Status: Updated strategic position Date: July 7, 2026 Author: Ditto (strategy memo draft) Stakeholders: Peyton Spencer, Sebastiano Rubino, Omni Aura LLC
Note: This memo is a strategic product and licensing position, not legal advice. Final license implementation should still be reviewed by counsel.
This memo supersedes the earlier AGPL-oriented framing.
After clarifying the actual scope of the codebases, the right licensing posture is:
@ditto-assistant/ditto-subnet → MIT@ditto-assistant/dittobench-starter-kit → MIT@ditto-assistant/ditto-harness → MITThe core reason is simple: none of these repositories contain the real secret sauce. They are ecosystem, trust, and adoption surfaces — not the proprietary heart of Ditto or Omni Aura.
If the subnet code, starter kit, and harness do not expose the defensible product moat, then using AGPL creates more friction than protection. In that case, the higher-leverage move is to make these repos maximally easy to inspect, adopt, integrate, and extend.
That approach fits both:
| Repo | Role | Recommended License | Why |
|---|---|---|---|
@ditto-assistant/ditto-subnet |
Subnet-facing infrastructure and integration surface | MIT | Low-friction adoption, ecosystem trust, standard open infrastructure posture |
@ditto-assistant/dittobench-starter-kit |
Templates / onboarding / developer bootstrap | MIT | Encourages copying, experimentation, contribution, and fast builder uptake |
@ditto-assistant/ditto-harness |
Public execution / validation / agent runner surface that the subnet product produces | MIT | If it does not expose the real moat, permissive licensing is a distribution advantage rather than a risk |
Open the trust surface. Protect the actual moat elsewhere.
The earlier AGPL framing depended on one key assumption:
ditto-harnessitself contained the proprietary strategic engine that needed to be legally defended from closed-source capture.
Your update changes that.
If ditto-harness, dittobench-starter-kit, and ditto-subnet do not expose the hard-to-copy value, then AGPL is solving the wrong problem.
It adds legal hesitation without protecting the real product moat. If the important value is in private systems, hosted infrastructure, internal orchestration, product quality, enterprise relationships, or closed operational know-how, AGPL on the public repos does not meaningfully defend that value.
It introduces avoidable friction for miners, validators, partners, and contributors. In Bittensor and adjacent open infra ecosystems, permissive licensing is often treated as the default expectation. AGPL invites second-order questions around obligations, linking, operational risk, and compliance.
It weakens distribution when distribution is the point. If these repos are meant to become trusted public building blocks, then the strategic goal is broad use, rapid integration, easy forking, and low-friction auditability.
It can mis-signal where the moat lives. Heavy copyleft on non-secret repos can accidentally suggest that the company is trying to create scarcity around code that is not actually the main differentiator.
In short: if the moat is elsewhere, AGPL taxes growth more than it defends value.
MIT is the cleanest invitation to the ecosystem:
That matters if the goal is to make Ditto-adjacent tooling feel like infrastructure rather than a guarded attachment.
For a subnet-related product surface, transparency matters.
If outside builders, miners, validators, or partners can easily inspect the harness and surrounding repos, that reduces suspicion and increases confidence that the public-facing components are doing what they claim.
This is especially useful when the product story depends on reliability, reproducibility, validation integrity, or workflow trust.
The culture around Bittensor tends to reward:
MIT aligns with that environment better than AGPL.
If the objective is category influence, ecosystem spread, and developer goodwill, permissive licensing is not just acceptable — it is usually the stronger move.
A uniform license posture across subnet repo, starter kit, and harness reduces confusion.
That simplicity helps with:
Instead of explaining why one public repo is “special” from a licensing perspective, you can say plainly:
These repos are open infrastructure. They are MIT because we want the ecosystem to use them.
That is easy to understand and easy to trust.
Choosing MIT here does not mean giving away the business.
It means being precise about where the moat actually is.
Managed memory infrastructure The hosted persistence, retrieval quality, indexing, graph management, provisioning, and operational reliability behind Ditto.
Backend orchestration and internal systems Private workflows, service coordination, deployment logic, internal quality gates, observability, security posture, and production reliability layers.
Product UX and managed experience The frontend experience, polished workflows, trust signals, account system, memory UX, enterprise usability, and the overall product feel.
Private evaluation assets and operational know-how Internal prompts, datasets, tuning methods, deployment heuristics, test corpora, scoring adjustments, performance learnings, and non-public operational playbooks.
Enterprise distribution and support SLAs, integration assistance, onboarding, hosted offerings, reputation, customer trust, support quality, and long-term product velocity.
Brand and ecosystem position Ditto and Omni Aura can still defend value through brand trust, product depth, and the quality of the managed system even when public repos are permissive.
MIT does not weaken the moat if the moat is not in these repos.
It simply puts the open-source boundary in the right place.
If you go MIT across all three repos, the real discipline shifts from legal restriction to repo boundary hygiene.
That means being extremely intentional about what does and does not live in public.
If MIT is the choice, repo scope discipline becomes the protection mechanism.
Each repo should clearly state:
This avoids confusion about what is open and what is proprietary.
Open source the code, but do not unintentionally open the brand.
You can keep strong control over:
MIT covers copyright licensing of the code. It does not require surrendering trademark control.
Even under MIT, you still have a strategic choice:
Use this if Omni Aura wants:
Use this if you want:
A simple rule of thumb:
The strongest MIT strategy is not “open everything and hope.”
It is:
A clean public framing could be:
We’re licensing
ditto-subnet,dittobench-starter-kit, andditto-harnessunder MIT because we want the ecosystem to inspect, use, extend, and build on them with minimal friction. These repos are open infrastructure and trust surfaces. The deeper Ditto / Omni Aura product moat lives in the managed systems, product experience, and operational layer around them.
That message is strong because it signals:
Given the clarified scope, MIT across ditto-subnet, dittobench-starter-kit, and ditto-harness is the right strategic choice.
That decision works because these repositories are not the secret sauce. They are the adoption surface, the trust surface, and the public infrastructure layer.
If the real defensible value of Ditto and Omni Aura lives in the managed memory system, backend orchestration, private operational know-how, enterprise execution, and product experience, then permissive licensing is not a giveaway.
It is a growth lever.
The winning posture is:
That is a coherent open-source strategy for both Ditto and Omni Aura.