Here is your today’s AI Dev Brief from Marktechpost, covering core research, models, infrastructure tools, and applied updates for AI developers and researchers.
Google Health AI Releases MedASR: a Conformer Based Medical Speech to Text Model for Clinical Dictation
Google Health AI has released MedASR, a 105M parameter Conformer based medical speech to text model trained on about 5000 hours of de identified clinical dictations and conversations, delivering state of the art word error rates on radiology, general medicine, family medicine and Eye Gaze benchmarks, and giving healthcare developers an English only, open weights ASR backbone that runs with Transformers via Hugging Face or Vertex AI and plugs cleanly into downstream models like MedGemma for clinical summaries and SOAP note generation, under the Health AI Developer Foundations license.. Read the full insights/article here.
InstaDeep Introduces Nucleotide Transformer v3 (NTv3): A New Multi-Species Genomics Foundation Model, Designed for 1 Mb Context Lengths at Single-Nucleotide resolution
NTv3 is InstaDeep’s new long range genomics foundation model that operates on 1 Mb DNA windows at single base resolution and supports 24 animal and plant species. It is pretrained on 9 trillion base pairs from OpenGenome2 and post trained with a joint objective on about 16,000 functional tracks and genome annotations, which enables state of the art performance on the Ntv3 Benchmark with 106 long range sequence to function tasks. The same backbone also supports controllable enhancer sequence generation via masked diffusion language modeling, with designs experimentally validated using STARR seq…. Read the full insights/article here.
CopilotKit v1.50 Brings AG-UI Agents Directly Into Your App With the New useAgent Hook
Agent frameworks are now good at reasoning and tools, but most teams still write custom code to turn agent graphs into robust user interfaces with shared state, streaming output and interrupts. CopilotKit targets this last mile. It is an open source framework for building AI copilots and in-app agents directly in your app, with real time context and UI control. The release of of CopilotKit’s v1.50 rebuilds the project on the Agent User Interaction Protocol (AG-UI) natively. The key idea is simple; Let AG-UI define all traffic between agents and UIs as a typed event stream to any app through a single hook, useAgent.....
This AI Paper from Stanford and Harvard Explains Why Most ‘Agentic AI’ Systems Feel Impressive in Demos and then Completely Fall Apart in Real Use
Researchers from UIUC, Stanford, and collaborators propose a formal framework for adapting agentic AI systems that sit on top of large language models and tools. They classify existing work into 4 paradigms based on what is optimized, the agent or tools, and where the supervision signal comes from, tool execution or final outputs. The survey connects methods like Toolformer, DeepRetrieval, s3, and AgentFlow into a single adaptation landscape and shows why future systems will mix occasional agent level updates with frequent adaptation of retrievers, searchers, simulators, and memory… Read the full insights/article here.
Project Notebooks/Tutorials
▶ [Open Source] Rogue: An Open-Source AI Agent Evaluator worth trying Codes & Examples
▶ How to Build a Proactive Pre-Emptive Churn Prevention Agent with Intelligent Observation and Strategy Formation Codes Tutorial
▶ How to Design a Gemini-Powered Self-Correcting Multi-Agent AI System with Semantic Routing, Symbolic Guardrails, and Reflexive Orchestration Codes Tutorial
▶ How to Design a Fully Local Agentic Storytelling Pipeline Using Griptape Workflows, Hugging Face Models, and Modular Creative Task Orchestration Codes Tutorial
▶ A Coding Guide to Build a Procedural Memory Agent That Learns, Stores, Retrieves, and Reuses Skills as Neural Modules Over Time Codes Tutorial