AgentsAgents
DeepSeek Opens dsh-superagent for queries from GitHub and Feishu
DeepSeek reported a reduction in p95 time to complete tasks in its own simulation after releasing an open-licensed research prototype.
Topic
Agents
AI agents and autonomous systems: workflow platforms, orchestration tooling and real deployments. We track what agents already do in production, where the limits of autonomy lie and how companies keep their actions under control.
Topic
Primary topic in 284 of 284 reports
- Agents
AgentsAgents
Google creates StreamVibe for timely video assistance
The video assistant StreamVibe should intervene only when necessary, but long videos remain a challenging task for models.
AgentsAgents
Factory released Droid 1.0 for programming
Factory introduced Droid 1.0 and reported results of its own tests on Terminal-Bench 4.
AgentsAgents
Microsoft Research linked Copilot assessment to synthetic data
Microsoft Research highlighted a conflict between the realism of tests and protecting sensitive corporate data.
AgentsAgents
SAP and Databricks link SAP data with enterprise AI tools
SAP and Databricks tied access to SAP data to a set of tools for enterprise AI, but details of the announcement are currently limited to the synopsis from a single source.
AgentsAgents
Databricks Blog linked client onboarding to context and verification
Databricks Blog links high-quality client onboarding to case context, evidence, and human oversight, but does not provide implementation results.
AgentsAgents
Claude Independently Chose Gene Therapy for a Brain Surgery on a Mouse
Claude chose gene therapy for a brain surgery on a mouse
AgentsAgents
Samaya AI released Sterling for financial research
Sterling achieved a higher FrontierFinance score at a claimed cost per question of 0,61 dollars against 6,89 dollars for Claude Fable 5.1.
AgentsAgents
Anthropic, OpenAI, Cloudflare, and Google Release AI Agents for Vulnerability Discovery
Four companies introduced tools to search for vulnerabilities in source code, but available data are insufficient to compare their effectiveness.
AgentsAgents
Zest introduced an AI concierge for choosing and booking restaurants
Zest described an AI concierge for joint restaurant selection and booking via text.
AgentsAgents
Infini-AI-Lab released Hermit, an Open-Source Harness Adapter Framework for AI Agents
Infini-AI-Lab introduced an open adaptation layer that sits on top of the model and the agent wrapper.
AgentsAgents
The UK Amazon Quick User Group held three meetings
Group held three meetings and gathered more than 150 participants, presenting client stories about agent-based AI and Amazon Quick.
AgentsAgents
Microsoft Research discussed new research directions
Microsoft Research dedicated the 97-th issue of the series to problems of code tools, workplace AI use, and weather forecasting.
AgentsAgents
Anthropic found that 50 experts took the AI's position as their own
Anthropic study shows that experts often accepted randomly assigned system positions as their own
AgentsAgents
OpenAI describes monitoring for inconsistency of frontier models
OpenAI described a multi-level monitoring system to detect inconsistency in frontier models and unauthorized behavior.
AgentsAgents
Wabi raised $55 million at a valuation of $1,25 billion
Wabi announced a round of $55 million at a valuation of $1,25 billion, but details about the deal and the product are still scarce.
AgentsAgents
Legora is building the agentic operating system for legal work
Legora links OpenAI, ChatGPT and Codex models to speed up legal work and contract analysis.
AgentsAgents
Google Opens Android Studio to Third-Party AI Agents
Android Studio expanded support for external AI tools, including Anthropic and OpenAI.
AgentsAgents
Bank of America kindly asks that you don’t share your banking password with your AI assistant
Bank of America reminded about the risk of sharing banking passwords with AI assistants and the client's responsibility for access-related actions.
AgentsAgents
OpenAI expands partnership with Oracle to bring its apps to ChatGPT
OpenAI and Oracle expand cooperation around business applications and AI agents within ChatGPT.
AgentsAgents
Mirendil released Fern 0.1 with a context of 10 million tokens
Mirendil unveils first model with context up to 10 million tokens; in code-agent development it lags behind leading systems.
AgentsAgents
OpenAI turns coding tools into dots, an always-on personal assistant
TechCrunch AI describes dots as an expansion of OpenAI's coding tools into a constantly-on personal assistant.
AgentsAgents
AWS describes caching of OpenAI context in Amazon Bedrock
AWS describes caching repeating context for multi-step systems in Amazon Bedrock.
AgentsAgents
The RealReal updates bot protection on purchases
The RealReal tightens site protection after reports of possible use of the 20-minute reservation mechanism by an automated tool.