Transparency
Nothing hidden.
Pinkk is an early-stage product. This page explains how we are funded, which infrastructure we rely on, how we use AI, and what we mean when we say we operate independently. We will keep it current as things change.
Startup credits & funding
Pinkk is primarily funded by its founder — personal capital put in to build the product and get it to revenue. No outside investors, no venture funding, no terms attached.
Alongside that, we have received startup credits from a number of technology programs. These are infrastructure credits, not investment — none of these partners hold equity, have influence over the product roadmap, or have any editorial say in how reports are written.
Credits allow us to run production infrastructure at reduced cost while the product is early and revenue is being established. When credits run out, we pay standard rates.
Founder capital
Pinkk is self-funded by its founder, Matt. No venture capital, no angels, no outside terms. Every product decision is made without external financial pressure.
Primary fundingAmazon Web Services
Compute, storage, and managed services via the AWS Activate program. Used for core application infrastructure.
Infrastructure creditsCloudflare
DDoS protection, CDN, DNS, bot management, and Turnstile for form protection via the Workers Launchpad / startup program.
Infrastructure creditsNotion
Internal documentation, knowledge base, and team operations via Notion for Startups.
Tool creditsIndependence note
Pinkk is funded by its founder and operates without outside investors or board influence. None of the technology credit programs listed above have visibility into our data, influence over product decisions, or any commercial arrangement beyond the credits themselves.
How we use AI
AI is central to Pinkk. Every report you run is analysed and structured using large language models. We think it is important you know exactly what that means in practice.
Which models we use
Reports are processed using models from OpenAI (GPT family) and Anthropic (Claude family), accessed via their commercial APIs. Configured Sable workloads use governed provider pools and routing rather than coupling the wider platform to one model vendor. We evaluate and update model choices as the landscape evolves.
What the AI does
The AI reads structured data we collect about a public profile — follower counts, engagement rates, content history, audience signals — and produces a confidence-rated findings report. It does not access private data, does not make legal judgements, and does not provide medical or financial advice.
Your data & model training
Data sent to OpenAI and Anthropic via their APIs is not used to train their models — this is a contractual protection in their API terms of service. We do not share your account data, uploaded assets, or report contents with any AI provider beyond what is necessary to generate a specific analysis.
Limitations & failure modes
AI can be wrong. Models can hallucinate, misread context, or reflect biases present in their training data. This is why every Pinkk finding carries a confidence rating — uncertain findings are flagged, not dressed up as facts. Reports are a structured starting point for research, not a final verdict.
Human-reviewed prompts
The instructions we give to AI models are written, reviewed, and tested by our team. They are not auto-generated. We iterate on them when output quality degrades or when new failure patterns are identified.
No automated decisions
Pinkk reports inform human decisions — they do not trigger automated actions on your behalf. A report never blocks an account, cancels a contract, or takes any action in a third-party system. That judgement stays with you.
Continuous improvement
We use internal quality signals, research sufficiency checks, confidence calibration, and human review to improve the product. When a finding or profile insight is corrected, the previous value and correction can be retained for review rather than silently overwritten. Changes to prompts, policies, and workflows are introduced deliberately and remain operator-controlled.
Sable — Pinkk's in-house intelligence layer
Sable is Pinkk's own intelligence and AI control plane, built and operated by us. For configured research and agent workloads, it coordinates provider routing, operational context, evidence handling, governance, and confidence controls around the external AI models and services we use.
Sable is not a third-party AI service. External AI providers are dependencies behind Sable, rather than integrations that each part of the platform manages directly. Its separate operational store and provider abstraction let us evolve capabilities without defining Pinkk around a single model vendor.
Intelligence orchestration
Sable coordinates the work around a model: which configured provider and model can handle a task, what operational context is available, and which governance policies apply. Models supply capabilities; Sable provides the consistent operational layer around them.
Provider-independent
Sable uses configurable provider pools and routing. Changing or adding a provider is a Sable-level decision, allowing us to assess models by capability, reliability, latency, cost, and task suitability instead of building the product around one vendor.
Governed by design
Governed Sable actions are evaluated against a control stack that can include kill switches, circuit breakers, per-capability approval policies, rate limits, spend budgets, previews, and rollback where available.
Isolated memory & context
Sable keeps its persistent operational memory, decisions, and audit data in a dedicated database separate from Pinkk's primary data stores. Context is selected by the relevant workflow, and report evidence is not inferred from memory alone.
Claim & evidence lineage
Where reports contain explicit source citations, Sable can associate claims with their supporting evidence and surface grounded, partial, or unlinked coverage. It does not infer a source relationship from proximity or similarity.
Full audit trail
Sable records the operational context around governed actions, including the capability involved, policy applied, rationale, preview, affected scope, execution status, outcome, and rollback state where available. This gives operators an auditable record of controlled actions and their results.
Infrastructure & third-party dependencies
These are the services Pinkk depends on to operate. We list them so you understand where your data flows.
Hosting
Application servers run on AWS infrastructure. Databases are managed and backed up daily.
CDN & Security
Cloudflare sits in front of all traffic for DDoS protection, caching, and bot filtering.
Payments
Billing is handled by Polar.sh. Pinkk never stores card details — payment data flows directly to Polar.
Transactional email is sent via Resend. Only your email address and the content of the specific email are passed to their service.
Monitoring
Uptime and incident tracking via BetterStack. Publicly visible on our status page.
Analytics
Privacy-first analytics with no cross-site tracking, no fingerprinting, and no personal data storage.
AI providers
OpenAI and Anthropic APIs. Data sent is scoped to the current analysis only. Not used for model training.
Error tracking
Application errors are captured to help us debug and fix issues. Error reports are sanitised before storage — no report content or account credentials are included.
Get involved
Pinkk is early. If you work in due diligence, creator partnerships, brand safety, or iGaming compliance and want to shape the direction of the product — we would like to hear from you.
We are open to conversations with potential advisors, design partners, and researchers who care about evidence-based decision-making. This is not a sales pitch — it is a genuine invitation to people who think the problem is worth solving.
Early access
Get early access to features before they ship. We share work-in-progress with users who have agreed to give honest feedback.
Express interestAdvisory & partnerships
If you have domain expertise in influencer marketing, brand safety, or iGaming compliance and want to work with us formally, get in touch.
Get in touchFeedback
Found something wrong in a report? Have a feature that would make Pinkk more useful for your workflow? Tell us directly.
Send feedbackLast updated
This page was last reviewed in August 2026. We update it whenever the material facts change — new funding, new providers, or changes to how AI is used in the product. Questions or corrections? Contact us.
See it for yourself.
Run a free intelligence report and see exactly how our AI structures and confidence-rates its findings. No credit card required.