization, BlogPosting, Book, Course, FAQPage, BreadcrumbList). All 16+ detail pages now export generateMetadata() and include structured data scripts. Sitemap expanded from 5 to 17 document types (~4,478 URLs). Added robots.ts and fixed metadataBase. Build verified: 4,009 static pages, 0 errors. PR #15 created.

Researcher Profiles: Enriched the person/researcher schema with professional title, importance/key contributions, biography (rich text), institutional affiliations (linked to stakeholders), specialization tags, and social/professional links (Google Scholar, LinkedIn, personal website). Redesigned persons index and detail pages to follow design guidelines: gray-scale palette, standard card patterns, orange only for hover/CTA states. Sanity schema deployed to Content Lake.

Today's update significantly enriches the data behind our researcher and expert pages. We've added comprehensive professional biographies, institutional affiliations, and direct links to academic and social profiles. We also completed a major SEO overhaul, ensuring Blossom's deep research database stays visible and accessible across the web.

We've upgraded our Interactive Research Map with detailed facility-level coordinates for over 900 clinical trial sites globally. Instead of just seeing the headquarters of trial sponsors, you can now see the actual hospitals and research centers where the work is happening, from leading universities in the UK to clinics across Australia.

Migrated 10 legacy "Top" lists via automated URL-parsing and fuzzy matching script. Created /top-10 overview hub and dynamic frontend routes securely utilizing Sanity NumberedPaperItem blocks.

A massive update day focusing on global reach and user engagement. We launched Country Research Hubs, enabling users to explore psychedelic research activity across 100+ countries with localized stats and interactive maps. We also automated our Weekly Newsletter workflow using Resend and implemented fine-grained notification preferences in the user dashboard. Finally, we synchronized our user database with Resend via Clerk webhooks for seamless communication.

Today's updates focused on refining our public-facing roadmap and advancing our automated data systems. We resolved a visibility issue in the Roadmap, bringing 170+ community requests back into view, and added premium sideways labels to improve column legibility. On the backend, we launched a Hybrid Event enrichment workflow, combining agentic AI research with human-in-the-loop validation for event metadata. Finally, we updated the Sanity ecosystem to v5.5.0 and migrated to modern CLI-based typegen for a smoother developer experience.

Today we significantly improved the events search and exploration experience. Key updates include the implementation of 'Smart Date' sorting which prioritizes upcoming events while keeping past events accessible, and the integration of a dynamic map that syncs with search results. We also added new filters for location (US/EU) and time (Future/Past), standardized the filter hierarchy, and updated the visual design with Blossom's signature orange hues. Additionally, descriptions were added for 12 new events to enrich the database content.

Happy New Year! We're kicking off 2026 with a massive series of updates to how research is organized and verified in Blossom.

🚀 Key Highlights

Category Pages: You can now browse papers and trials by Compound or Topic directly. Using our new polymorphic routing, URLs like /papers/psilocybin or /trials/mdma naturally display filtered research without needing to search.

Paper Verification Workspace: We've launched a new internal tool to help our auditors classify research into 6 population tiers: Clinical Trial, Observational-Clinical, Non-Clinical Patient, Healthy, Other, and Unverified. This ensures the highest data quality for your discovery journey.

🛠 Experimental & Internal Updates

Enhanced Trials Workflow: Improved tracking of linked papers and clinical verification flags.

Automated Sync Protocol: Our AI agents are now equipped with a unified synchronization protocol to keep the Roadmap, Changelog, and Todo list perfectly aligned after every update.

Radiant Design Refinements: Extracted and refactored reusable Paper and Trial cards to ensure a consistent, premium aesthetic across all dynamic views.

We kicked off the year with several major enhancements to Blossom. We introduced automated categorization for all research papers and trials by Compound and Topic, making it easier than ever to dive into specific areas of research. We also launched a new Paper Verification Workspace for clinical classification and standardized our AI workflows to ensure your roadmap and project updates remain perfectly in sync.

Public Roadmap: Over 140 new cards added to our public facing roadmap/feedback system.

Released User Library: Save, rate, and comment on papers and trials.

Shelves: Organize your saved research into custom collections.

Personalized Search: The homepage now greets you by name.

SEO Overhaul: Rich social sharing cards and structured data.

Account Dashboard: New billing history and roadmap tracking.

Search Refresh: Full-width results layout with streamlined navigation.

Advanced Filters: New granular filters for Topics, Compounds, and Study Characteristics.

Public Roadmap: Launched our initial feedback board for users to suggest features.

Performance: Resolved build bottlenecks and optimized image loading.

Pricing & Checkout: Launched our premium tiers with Stripe integration.

Interactive Roadmap UI: Visual tracking for upcoming site features.

Refined Aesthetics: Updated various UI components to our new premium design system.

Data Import: Imported thousands of peer-reviewed papers and clinical trials.

Core Experience: Built the search engine and individual detail pages.

Site Infrastructure: Implemented base CMS structure and global navigation.