Crate Digger
Finds music through context, not popularity alone.
Crate Digger is a trust-first music discovery experiment built by Darwin Hernandez. It turns expressive briefs like "nocturnal, elegant, reverb-drenched" into a calibrated listening path rather than a generic list of familiar albums.
The listener chooses what must survive the interpretation, how far to travel across genres, and how visible the artists should be. Crate Digger then researches candidates, verifies album identities, and presents one album at a time with an inspectable reason for why it belongs.
Music discovery should explain why, not just what.
Most recommendation systems optimize for the next likely click. Crate Digger starts with the context behind a listener's curiosity.
A prompt can contain mood, texture, era, scene, energy, and a reference artist at the same time. Flattening that into one similarity score loses the meaning of the request. Asking a language model directly can produce compelling answers, but confidence often arrives without visible evidence.
Crate Digger treats discovery as a guided interpretation. It calibrates the request, makes the evidence boundary visible, and returns an honest coverage state when nothing is supported strongly enough.
A good recommendation should show its path, not just name a destination.
Context-led discovery
Translates expressive briefs into a calibrated path across interpretation, genre distance, and artist visibility.
Provenance as UX
Separates verified identity, curator judgment, third-party signals, and model inference instead of blending them into one confident answer.
Honest failure states
Reports no match, known but unsupported, or unknown when the available evidence cannot support a useful recommendation.
Taste is not a prompt-completion problem.
Crate digging has always been about context: labels, scenes, collaborators, liner notes, adjacent movements, and the strange bridge between two albums.
Digital recommendation often compresses that texture into popularity, behavioral similarity, or a black-box answer. Darwin built Crate Digger to test whether AI could help recover the path-making part of discovery without pretending that a model's interpretation is verified truth.
The product is deliberately more listening room than infinite feed. It slows the experience down, presents one album at a time, and keeps the reasoning close enough for the listener to inspect or reject.
Listeners who want a trail to follow, not another pile of content.
Crate Digger is for curious listeners who can describe a feeling, scene, reference, or listening situation but do not want popularity to decide where the search ends.
The curious listener
Knows the feeling or sonic quality they want, but not the album, artist, or genre name that gets them there.
The crate digger
Wants overlooked albums and credible bridges across labels, scenes, eras, and musical lineages.
The skeptical explorer
Likes AI-assisted discovery but wants to know which parts are verified, curated, sourced, or inferred.
The product-minded visitor
Sees how transparency, constraints, calibration, and failure states can turn an AI answer into a more trustworthy product.
It turns a mood-rich brief into an inspectable listening decision.
Discovery distance and artist obscurity are different choices.
Moves beyond the popularity loop
Artist visibility is an explicit control, so underheard music can surface without forcing the listener into a completely different genre.
Preserves what the listener actually meant
Request-specific calibration separates the quality that must survive from how far the recommendation should travel.
Makes evidence inspectable
Each result carries reasons, confidence, and source labels, with album identity verified independently through MusicBrainz.
Refuses the confident guess
When coverage or evidence is insufficient, the system says so instead of filling the page with plausible-looking recommendations.
A recommendation workflow with trust boundaries built in.
Crate Digger combines human-curated anchors with controlled interpretation, external candidate research, identity verification, faithful explanations, and explicit coverage states. The anchors guide the system without becoming a closed recommendation inventory.
Listening brief
The listener begins with a mood, scene, image, reference artist, listening situation, or open-ended prompt.
Request calibration
Adaptive questions identify the quality that must survive, genre distance, and desired artist visibility.
Curated anchors
A human-maintained reference library and controlled vocabulary provide calibration evidence and stable interpretive concepts.
Candidate discovery
Optional model-guided discovery proposes outside albums without allowing model recall to become verified fact.
Research pass
The system looks for support for the calibrated bridge and rejects candidates whose contextual connection is too weak.
Identity verification
MusicBrainz confirms album and artist identity; it does not certify mood, meaning, or musical fit.
Evidence and confidence
Reasons remain labeled as curator, MusicBrainz, third-party, or model-inferred evidence with numeric confidence.
Listening path
Recommendations appear one at a time with the reason, evidence trail, listening profile, and routes for deeper exploration.
Evaluation loop
Anonymous journey feedback and a frozen evaluation harness support learning without pretending early signals are comparative proof.
Product positioning, trust architecture, and working software.
Darwin Hernandez shaped Crate Digger across product concept, positioning, listener experience, recommendation policy, curation strategy, evaluation design, and implementation.
He defined the core product choices: context before popularity, independent controls for genre distance and artist visibility, one-album-at-a-time discovery, visible confidence, and a clear line between identity verification and musical interpretation.
He also designed the credibility rules around provenance, controlled vocabulary, MusicBrainz verification, data rights, secret handling, public-mode restrictions, and honest coverage states.
On the build side, Darwin created the Python and FastAPI application, deterministic data pipeline, SQLite data model, listener and evaluation rooms, test suite, public feedback flow, and Render deployment configuration.
The recommendation is only as trustworthy as the path behind it.
Crate Digger treats confidence, provenance, and failure states as parts of the product experience rather than technical details hidden behind the answer.
The recommendation logic can travel wherever taste is hard to describe.
Crate Digger's transferable pattern is not music-specific: interpret an expressive request, calibrate what must survive, separate familiarity from distance, verify what can be verified, and explain why each result fits.
Films and series
Turn requests like “quiet, unsettling, visually precise” into recommendations shaped by mood, pacing, era, visual texture, and familiarity, then show the evidence behind each match.
Places and cultural experiences
Help someone find a gallery, neighbourhood walk, live event, restaurant, or other experience by translating atmosphere, energy, social context, and willingness to explore into an explainable path.
Crate Digger at a glance.
What is Crate Digger?
Crate Digger is a trust-first music discovery experiment built by Darwin Hernandez. It turns expressive listening briefs into calibrated recommendations with visible evidence and confidence boundaries.
How is it different from a streaming recommendation feed?
Crate Digger begins with context and listener intent rather than engagement history. It separates genre distance from artist visibility and explains the supported path behind each recommendation.
Does it treat AI output as fact?
No. MusicBrainz verifies album identity. Human curation, third-party signals, and model inference remain distinct sources, and musical fit is labeled as interpretation rather than verified fact.
Is it finished?
No. Crate Digger is in its MVP phase. The core discovery flow, listener interface, and evidence model are working while real-world testing and product refinement continue.
Can AI discovery feel more like a trusted guide than a popularity engine?
That is the bigger question Darwin Hernandez is testing with Crate Digger.