If you're using dydact to apply existing structural intelligence to known classes, it's a transaction. If you're using dydact to find something that didn't exist before, dydact is a stakeholder in what comes out. Not a licensing fee. An equity + IP-sharing partnership. Non-negotiable entry condition for discovery-class endpoints.
Apply the trained v0.3 operator to known compound classes. Fast, cheap, no IP claim on output. Most use cases live here.
POST /api/v1/predict/tcPOST /api/v1/predict/tc/sweepPOST /api/v1/holdout/runGET /api/v1/calibration/:basisGET /api/v1/modelsDiscovery endpoints run void-boundary inversion and candidate synthesis — they find compounds and configurations that didn't exist in training. Access is not free and not contingent on outcome. You pay a retainer for the privilege of running the substrate, and dydact holds equity + IP rights in anything that ships. Both terms apply independently.
Hypothetical discovery has zero value to dydact. Real substrate time has real cost — bare-metal GPU bursts, calibration maintenance, operator support, schema auditing. The retainer funds those. The equity captures upside when real discoveries emerge. dydact is not a charity experiment stand.
POST /api/v1/oracle/synthesizePOST /api/v1/candidates/rankPOST /api/v1/training/initBlocked until retainer cleared + agreement signed.
Annual retainer or per-engagement fee. Funds substrate time, compute bursts, operator support, calibration maintenance. Scaled by vertical: pharma ≫ semis > advanced materials > other. Paid regardless of whether the engagement produces a commercial discovery.
No retainer → no access. Quoted per engagement after intake call.
The retainer filters. Serious engagements commit real budget before first call; tire-kickers self-select out. It also funds the real cost of running discovery — bare-metal GPU bursts, operator time, calibration runs, response schema auditing — which are not free to dydact regardless of whether your experiment converges.
The equity captures upside the retainer can't price at agreement time. A 2% royalty on a breakthrough caps at single-digit millions a year for a bounded patent window. A 33% stake compounds indefinitely across every product, every next-gen iteration, every adjacent market the venture expands into.
We keep the spigot. Post-discovery, optimizations and extensions still run through dydact's operator. The partner keeps the discovery; dydact keeps the ongoing refinement channel. That alignment is why the structure holds long-term.
We don't chase every field. Discovery partnerships concentrate on verticals where the geometric substrate has a structural advantage over existing methods and where the IP value justifies the partnership model.
Tc prediction + novel-compound synthesis across BCS / iron-pnictide / cuprate / hydride regimes. Paper validates the methodology.
Target-binding affinity + molecular candidate generation under physiological conditions. Requires HIPAA + pharma GMP compliance overlay.
Novel semiconductor and qubit-class material candidates. Cross-modal substrate (molecular × cad × reaction_conditions).
High-temperature alloys, metamaterials, catalysts. Corpus expansion in progress for cond-mat scope.
Pharma trial candidate shortlisting with confabulation-rated confidence. Enterprise tier + GMP overlay.
Discovery work in a field we haven't opened yet? Email access@dydact.io with what you're attacking and why the geometric substrate applies.
Regulatory surface is orthogonal to prediction-vs-discovery. A pharma discovery partnership stacks HIPAA + GMP on top of the discovery agreement; a materials prediction corporate contract might only need GDPR + SOC 2.
EU residency (default — substrate runs in EU jurisdiction). Right-to-erasure on non-anonymized inputs.
Annual attestation. Access control reviews + audit retention. Enterprise add-on.
BAA + PHI firewall + encrypted-at-rest everything. Shared compliance stack with scraiv.io. Discovery-tier pharma partnerships.
ISMS alignment. Enterprise default.
Production readiness for regulatory submissions. Contracted per engagement.
US-jurisdiction substrate + air-gapped shard. On-request.
One key per department per university. 1,000 predictions/day default. Hold-out endpoints for paper reproduction. Cite the dydact paper when you publish; we track paper count as the primary academic-tier metric.
Academics DO NOT get discovery-endpoint access by default — the IP structure there is built for commercial partnerships. If your academic work legitimately requires discovery-endpoint use (e.g. a grant-funded program with pre-negotiated IP terms), email access@dydact.io and we'll scope a dept-level discovery arrangement.
Apply for academic access →