consciousnesslibrary

What it is in one line

A free, open library that pulls the world's psychedelic and consciousness research into one place: every 20 minutes it ingests new papers from a dozen-plus academic APIs, dedupes them, organizes them by topic, writes plain-language summaries, and can synthesize evidence across the whole corpus — while correctly keeping "LSD the psychedelic" apart from Lumpy Skin Disease.

Who built it

Posted on HN (Show HN, 30 points / 12 comments). The author writes that he had just finished a master's in Psychedelics and Consciousness Studies, built the library he wanted while studying, shelved it for a year, then picked it back up after his practicum advisor pushed him to go hard on it. The project is registered as a 501(c)(3), legal entity Mindscape Collective, self-funded so far. The pool lists the builder as elliotec.

Read: a library that "reads the whole field" can only be built by someone who actually read the field for a degree and uses it daily — the barrier is not technical, it is knowing what to collect and what "complete" means. This author has both.

What it actually does

• Aggregates and dedupes → pulls from a dozen-plus academic APIs (PubMed, OpenAlex, Europe PMC, bioRxiv, more) every 20 minutes, automatically deduplicated
• Relevance judging → a two-stage pipeline of cheap keyword prefilter plus an LLM rubric that accepts or rejects papers, resolving the hard cases: LSD the psychedelic vs Lumpy Skin Disease, ketamine for depression vs anesthesia for cats
• Plain-language summaries → every paper gets a "Study at a glance": design, sample size, key findings, in plain language
• Topic organization → 35,787 papers across 37 topics grouped under psychedelics, practices (meditation), experience, humanities, neuroscience
• Evidence synthesis → topic pages carry evidence overviews, and you can ask across the whole corpus; the AI synthesizes consensus, conflicts, and unknowns
• Clinical trial tracking → refreshed daily from ClinicalTrials.gov: recruitment status, phase, enrollment, sponsor
• A map of the whole corpus → one dot per paper, so the shape of the field is visible at a glance

What it deliberately does not do: no commercial search platform, no ads, no login required to read; free-text synthesis needs a free account.

What old behavior it replaces

Researching psychedelics used to mean searching PubMed, bioRxiv, and ClinicalTrials.gov separately, manually deduping, and manually reading dozens of abstracts to find consensus and disagreement. A literature review per topic ran on a weekly timescale.

This library replaces three manual chores: cross-database search (one query covers everything), reading abstracts to find consensus (synthesis names what agrees and what is still contested), and tracking clinical trials (daily refresh instead of weekly manual checking).

Business model

Non-commercial. A 501(c)(3) nonprofit: free, ad-free, readable without an account, funded by donations (Stripe, PayPal, Buy Me a Coffee) plus the author's own money.

Read: this is not a business, it is infrastructure. But "build the complete corpus of a narrow field as a free public good at near-zero cost" is itself a signal for anyone making vertical AI products — the asset pays off elsewhere (influence, data partnerships, foundation support), not in subscription fees.

Hard numbers

• 35,787 papers / 82,200+ authors / 37 topics, growing daily
• Infrastructure: Rails 8.1, a single Postgres (full-text, vectors, and job queue all in one), a $24/month DigitalOcean droplet (just upgraded from the $12 tier), deployed with Kamal
• Total LLM spend to date: ~$28 (DeepSeek-V4-Flash), trending down after a one-time backfill push
• Ingestion cadence: polls a dozen-plus academic APIs every 20 minutes
• Team: essentially one person (the site lists a few collaborator avatars); revenue, funding: none (donation-based)

Four-way read

The call

A template for "turn a very narrow field into a complete corpus at near-zero cost." Low commercial value, high mechanism value.

The most stealable part is the relevance pipeline: cheap keyword prefilter first, LLM rubric accept/reject second. It sounds plain, but the author says relevance was by far the hardest part — generic search fails here because Lumpy Skin Disease also contains "LSD," and vector retrieval cannot tell the semantic camps apart. The two-stage pipeline brought the whole corpus's LLM cost down to $28, and it transfers to any vertical corpus.

The one thing it did right was registering as a nonprofit instead of a company. Trust barriers are high for narrow-field public libraries, and "we don't sell data, we don't sell ads" is part of the content.

The limits are obvious too: solo operation, donation funding, no public retention metrics. The biggest risk in this project is not the product, it is the maintainer's energy — once the author lands a job, update cadence will drop.

What to watch next

① Whether paper count keeps growing daily and the 20-minute cadence actually holds — is maintenance alive ② Whether registered users and synthesis-query usage ever go public — real use, or self-indulgence ③ Whether institutional partners appear (universities / foundations / journals) — the hinge between a personal project and public infrastructure

What you can take from it

Product logic: any vertical can copy the trio of "complete corpus + plain summaries + evidence synthesis." The key move is defining "complete" first — subscribe to a dozen APIs and poll every 20 minutes; completeness beats retrieval quality as the initial claim.

Pricing structure: none. But "nonprofit + donations + self-funding" is a viable structure for narrow-field tools that need public credibility.

Verdict

Worth watching — as a mechanism sample, not an investment. Complete corpus, near-zero cost ($24/month infra plus $28 of LLM spend), and a two-stage relevance pipeline that transfers directly. It proves one person, one $24 server, and a few dozen dollars of LLM calls can build a field's public infrastructure. Don't expect a business; do copy the mechanism.