AGENTS · AUTONOMOUS · VERTICAL
The application layer — Harvey, Sierra, Devin, Adept, and the long tail of vertical AI agents.
Legal AI · Litera competitor.
Customer-service agents — Bret Taylor's startup.
Software-engineering agent — Cognition AI.
Action transformer — partially absorbed into Amazon.
Code-generation agent inside Replit.
Browser-driving agent.
Foundation models (Claude, ChatGPT, Gemini) get the headlines, but the dollars are increasingly flowing into the application layer — agents that do specific work end-to-end. OSS tracks the vertical leaders (Harvey for legal, Sierra for customer service, Devin for engineering, Adept for browser automation) plus the general-purpose contenders.
Cognition AI Estimates AI Software Engineer Productivity
Cognition AI has developed a system to estimate the productivity of its autonomous AI software engineer, Devin, by measuring how many engineering hours each session is worth. The company faced challenges in predicting real ROI but focused on estimating human engineering hours saved. After refining their model, they achieved an r-squared value of 0.54, indicating a reasonable level of accuracy for aggregated estimates. The system classifies sessions as productive or not and compares them to human estimates. This approach helps translate AI output into business value using engineering salaries.
Cognition AI — Blog (Devin)
Endava Redesigns Software Delivery with AI Agents
Endava is leveraging AI agents, ChatGPT Enterprise, and Codex to transform software delivery. The company aims to accelerate workflows and foster an AI-native culture across enterprises. This approach focuses on automation and efficiency in software development processes.
OpenAI — News
Microsoft's Project Solara: An Android OS for AI Agents
At Build 2026, Microsoft introduced Project Solara, a new Android-based operating system designed to run AI agents rather than traditional apps. Based on an open-source build of Android (AOSP), Solara aims to generate contextual user interfaces dynamically, adapting to different devices and user needs. The platform is part of Microsoft’s broader push into AI and is intended to simplify the development of specialized hardware by offloading interface creation to AI agents. While still in the concept phase, Solara represents a shift in how Microsoft envisions future computing, with AI handling much of the user interaction and device functionality.
Ars Technica
Devin Desktop and the multi-agent future of software engineering
Cognition AI has announced Devin Desktop, an evolution of Windsurf designed for managing multiple AI agents in software development. The platform allows developers to oversee local and cloud agents, streamline workflows, and collaborate through shared context. It supports the Agent Client Protocol (ACP), enabling compatibility with various agents and maintaining full IDE functionality.
Cognition AI — Blog (Devin)
Posts to your status feed
Pick the closest match below, edit the body, and post. Your report carries the #agents tag automatically so it surfaces here + in the trending-tags rail. Reporting also follows AI Agents so you’ll get status updates.
Enterprise contracts (Harvey/Sierra) · Devin $500/mo · Replit Agent in Core/Teams plans
Best-effort summary — confirm on the provider's own pricing page; tiers and prices drift.
Plaud
Plaud Note
Credit-card-thin AI voice recorder — magnetises to the back of an iPhone, records call + room audio, ships transcripts + GPT-4/Claude summaries via the Plaud app.
Plaud
Plaud NotePin
Wearable pendant variant of the Plaud Note — clip it on, button-press records, AI summary on the dashboard. Direct rival to the Limitless Pendant.
HiNotes
HiNotes AI Recorder
Plaud-style portable AI recorder with chapter summaries + action-item extraction. Differentiator vs Plaud is the per-user mode lock for sensitive meetings.
Claude Code sessions are no longer trapped in one terminal. Because a session is tied to your Claude account, it follows you across your devices — start a task at your desk, check on it from your phone, steer it from the couch. What it is, where it runs (CLI, desktop, web, IDE), and how to use the everywhere part — including why coding you can supervise from anywhere is a real accessibility win.
Jul 17, 2026
xAI's Voice Agent Builder is a no-code platform that lets you create and deploy human-like voice agents that answer full phone calls — in under two minutes.
Jul 12, 2026
Cognition AI Estimates AI Software Engineer Productivity
Cognition AI has developed a system to estimate the productivity of its autonomous AI software engineer, Devin, by measuring how many engineering hours each session is worth. The company faced challenges in predicting real ROI but focused on estimating human engineering hours saved. After refining their model, they achieved an r-squared value of 0.54, indicating a reasonable level of accuracy for aggregated estimates. The system classifies sessions as productive or not and compares them to human estimates. This approach helps translate AI output into business value using engineering salaries.
Cognition AI — Blog (Devin)
Endava Redesigns Software Delivery with AI Agents
Endava is leveraging AI agents, ChatGPT Enterprise, and Codex to transform software delivery. The company aims to accelerate workflows and foster an AI-native culture across enterprises. This approach focuses on automation and efficiency in software development processes.
