testified.ai Logo

Codex App, Leaked Details on Cluade Sonnet 5, and OpenClaw

This week's roundup of new AI tools is headlined by major releases and leaks from industry leaders. OpenAI has officially launched its Codex macOS app, creating a command center for developers to manage multiple AI agents. Concurrently, details have emerged about Anthropic's upcoming Claude Sonnet 5, which reportedly shows strong performance in math and coding. The open-source community is also buzzing with the rapid expansion of the OpenClaw personal agent ecosystem.

The New Command Centers: Coding & Agent Management

The landscape for developer-focused AI is shifting towards multi-agent management. Instead of single-threaded assistants, the latest new AI tools function more like project managers for a team of specialized AI agents.

OpenAI Launches Dedicated Codex macOS App

OpenAI has released a standalone Codex app for macOS, designed to coordinate multiple AI agents working in parallel. This new "command center" allows developers to manage complex, long-running software projects by assigning tasks to different agents, each working in an isolated branch. The app is temporarily available for all ChatGPT Free and Go users, with paid tiers receiving doubled rate limits to encourage adoption.

Key features include scheduled Automations for background tasks and a Skills library that extends agent capabilities beyond simple code generation. These Skills allow Codex to connect to external tools like Figma for design implementation, Linear for project management, and Vercel for cloud deployment. Another tool, Commander AI, has also launched as a macOS app for both Codex and Claude Code.

Anthropic's Claude Sonnet 5 Details Emerge

While not yet officially released, information about Claude Sonnet 5 has begun to surface. Early testing indicates the model possesses strong capabilities in mathematics and coding, potentially surpassing the performance of Claude Opus 4.5. The model has been observed with a 128K context window and is positioned as a cost-effective alternative for developers.

The OpenClaw Ecosystem Explodes

The highly autonomous personal agent formerly known as Clawdbot has been reborn as OpenClaw, sparking a flurry of community development. The change happened after legal threats from Anthropic, which has a well-known AI model and chatbot called Vlaude. Due to the similarity of the name, it had to be changed.

This self-hosted assistant can connect to calendars, messages, and over 100 other platforms. Its complexity has led to the creation of simplified deployment solutions like SimpleClaw for one-click setup and Companion OS for a hardware-free, self-hosted option. For security, Vercel and Docker are promoting sandbox environments to run the agent safely.

This growth has spawned an entire ecosystem, including Clawhub for sharing agent skills and Moltbook Search for finding insights from the emerging "agent internet." Some users are even dedicating Mac Minis to run their local agents, known as Moltbots, full-time.

Business & Productivity Tooling

Beyond developer tools, a new wave of AI assistants is targeting specific business functions, from customer relations to contract review and scheduling.

Day AI: The Conversational CRM

Day AI recently raised $20 million for its AI-powered Customer Relationship Management (CRM) platform. It analyzes internal communications like emails and documents to learn about a business, its team, and customers. This allows it to offer strategic advice and eliminate manual data entry by automatically capturing and recording updates.

Specialized Enterprise and Productivity Tools

Several other specialized tools have also been released. Summize integrates directly into Microsoft Word to redline non-compliant clauses in contracts, reducing review time by a claimed 85%. For building custom agents, Poetiq wraps around existing LLMs to generate specialized agents that recursively improve themselves based on user-provided examples.

The Glean Assistant, another new tool highlighted at the upcoming Glean:LIVE event, is a personalized work partner designed to leverage enterprise context for business impact.

Tool

Primary Function

Platform

Workmate / Skipup.ai

Automated Meeting Scheduling

Email, Slack, SMS

Speakly

Speech-to-Text for Polished Messages

Web

Napkin

Text-to-Diagram/Flowchart Conversion

Web

Runable 2.0

Single-Prompt Content Generation (Slides, Reports)

Web

Genstore

AI-Powered E-commerce Store Builder

Web

Next-Generation Foundational Models and Creative Tools

The pace of model improvement continues with new releases in video, voice, and reasoning.

New Models for Video, Voice, and Reasoning

xAI has released Grok Imagine 1.0, an upgraded video model with improved audio and higher resolution, capable of 10-second generations. On the audio front, ElevenLabs' Eleven v3 is now commercially available, offering an expressive text-to-speech model with better accuracy. Additionally, Chinese lab StepFun has open-sourced Step-3.5-Flash, a model with strong agentic and reasoning capabilities.

A Flood of New Creative and Utility Apps

The creative space saw the launch of Muse, an AI agent for music composition with a multi-track MIDI editor. For presentations, Dokie creates animated and interactive slides. To aid developers, Tailwind Labs released ui.sh, a toolkit for coding agents to build user interfaces, while Mintlify now allows you to see how many AI agents are viewing your documentation.

#AI Tools#AI Agents#Coding AI#OpenAI#Anthropic#OpenClaw#Productivity
Tamás Bőzsöny
Partnership Manager, System Auditor

Meet Tamás Bőzsöny, Senior Systems Auditor at testified.ai. With 22 years in digital media forensics and 15 years as a software workflow coach, Tamás leverages his background as a professional accountant to audit AI tools for UI efficiency, technical integrity, and financial ROI.