1. OpenAI CFO Sarah Friar stated the company is not planning an…
OpenAI CFO Sarah Friar stated the company is not planning an IPO in the near term, focusing instead on growth and R&D over short-term profitability. The firm seeks government support in financing for increased data center capacity, noting challenges in raising debt due to uncerta
Category: Newsletter
2. Google Gemini’s “Deep Research” feature now integrates Gmail…
Google Gemini’s “Deep Research” feature now integrates Gmail, Drive, and Chat data to generate AI-powered research reports. Available on desktop with planned mobile rollout, this tool combines workspace data with web searches to create comprehensive research documents.
Category: Newsletter
3. Cursor improved its coding agent’s accuracy and user retenti…
Cursor improved its coding agent’s accuracy and user retention by implementing custom semantic search across codebases, resulting in fewer user iterations and better handling of code changes.
Category: Newsletter
4. Analysis using METR data suggests AI’s task completion capab…
Analysis using METR data suggests AI’s task completion capabilities double approximately every six months, potentially leading to the automation of full workdays within 18 months and industrial revolution-scale economic gains within two decades.
Category: Newsletter
5. OutSystems Agent Workbench provides a single platform with n…
OutSystems Agent Workbench provides a single platform with no-code tools, security features, and automated quality reviews to streamline the deployment of agentic AI systems, with customers like Bosch and Heineken already utilizing it.
Category: Newsletter
6. Perplexity released code allowing trillion-parameter mixture…
Perplexity released code allowing trillion-parameter mixture-of-experts models to run efficiently across multiple AWS servers, overcoming prior hardware and networking limitations that restricted large model deployment.
Category: Newsletter
7. Google enhanced the Gemini API with expanded JSON Schema sup…
Google enhanced the Gemini API with expanded JSON Schema support and improved property ordering for more reliable structured outputs in multi-agent workflows and database tasks.
Category: Newsletter
8. Pinterest plans to adopt open-source AI models which offer e…
Pinterest plans to adopt open-source AI models which offer effective user value at a reduced cost compared to large model providers, and will monitor user demand before further deploying AI shopping agents and personalized AI-curated boards.
Category: Newsletter
9. Perplexity will be integrated into Snapchat’s chat interface…
Perplexity will be integrated into Snapchat’s chat interface by early 2026, with a $400 million payment to Snap for the collaboration. Snap may seek additional partnerships with other AI companies to expand AI integration.
Category: Newsletter
10. Microsoft Research introduced SentinelStep, a mechanism with…
Microsoft Research introduced SentinelStep, a mechanism within its Magentic UI AI agent framework, enabling AI agents to efficiently manage polling intervals and maintain context during long-duration monitoring tasks, significantly improving task success rates.
Category: Newsletter
11. Anthropic’s new approach presents MCP servers as code APIs i…
Anthropic’s new approach presents MCP servers as code APIs in a filesystem structure, enabling AI agents to write programs that interact with tools and drastically reduce token usage, improving context management and privacy in AI workflows.
Category: Newsletter
12. Google launched the Jules extension for Gemini CLI that allo…
Google launched the Jules extension for Gemini CLI that allows users to assign coding tasks to autonomous agents running in the background, enabling uninterrupted workflows and automated code modifications.
Category: Newsletter
13. Two Indian founders released Maya1, a 3B parameter open-sour…
Two Indian founders released Maya1, a 3B parameter open-source voice model capable of handling over 20 emotions with sub-100ms latency on a single GPU, available under Apache 2.0 license with vLLM support.
Category: Newsletter
14. Microsoft’s new Azure ND GB300 v6 virtual machines achieved a record 1
Microsoft’s new Azure ND GB300 v6 virtual machines achieved a record 1.1 million tokens per second for Llama 2 70B inference, surpassing previous benchmarks and enabling faster enterprise-scale AI workloads.
Category: Newsletter
15. Giga AI raised $61M Series A for its self-improving AI agent…
Giga AI raised $61M Series A for its self-improving AI agents that automate customer operations, offering rapid two-week deployments and multilingual voice/chat experiences optimized for scaling operations.
Category: Newsletter