Together AI and Pearl Research Labs Team Up to Reduce the Cost of AI Inference
On Thursday, May 14, 2026, a flurry of updates emerged in the AI and data engineering landscape, with a mixed bag of announcements from major players. The day's news spanned partnerships, new features, and insightful blog posts, offering a glimpse into the evolving world of artificial intelligence and data management.
Together AI has partnered with Pearl Research Labs to launch a discounted Pearl-powered inference endpoint for Gemma-4-31B-it-pearl, utilizing Proof of Useful Work to turn AI workloads into crypto emissions. This move is set to reduce the cost of AI inference, with the endpoint supporting a range of applications.
OpenAI published a series of case studies showcasing how different teams can leverage Codex, including sales teams, data science teams, and business operations teams. These case studies demonstrate how Codex can be used to create pipeline briefs, meeting prep packets, and initiative briefs, among other applications. Additionally, OpenAI unveiled a new personal finance experience in ChatGPT for Pro users in the U.S., allowing users to securely connect their financial accounts and receive AI-powered insights.
Databricks brought GPT-5.5 to enterprise agent workflows, following the model's impressive performance on the OfficeQA Pro benchmark. This integration is expected to enhance the capabilities of enterprise agents, enabling them to handle more complex tasks. Estuary, meanwhile, published a blog post exploring why AI agents make wrong decisions, highlighting the importance of access to fresh data for LLM-driven AI agents.
On the infrastructure side, Databricks expanded interoperability with Unity Catalog Open APIs, designed for the open lakehouse. This update allows data teams to break free from traditional data silos and work more efficiently. Confluent announced new Kafka observability updates in Confluent Cloud, including metrics to understand consumer groups and elastic clusters. The company also introduced Tableflow, which helps feed data lakes with real-time, analytics-ready tables for 30-50% lower cost.
Hugging Face released Granite Embedding Multilingual R2, an open Apache 2.0 multilingual embedding with 32K context, offering best-in-class sub-100M retrieval quality. This new embedding is set to improve the performance of various NLP tasks. Confluent also published an architectural guide to event-native governance, covering schemas, lineage, security, retention, and compliance patterns for reliable data platforms.
AWS AI/ML announced several updates, including Amazon Lex Assisted NLU, which improves bot accuracy, and real-time voice agents with Stream Vision Agents and Amazon Nova 2 Sonic. The company also introduced cross-account Athena access for Amazon Quick, enabling customers to query Athena data in other AWS accounts using IAM roles. Furthermore, AWS AI/ML demonstrated how to control where AI agents can browse using Chrome enterprise policies on Amazon Bedrock AgentCore.
A quieter note from Databricks came in the form of a blog post exploring how AI agents are transforming electric grid operations, highlighting the potential for autonomous systems to improve efficiency and reliability. ClickHouse, meanwhile, published a deep dive into pg_clickhouse's Foreign Data Wrapper, explaining how it decides what SQL to push down to ClickHouse versus execute locally in Postgres.
Google Cloud AI shared perspectives from the Cloud CISO, discussing how Google and Wiz are changing multicloud strategy for CISOs. ClickHouse also compared its performance to Prometheus for high cardinality, part of a series exploring the architectural tradeoffs of series-based systems. NVIDIA announced that Subnautica 2 is now available on GeForce NOW, day-and-date with its launch.
OpenAI announced that Codex can now be used from anywhere, thanks to the ChatGPT mobile app, allowing users to monitor, steer, and approve coding tasks in real-time across devices and remote environments. Snowflake published two blog posts, one on how Cortex Code helps FP&A move from static reporting to near real-time insight, and another on the importance of a reliable semantic layer in financial services AI risk.
On funding, several companies made notable announcements, although details were scarce. Meanwhile, at Sea Limited, the company's CPO explained why they are deploying Codex across engineering teams to accelerate AI-native software development in Asia.
That's the Daily AI Brief for Thursday, May 14, 2026. More tomorrow.
Sources covered
1 story · 25 sources (0 news, 25 company, 0 voices)