- Sapien Dispatch
- Posts
- Sapien Weekly Digest - February 13th
What we’ve been up to:
New Year, New Us:
@JoinSapien on X is now @BuildOnSapien as we’re getting ready to launch the trust layer for Artificial Intelligence for builders across the globe!
The new Task UI:
Here’s the long awaited first look at what your new UI will look like:
The new environment that tasks will live in, this will be your home as a user!
We’re proud to finally reveal the first sneak peek, there will be much more to come where this came from. We were busy.
If you’re a builder, hit us up at [email protected] - We want to hear from you!
We’re going to ETH Denver!:
A trip down to Colorado! It’s time to show off what we’ve been building to the attendees at ETH Denver, in form of our Product Manager Ali Malik and our Director of Solutions Architecture Lukas Grapentine.
If you’re in Denver, hit us up and we’ll connect!
Sapien Team Takeover:
”I would Mandate open-weight models for all frontier systems above a certain capability threshold.”
Today’s Discord Team Takeover featured our Software Engineer, Jean-Philippe Melanson, on why Sapien uses blockchain to make quality verifiable at scale. JP spoke on the missing layer of accountability in AI and broke down Proof of Quality as the operational control system for AI data pipelines: immutable attestations, transparent verification, and cryptographic provenance that remove reliance on centralized gatekeepers.
He also has excellent taste in Hockey players.
Read the full transcript right here!
🚀 What’s New in the App?
While we’re still in the transition phase, the following tasks are available and taken care of for you!
Current task Overview:
😂 Comic Lines
😂 Comic Lines QA
👩🍳 Gastro Tag – Easy / Intermediate / Hard
👩🍳 Gastro Tag QA – Easy / Intermediate / Hard
💬 Emotion Prompt – Easy / Intermediate / Hard
💬 Emotion Prompt QA – Easy / Intermediate / Hard
📦 Logic Path Vietnamese QA
Our Voices in the World
What does this AI bubble really mean for builders and developers?
Our Product Manager Ali Malik was invited to sit down with the panel organised by Lucid AI to chat about the opportunities that AI enables and how it’s being used today!
Find the recording of the Space right here!
What else happened in AI?
The World outside of Sapien:
OpenAI begins testing ads in ChatGPT:
OpenAI started a limited US test of ads inside ChatGPT for logged in adult users on the Free and Go tiers. Paid tiers including Plus, Pro, Business, Enterprise, and Education remain ad free. OpenAI states that ads do not influence answers, and that conversations remain private from advertisers. Ads are shown as clearly labeled sponsored units and are visually separated from organic responses. Ad selection in the test uses conversation topic, prior chats, and prior ad interactions, while restricting ads near sensitive or regulated topics and excluding under 18 users, including in chats about health, mental health, or politics. OpenAI has positioned ads as a way to fund free and low cost access while continuing to invest in infrastructure and model capability.
According to the announcements, advertisers do not get access to individual people’s chat content, chat history, memories, or personal details, and only receive aggregate performance metrics. Retail marketers are treating the rollout as a test of whether “conversational ads” can work without breaking user trust, with Target already publicly describing labeled, separated ad placements that do not affect answers
Anthropic raises $30B Series G at $380B post money valuation:
On February 12th, Anthropic announced a $30 billion Series G led by GIC and Coatue, valuing the company at $380 billion post money. The company states the round will fund frontier research, product development, and infrastructure expansion. Anthropic lists a large set of additional investors and notes that the round includes a portion of previously announced investments from Microsoft and NVIDIA. As a part of this announcement, Anthropic claimed a $14 billion run rate revenue, growing more than 10x annually over each of the past three years. The company also stated enterprise adoption has accelerated, including growth in customers spending more than $100,000 annually and an increase to more than 500 customers spending more than $1 million annually. Anthropic also recommitted to multi cloud availability across AWS, Google Cloud, and Microsoft Azure and a diversified hardware stack across Trainium, TPUs, and NVIDIA GPUs.
While the stock market increasingly looks towards large scale investments at the top, this round resets the capital intensity expectations for frontier model competition and makes infrastructure scale an even more important factor. It also signals that enterprise distribution and multi cloud deployment are becoming decisive differentiators.
Z.ai releases GLM 5 as a frontier open weights model aimed at agentic engineering:
On February 12th, Z.ai announced the release of their latest model, dubbed GLM 5, framing it as a model designed for complex system engineering and long range agent tasks, framing it as a shift from “coding” to “engineering.” Z.ai stated that GLM 5 targets deep reasoning performance for backend architecture, complex algorithms, and difficult bug fixing. The release notes claim direct benchmarking against Claude Opus 4.5 on code logic density and systems engineering capability, and describe integrating DeepSeek Sparse Attention to improve token efficiency while maintaining long context quality.
Z.ai also published open weights artifacts, with the Hugging Face model card describing large improvements over GLM 4.7 and claiming best in class performance among open source models across reasoning, coding, and agentic tasks. The official repository describes a scale up from 355B to 744B parameters with 40B active parameters, plus pretraining data growth from 23T to 28.5T tokens. The model is licensed under MIT per the Hugging Face listing.
GLM 5 increases the competitive pressure on frontier closed models by raising the ceiling for open weights agentic coding and systems work. It also shows that architectural efficiency choices like sparse attention are now core to deployability.