We broker and license the complete engineering histories of private software companies, operating and legacy alike. We certify their chain of title and structure each license to maximize its value to the AI labs and data companies training agentic models.
Named for Scapa Flow, Orkney, where the world's low-background steel is salvaged from the seabed. We do the same for code.
Labs need provably human, provenance-clean engineering data, and it stopped being produced at scale in 2022.
High-quality open repositories are already trained on, increasingly AI-contaminated, and the subject of active copyright litigation.
Models need causal traces of how teams actually build: design arguments, failed experiments, code review, months of system evolution.
No new pre-AI code will ever be written. Supply is fixed, shrinking, and appreciating. Every corpus certified is permanently off the board for everyone else.
A data-rights origination company: mandates from operating founders and wind-downs, with every certified asset licensed on the structure that earns it the most.
Commission-based brokerage mandates from operating founders at zero upfront cost, plus outright acquisition of wind-down assets via ABCs and bankruptcies, permanently growing our library.
Chain-of-title and IP-assignment audit, open-source/copyleft scan, PII and customer-data scrub, and context packaging: commits, PRs, issues, and design docs bundled into a licensable dataset.
Exclusive, time-limited exclusive, or non-exclusive: a competitive process prices each structure on its merits. Exclusivity commands a premium; non-exclusive terms keep re-licensing and recurring subscriptions on the table.
A tiered standard every buyer can trust and reference in their own downstream contracts, whether or not the asset trades through us.
Engineering history authored entirely before the era of AI code generation. Provably human, dated by commit metadata, and permanently scarce.
Modern corpora with documented AI involvement: tooling used, commit-level attribution, and review threads showing humans correcting AI output. That is the signal agentic models need most.
Ongoing engineering history from operating companies, delivered on a time lag as a recurring subscription. Non-dilutive revenue from work already done.
We turn it into non-dilutive revenue without selling your company or your edge, structured as a one-time premium, a time-limited exclusive, or recurring licensing, whichever earns the most. There is no upfront cost: we are paid on closing, and certification is included.
| Selling direct | Through Orkney Labs | |
|---|---|---|
| Price discovery | One buyer, first offer. No competitive tension, and exclusivity is given away free. | Multi-buyer process: certified assets priced by competition, exclusivity sold at a premium. |
| Diligence | Raw, unverified rights mean buyers discount heavily or walk after diligence. | Certification closes the diligence discount; warrantable data commands top of range. |
| Liability | Founder personally warrants IP claims they never actually checked. | Reps backed by a documented audit: seller liability drops, deals close faster. |
| Your business | PII and customer data buried in years of git history is liability shipped to a stranger. | PII scrubbed, delivery time-lagged, field-of-use limited. Engagements confidential by default. |
| Upside | First offer sets the terms, and exclusivity is handed over without being priced. | Every structure valued on its merits: exclusivity priced at a premium, or non-exclusive terms that preserve re-licensing and subscription upside. |
Scope stays yours to define: engagements cover only the named repository you choose to license, never your current products, data, or future IP. You approve every deal in writing before it is signed. For founding clients, our flat certification fee is waived in full.
Everyone in this market needs clean provenance, including teams that never transact through us.
Flat-fee audit and packaging for companies licensing direct: chain-of-title report, OSS/copyleft scan, PII scrub, context bundle. You keep the deal.
Labs and data companies verify third-party corpora before they buy. Every raw dataset changing hands is a diligence engagement waiting to happen.
"Certified pre-AI" · "Attributed AI-collaboration" · "Living corpus": a standard buyers can reference in their own downstream contracts.
Whether you are an operating founder with a legacy repository, a wind-down with assets to place, or a buyer who needs provenance verified, tell us what you are working with.
Michael Silberberg · Founder, Orkney Labs