Sapien Weekly Digest

The Roadmap for 2026 is finally here!

What we’ve been up to:

Preparing for the new App:
Over the last phase of Sapien, one thing became clear. The current app taught us a lot, but it was built for a different stage of the product. Proof of Quality requires a new build that is built from the ground up with developers in mind: More open, more scalable, and designed around verification from day one.

As we move toward launch, we are deprecating the current Sapien app and transitioning to a new Proof of Quality-native experience built for AI developers and builders integrating quality verification into training pipelines, evaluation systems, agent workflows, and production safety gates.

Task access in the current app has closed at the time of this announcement. If you have pending USDC that has not yet been paid out, there will be a one week reporting window to flag it for review.

Reporting window:

• Opens: March 5, 5pm EST

• Closes: March 11, 5pm EST

Please submit your report through our standard customer service channels and include:

• Your account identifier, email or username, and wallet address if applicable

• The total pending amount shown in the app

• Any additional details that help us verify the claim

If you already submitted a report, there is no need to send it again. The team will follow up directly. Reports submitted after the reporting window may not be eligible for review as we wind down the current app payout workflow. Unstaking will remain available with no changes, there is no unstaking deadline at this time. Questions or issues related to unstaking will continue to be handled after the pending USDC reporting window ends.

This transition is about making it as seamless as possible for developers to make use of the PoQ protocol. Proof of Quality is designed for developers that need to prove how quality was established across training data, evaluations, and production decision flows, and the next version of Sapien will be purpose built for those builders from the start.

We are in final audits before launch.

Join the early adopters list for priority access when Proof of Quality goes live!

Engineering Corner:
The next version of the app is being designed with the intent of making it easy for anybody to build on Proof of Quality, whether they’re web3 natives or not. Builders should not have to learn an entire crypto workflow just to get through onboarding.

A lot of current engineering work sits exactly at that boundary. The team is revising the task UX, tightening the model for user authentication, and rethinking wallet structure so participation feels simple for Web2 native builders even if the protocol itself remains onchain. The goal is to remove as much friction as possible without compromising the underlying mechanics that make verification credible. Stay tuned for updates!

Our Voices in the World:

On Wednesday, our Product Manager Ali Malik sat down with a panel of experts invited by Tria to talk about the differences between a product that’s scaled to last and one that is carried by hype. Check out Ali’s thoughts right here!

What else happened in AI?

The World outside of Sapien:

OpenAI launched GPT-5.4:
OpenAI launched GPT-5.4 on March 5, 2026, positioning it as its new flagship model for professional work across ChatGPT, the API, and Codex. OpenAI says GPT-5.4 combines advances in reasoning, coding, and agent workflows, and is its first mainline reasoning model to incorporate GPT-5.3-Codex capabilities.
On OpenAI’s reported benchmarks, GPT-5.4 improved from GPT-5.2 on knowledge work and spreadsheet tasks, including 83.0% on GDPval versus 70.9% for GPT-5.2 and 87.3% on internal spreadsheet modeling tasks versus 68.4% for GPT-5.2. OpenAI also says GPT-5.4 is its “most factual” model so far, with individual claims 33% less likely to be false and full responses 18% less likely to contain errors than GPT-5.2 on a set of flagged prompts.

In the same breath, OpenAI also launched ChatGPT for Excel in beta on March 5, 2026, embedding ChatGPT directly inside Excel workbooks to help users build, update, and analyze spreadsheet models. OpenAI says the tool preserves Excel-native structure, formulas, and assumptions, and can explain changes, trace errors, and link responses to the exact cells it references or edits. The beta is rolling out to Business, Enterprise, Edu, Teachers, Pro, and Plus users in the U.S., Canada, and Australia. OpenAI also says ChatGPT for Google Sheets is coming soon.

U.S. government break with Anthropic:
The Trump administration’s clash with Anthropic over military AI safeguards has escalated into a formal break. After ordering federal agencies to phase out Anthropic tools, the Pentagon has now designated the company a “supply-chain risk,” immediately blocking Claude from Pentagon contract work and setting up a court fight over who gets to set limits on battlefield AI. Anthropic said it received formal notice on March 4, and the Pentagon’s “supply-chain risk” designation became effective immediately on March 5, barring government contractors from using Anthropic technology in work for the U.S. military. Anthropic says it will challenge the move in court.  

At the same time, investors and industry groups have pushed to de-escalate because this kind of designation has typically been used against foreign adversaries, not U.S. AI firms. This is the clearest test yet of who controls military AI boundaries: the government customer or the model provider. The outcome could reshape defense procurement, determine whether AI labs can enforce red lines on surveillance and autonomous weapons, and set a precedent for treating a domestic U.S. tech supplier like a national security supply chain threat.

White House power pledge with hyperscalers and AI firms:
On March 4, 2026, the White House secured a voluntary “Ratepayer Protection Pledge” from Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI. The companies agreed to secure the power needed for AI data centers and cover related infrastructure costs instead of shifting those costs onto households and small businesses. Under the White House framework, the firms would pay for that capacity even if they do not fully use it, and they also agreed to coordinate backup generation with grid operators to support reliability during emergencies.

The pledge was signed at the White House after Trump first previewed it in his February 24 State of the Union, as concerns grew that rapid AI data center expansion could raise electricity prices for consumers.  Signatories agreed to “build, bring, or buy” new power supply for their data centers, including electricity from new plants or from existing plants with expanded output. The companies also committed to pay for grid and power delivery upgrades tied to their data centers, and to negotiate separate rate structures with utilities and state governments. 

More to show off next week!