OpenAI — News
Microsoft's Project Solara: An Android OS for AI Agents
At Build 2026, Microsoft introduced Project Solara, a new Android-based operating system designed to run AI agents rather than traditional apps. Based on an open-source build of Android (AOSP), Solara aims to generate contextual user interfaces dynamically, adapting to different devices and user needs. The platform is part of Microsoft’s broader push into AI and is intended to simplify the development of specialized hardware by offloading interface creation to AI agents. While still in the concept phase, Solara represents a shift in how Microsoft envisions future computing, with AI handling much of the user interaction and device functionality.
Ars Technica
Devin Desktop and the multi-agent future of software engineering
Cognition AI has announced Devin Desktop, an evolution of Windsurf designed for managing multiple AI agents in software development. The platform allows developers to oversee local and cloud agents, streamline workflows, and collaborate through shared context. It supports the Agent Client Protocol (ACP), enabling compatibility with various agents and maintaining full IDE functionality.
Cognition AI — Blog (Devin)
Nvidia Launches AI-Powered 'Superchip' for Laptops and PCs
Nvidia has introduced a new 'superchip' designed to bring AI capabilities to laptops and desktops running Microsoft Windows. The RTX Spark PC chip aims to enable AI agents to replace traditional input methods like the mouse and keyboard. This development marks a new phase in the competition for AI chip dominance, positioning Nvidia against major rivals such as Intel, Apple, Qualcomm, and AMD.
The Guardian — Technology
Sierra Introduces Agent-Building AI
Sierra has launched a new platform feature called Ghostwriter, an agent-building AI that simplifies the process of creating customer experience agents. With Ghostwriter, users can define what they want and let the agent deliver it, eliminating manual tasks such as editing journeys and writing integrations. The feature is part of Sierra's reimagining of its platform around AI-powered agents, following the impact of coding agents like Codex and Claude Code on software development.
Sierra — Blog
Verifying Agentic Development at Scale
The team has expanded Devin's computer use capabilities, enabling it to take screenshots, move the mouse, and interact with applications. This has unlocked new capabilities for Devin, including testing its own work and confirming changes actually work. Automated cloud testing has saved an enormous amount of time, but increasing reliability was not smooth and required addressing failure modes.
Cognition AI — Blog (Devin)
Harvey Mobile: AI-Powered Legal App
Harvey has introduced Harvey Mobile, a legal AI app for iOS and Android devices. The app allows users to access Harvey's products on-the-go, enabling them to capture new information, get up to speed, and keep work moving from anywhere. With Harvey Agents, the app can execute complex legal work end-to-end, freeing up lawyers to focus on high-value tasks.
Harvey AI — Blog
Code Shapes AI Agent Behavior
A new review paper suggests that the software layer surrounding a language model, rather than the model itself, is the key to creating autonomous AI agents. This 'harness' includes tools, memory, testing, and permission boundaries that enable stateless models to function as working agents. A dedicated team at Deepseek in Beijing is already exploring this concept, with a core formula that supports the thesis: model + harness = AI agent.
The Decoder
AI Coding Agent Not Meant to Replace Humans
Cognition's AI coding agent Devin is not designed to replace human programmers. According to Scott Wu, Devin's purpose is not to supplant humans.
TechCrunch — AI
Robinhood Opens Trading to AI Agents
Robinhood is allowing traders to create separate accounts for AI agents, enabling automated stock trading. This feature lets users add a specific amount of money to the account, giving the agent control over buying and selling stocks across the market. Robinhood sees this as an opportunity for traders to explore new investment strategies.
The Verge — AI
Critical Vulnerability Found in Popular Open Source Package
A critical vulnerability known as 'BadHost' has been discovered in Starlette, a popular open source package used by millions of AI agents. The vulnerability poses a significant risk to the security and integrity of these systems. With over 325 million weekly downloads, the impact of this vulnerability could be substantial.
Ars Technica
Claude Code sessions are no longer trapped in one terminal. Because a session is tied to your Claude account, it follows you across your devices — start a task at your desk, check on it from your phone, steer it from the couch. What it is, where it runs (CLI, desktop, web, IDE), and how to use the everywhere part — including why coding you can supervise from anywhere is a real accessibility win.
Jul 17, 2026
No articles yet in AI Agents. Check back soon, or browse all sections.
No forum thread for AI Agents yet — start one with the button above.
Loading edit history…
xAI's Voice Agent Builder is a no-code platform that lets you create and deploy human-like voice agents that answer full phone calls — in under two minutes.
Jul 12, 2026
Related
xAI's Voice Agent Builder is a no-code platform that lets you create and deploy human-like voice agents that answer full phone calls — in under two minutes.
Related
Gardening is one of the most-cited losses when mobility declines. Automated watering, raised hydroponic towers, and robot mowers can hand it back — here's the honest map, subscription traps included.
