Sapien Weekly Digest

Hey there, Sapiens!

Welcome to November! We’re leaving the spooky season and moving rapidly into the holidays. That means we’re officially 85% into the year, but the news aren’t ready to slow down and get comfy just yet…

What we’ve been up to

Binance:
We’re now officially live on Binance! Binance has listed Sapien ($SAPIEN) and included it as the 57th project on HODLer Airdrops. The world’s largest exchange now also includes the world’s first decentralized data foundry.

Smart Contracts:
We have officially open sourced our smart contract repository, marking a major step toward full protocol transparency and community led development. This enables external developers, auditors, and users to verify that the contracts perform as intended, as well as transparency into how the smart contracts are designed to power staking, validation, rewards, and governance across the network.

🚀 What’s New in the App?

Since last week, we’ve updated to Sapien v0.187.0 - Here’s what’s changed:

• Top bar UI responsiveness

• Training hub task card UI Fixes

• Button sizing Cleanup

• Font size Cleanup

• Bounding Box Skills Assessment - coming soon

Current task Overview:

🍳 Breakfast Image Upload
🍽 Dinner Image Upload
🧩 Logic Path – Easy / Intermediate / Hard
🌐 Cadena Lógica (Español)
🗣 Njia ya Mantiki (Kiswahili)
🧠 Đường dẫn lô gic (Tiếng Việt)
🧬 Gastro Tag
🐞 Bug Hunters
📦 Breakfast Image Bounding Boxes

New Tasks:

Vehicle Model Mapping is coming soon!

This task helps build one of the most essential data layers for AI systems that interact with the physical world: understanding vehicles accurately. Modern AI models that support logistics, navigation, or vision systems all depend on structured vehicle data. You can be the one to provide that data!

We’re also preparing a new Skills assessment for you to show us that you have what it takes to provide quality data for future bounding box tasks!

Our Voices in the World.

“AI is learning faster than humanity is creating net new knowledge.” - check out our Founder Rowan Stone’s interview with TEACHMEDEFI on why gathering human expertise at scale will be invaluable as the need for data is ever increasing.

We joined NEXT Canada’s NEXT AI Toronto Open House event on November 5th, connecting with other founders in the Greater Toronto Area to bring the idea of Proof of Quality to a wide range of AI and business experts.

Our internal research articles this week are all about bias, drift, and the compliance issues companies run into by ignoring them. We explained what AI Model bias is in the first place, as well as how the existence of bias leads to the existence of AI Value Drift as a consequence. Companies are now expected to have a fully transparent and auditable record of how their AI model came to the result of its reasoning.

This week, we were joined by our friends at Recall, Pond, PepayLabs, and DeepSafe AI on a conversation about why every business will need AI on our hosted X Space! This morning, our very own Mo was invited by our partners at AITV to go into detail about the differences between AI companion and Human companions and what we still need to achieve to meet the demand for them in a secure manner.

Where are we going?

Next Friday, we’re inviting everybody to our monthly Town Hall! Apologies for the delay last time, but we’re all set and ready to go for the 14th.

We’ll make an announcement in our Discord closer to the time, but set your reminders for 2 PM Eastern Time! Our Founder Rowan Stone and our Head of Product Lauren Mushro will give you an overview on what we have planned for the near and long term future.

In the Pipeline:

Even easier Onboarding. We’re close to releasing explainer videos to coincide with a smoother onboarding process all around. We’re building Sapien for everyone, so even your grandparents should be able to find their way around.

Skills & Tasks. We’re aiming to release even more skill assessments and tasks for you to work on. The goal is to make sure diligent workers will always have a way to train AI.

Community initiatives. We’re creating more events to come for the holiday season. We’ll make a separate announcement across our socials for the events we’re planning on running.

What else is new?

Moonshot AI’s Kimi K2: The Next Open Source Inflection:
Moonshot AI has released Kimi K2 Thinking, a one trillion parameter mixture-of-experts (MoE) model with 32 billion active parameters per inference, trained on roughly 15.5 trillion tokens. Moonshot AI is backed by Alibaba Group, part of a broader national effort to close the capability gap with U.S. labs. The open source release under permissive licensing may serve both as a technical and geopolitical signal: demonstrating parity in agentic reasoning while exporting a lower cost development framework to global research ecosystems. If the performance data holds, Kimi K2 may become the first open model to rival proprietary frontiers such as GPT-5 and Claude 4.5 Sonnet on multitool benchmarks

Google is going to space:
Google just made public one of its most ambitious infrastructure projects to date: Project Suncatcher, a plan to build space based data centers powered by sunlight. The idea is to place solar powered satellites equipped with Tensor Processing Units in orbit, forming what Google calls a constellation scale AI compute fabric. Each satellite would connect to others using laser based interlinks, effectively operating as a distributed cluster above the atmosphere. According to Google’s research, solar panels in orbit can generate up to eight times more power than those on Earth due to constant sunlight and zero atmospheric loss. The company believes that by the mid-2030s, launch costs below $200/kg could make orbital compute cost competitive with terrestrial data centers on a per-kW-year basis. Two prototype satellites are planned for early 2027.

OpenAI & Amazon Web Services sign compute agreement:
OpenAI has entered a seven year, $38 billion infrastructure agreement with Amazon Web Services to secure the compute power required for its largest AI models. The deal gives OpenAI access to hundreds of thousands of NVIDIA GPUs and the ability to scale into the tens of millions of CPUs, with full deployment targeted before the end of 2026. It follows the removal of exclusivity clauses in OpenAI’s Microsoft partnership, signaling a clear move toward multi vendor diversification. For regulators, the concentration of compute power in a handful of hyperscalers presents new questions about competition, access, and national resilience in AI development.

See you next week!

Human insight continues to fuel our process, so invite your friends to join Sapien! We’ll continue to build for everyone, everywhere to be able to train AI and earn for their expertise. We wish you a great start into your November.

We’ll be building…