Databricks proposes structured chart extraction to improve enterprise agent retrieval accuracy
Latest signals
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
Cursor wins by owning the human-model interface, not competing on foundation models.
“We don't need to compete with anthropic and open AI on models right now. The interface between the human and the model is the key thing.”
Red Hat demonstrates KV cache-aware routing and prefill/decode disaggregation for agentic LLM workloads on Kubernetes.
“benchmarks actually don't show you is the chaotic reality of multi-turn interactions, massive context fluctuations which are very typical of agentic workloads”
LangChain defines 'Deep Agents' as long-running, high-context agents like Claude Code and deep research tools
“an agent like Claude Code, we would consider a Deep Agent because it's doing those longer-running jobs. It needs lots of context and to be able to manage that context over a long horizon.”
Databricks positions Lakebase Postgres and agentic AI as key drivers for vertical industry transformation
NVIDIA delivers Vera Rubin and Vera C servers to AWS in Seattle
Google DeepMind is piloting the world's first double-blind AI evaluations
AI agents with legitimate permissions create a new high-velocity enterprise security threat category
“the same type of threat is coming from non-human actors, agents that essentially have legitimate access to the environment with legitimate permissions”
Randomized study of 1,000+ students examines ChatGPT's impact on critical thinking and academic performance
Databricks Lakebase reimagines Postgres with object storage and WAL for agentic AI workloads
“Agents that interact with a traditional OLTP database often create bottlenecks at the storage layer.”
Parse error
NVIDIA acquires HuggingFace for $13B, nearly double its initial $7B offer.
OpenAI's Jalapeño chip claims 1.5–1.9x better performance per watt than NVIDIA GB200/GB300.
Google launches Gemini 3.5 Transcribe, an LLM-based transcription model with 70+ language support
“This model is an LLM based transcription model. So it can do really, really well in transcribing alphanumerics or email addresses, anything that normal transcription models might struggle with.”
Qwen releases 125B MoE model with only 6B active parameters as Qwen4 architecture preview
“a multimodal MoE model that also serves as an early preview of the architecture used in Qwen4”
SwiftUI adds a two-step reorderable drag-and-drop API in new WWDC release
DHH reversed AI skepticism after Claude Opus 4.5, calling it the dividing line to agentic coding
“I give it a couple of tasks, and I realize that the quality of the output is uncannily close to what I would've written. And I remember just leaning back and thinking, "What just happened?"”
Google launches Gemini 3.5 Transcribe, its first LLM-based transcription model with live streaming support.
“this model is both available for unary on the interactions API but also for life transcription on the life API”
NVIDIA NVLink Fusion enables NVHBM memory for custom XPU accelerators at hyperscale
Agent Plugins 1.0 GA unifies MCP and skills into one cross-platform open standard
“One package, multiple clients, one thing to version.”
Microsoft Developer hosting free 5-hour MCP Live stream on September 9th celebrating new stateless protocol release
“We're gonna have a five hour free live stream on September 9th in order to celebrate MCP and its new release with its stateless protocol.”
NVIDIA demonstrates AI agents training cross-embodiment robot navigation policies across robot types
AWS Bedrock AgentCore Evaluations uses OpenTelemetry to score agents from any framework
“Amazon Bedrock AgentCore evaluations solves this fragmentation by decoupling evaluation from the framework choice.”
Rich Sutton argues continual learning is just learning — the field's framing is the anomaly
“I'm not weird. The field is weird.”