Sapien

Most AI failures are not “mystery bugs.” They are predictable outcomes of unverified judgments made somewhere in data capture, evaluation, or review. Proof of Quality is built to make those judgments auditable and accountable. Today we are publishing the Sapien roadmap so builders can see exactly what is being shipped, when, and why. It is organized around one goal: make Proof of Quality a drop in primitive for any AI pipeline: Sapien.io/roadmap

With increasing autonomy, agents will need to be able to verify their actions. The missing piece is a record of the standard, the reviewers, and the path to the outcome. Proof of Quality is that record.

Airports are struggling with a growing wave of AI generated articles and websites publishing inaccurate information about airport and flight operations. The fabricated stories appeared in search results, copied the visual language of legitimate news, and directed travellers toward misleading operational claims. Information that looks correct can still come from stale or fabricated data. Operational AI work needs a verifiable record showing which sources were checked, when they were checked, and how publication approval was reached.

Meta’s AI detection tool was routinely fooled by minor adjustments to images created by Meta’s very own Muse Image. The detector claimed that 55% of the AI images were identified as real after they were cropped to roughly one third or one half of their original size. The useful question here is wider than “Does the AI detector work?” We need to verify which edits it survives, how often it fails, who tested it, and where reviewers disagreed. Proof of Quality can preserve those answers.

The NHTSA issued a call to action after documenting a pattern of autonomous vehicles entering emergency scenes, blocking ambulances and firefighters, and failing to respond to lights, fire, cones, and hand signals. Autonomous vehicles can navigate ordinary traffic and still fail when the environment changes. The builder question is simple: What guarantee do we have that the AI running the vehicle will get it right every time when we can't afford mistakes? Proof of Quality creates a verifiable record of the evaluation.

A lawyer faced the Connecticut Supreme Court after AI assisted legal briefs included nine errors, including invented quotes and incorrect citations. The filing entered a consequential workflow before the work received adequate verification. Proof of Quality prevents this.

Right here: directory.sapien.io/pages/sapien-c…

Applications for the Sapien Council close tonight! We are opening just 20 seats in the Sapien Council: a small, hand picked group of builders who get early access to Sapien’s infra, direct time with our eng team, and a real voice in what we build next Council members get: - Under the hood access — live demos before new features ship publicly - Builder Fridays — biweekly product previews and Q&A with the core team - Real influence — feedback that directly shapes our roadmap - Founding recognition — recognized as a founding Council member across Sapien channels We have already received dozens of applications from all over the world, with only 12 hours left to apply This is for builders working with AI agents, or anyone who cares deeply about trust, verification, and data quality The application takes about five minutes, and our team reads every submission personally. Know someone who would be a great fit? Send them this! Apply below:

Apply here: directory.sapien.io/pages/sapien-c…

Applications Closing Soon: Join the Founding Sapien Council Response has been strong, and our Sapien builder council spots are filling up. Applications to the council closes on June 7, 2026.

We are reviewing for technical depth and active builder context. If you are building around AI verification, AI data quality, agent evaluation, or data pipelines, share your GitHub, portfolio, project, or current technical challenge. Apply below!

If that sounds interesting, apply here: directory.sapien.io/pages/sapien-c… The application takes about 3 minutes. We'll review every submission personally, and applications will close once all 20 seats are filled.

Applications are open for the founding Sapien Council! We're inviting 20 builders to join a private group helping shape the future of AI verification. If you're working on AI agents, model evaluation, data quality, review workflows, or AI verification, we'd love to hear from you. As a council member you'll get: - Early access to new infrastructure and features - Builder Fridays with demos, previews, and Q&A with our engineering team - A direct voice in what we build next - Recognition as a founding council member

Hallucinated URLs have moved from model flaw to attack surface. A model can invent a plausible domain, an attacker registers it, and the bad link can travel through search, chat, or agent workflows. Hallucinated domains are a clean example of faulty AI work becoming real world infrastructure. The builder lesson from phantom squatting is simple. Every generated domain needs to be treated as an artifact that needs verification to be trusted.

Bank of England Deputy Governor Sarah Breeden said agentic AI may require regulatory reform as autonomous payments and trading expand. As Agentic AI is moving into finance, payments, and trading, builders should hear the signal clearly: autonomous systems need verifiable review processes before their actions affect customers, infrastructure, or reputation. Proof of Quality helps create a record of what standard was applied, who reviewed the work, and how the final outcome was reached.

DuckDuckGo AI search repeated a fake claim from poisoned web content. This is a clean example of why "the model cited sources" is not enough of a quality signal. A bad source can still produce a polished answer. Proof of Quality gives AI work a verifiable record, so downstream users can inspect what standard was applied and how the result was reached.

Interested in running a pilot? Reach out to build@sapien.io and we’ll help you verify your AI work and work with you as part of our early adopter program.

Teams do not need to rebuild their security audit workflow to test Proof of Quality. Start with one queue, one checkpoint, and one proof artifact. You can take the findings already moving through review, add expert validation where trust breaks down, and see whether the output becomes easier to accept, reject, or escalate.

Avon and Somerset Police built a broad predictive analytics system across at least 23 models, one burglary model even ran with precision under 10 percent for more than three years. Many of the models were abandoned after staff lost confidence in their accuracy and transparency. AI work that can affect people’s lives needs a verifiable record of how it was evaluated. Without an inspectable record, downstream users inherit uncertainty.

Security audit teams do not need another black box in the review process. - Product wants faster cycles. - Security ops needs reliable findings. - Technical leadership needs a defensible record. Verification has to serve the whole organization. That’s why Proof of Quality provides all three.

Bloomberg Law reports that every large law firm in its survey used legal AI tools in 2025. That creates a clear builder problem: AI work in regulated industries needs an auditable verification process before people can rely on it. Proof of Quality adds that verification layer.

Estonia giving agents digital IDs to AI agents is a clear market signal: Autonomy needs attribution. Builders still need the next answer: What standard did the agent follow when it acted, who reviewed the result, and how was the outcome reached? Proof of Quality turns that review into a verifiable artifact.