Interesting AI Releases: Salesforce WALT, Apple UltraCUA, Google VISTA, and many more...

[UI Agents] Salesforce AI Research Introduces WALT (Web Agents that Learn Tools): Enabling LLM agents to Automatically Discover Reusable Tools from Any Website. WALT converts brittle click sequences into deterministic tools that wrap website functions such as search, filter, sort, post comment, and create listing. The system discovers tool candidates offline, stabilizes selectors, promotes URL and parameter operations, induces input schemas, and registers tools only after end to end validation. On VisualWebArena WALT reports 52.9% average success and on WebArena it reports 50.1%. Toolization cuts actions by about 21.3% and ablations attribute gains to a multimodal DOM parser at plus 2.6% and an external verifier at plus 3.3%.

[Protein Model] Anthrogen Introduces Odyssey: A 102B Parameter Protein Language Model that Replaces Attention with Consensus and Trains with Discrete Diffusion. Odyssey is Anthrogen’s multimodal protein language model family that fuses sequence tokens, FSQ structure tokens, and functional context for generation, editing, and conditional design, it replaces self attention with Consensus that scales as O(L) and reports improved training stability, it trains and samples with discrete diffusion for joint sequence and structure denoising, it ships in production variants from 1.2B to 102B parameters, it claims about 10x data efficiency versus competing models in matched evaluations, and API access is opening for external users to test real design workflows

[Computer-Use Agents] UltraCUA: A Foundation Computer-Use Agents Model that Bridges the Gap between General-Purpose GUI Agents and Specialized API-based Agents. UltraCUA defines a hybrid action policy for computer use agents that mixes GUI primitives with programmatic tool calls, which reduces long brittle action chains and improves reliability. The team builds an automated tool library and a synthetic data engine that produces 17,000 plus verifiable tasks, then trains 7B and 32B models with supervised fine tuning followed by online reinforcement learning. On OSWorld, UltraCUA reports an average 22 percent relative improvement with 11 percent fewer steps. On WindowsAgentArena, the 7B model reaches 21.7 percent without Windows specific training, which shows cross platform transfer.....

[AI Agent Research] Google AI Introduces VISTA: A Test Time Self Improving Agent for Text to Video Generation. VISTA is a multi agent framework that improves text to video generation during inference, it plans structured prompts as scenes, runs a pairwise tournament to select the best candidate, uses specialized judges across visual, audio, and context, then rewrites the prompt with a Deep Thinking Prompting Agent, the method shows consistent gains over strong prompt optimization baselines in single scene and multi scene settings, and human raters prefer its outputs.

_{You should not miss this one}

[Open Source AI Agent] PokeeResearch-7B: An Open 7B Deep-Research Agent Trained with Reinforcement Learning from AI Feedback (RLAIF) and a Robust Reasoning Scaffold. PokeeResearch-7B is a 7B deep research agent that combines Reinforcement Learning from AI Feedback with an RLOO policy gradient and a chain of thought, multi call scaffold that adds self verification and recovery. It runs web search and page reading through a local tool server that uses Serper and Jina, then synthesizes multiple research threads at test time. The release targets semantic correctness, citation faithfulness, and instruction adherence, reports mean at 4 accuracy across 10 text benchmarks, and shows larger gains on GAIA, HLE, and BrowseComp. Code and weights are public under Apache 2.0.

▶ How I Built an Intelligent Multi-Agent Systems with AutoGen, LangChain, and Hugging Face to Demonstrate Practical Agentic AI Workflows Codes Tutorial

▶ A Coding Guide to Build a Fully Functional Multi-Agent Marketplace Using uAgent Codes Tutorial

▶ A Coding Implementation of Secure AI Agent with Self-Auditing Guardrails, PII Redaction, and Safe Tool Access in Python Codes Tutorial

▶ An Implementation to Build Dynamic AI Systems with the Model Context Protocol (MCP) for Real-Time Resource and Tool Integration Codes Tutorial

▶ How to Build an Advanced Voice AI Pipeline with WhisperX for Transcription, Alignment, Analysis, and Export? Codes Tutorial

▶ How to Design a Fully Functional Enterprise AI Assistant with Retrieval Augmentation and Policy Guardrails Using Open Source AI Models Codes Tutorial