Alibaba has open-sourced Qwen-UI-Agent, a GUI agent foundation model that operates across mobile, desktop, web, and deep-search environments on real hardware rather than relying on simulation. It achieves top benchmark results: 82.1% on MobileWorld, 79.5% on OSWorld-Verified, and first on WebArena. It also introduces MobileWorld-Real, a 400+ task benchmark on 100+ phones and 150+ apps, with a 92.2% success rate.
- Supports command-line execution alongside standard GUI operations and batches multiple actions into a single decision step to shorten trajectories.
- Built-in safety layer refuses illegal or high-risk requests outright and pauses at sensitive operations (payments, data deletion, privacy grants) for explicit user confirmation.
- Trained via online reinforcement learning on trajectories exceeding 100 steps, paired with adaptive curriculum learning to progressively tackle longer tasks.
- Supports command-line execution alongside standard GUI operations and batches multiple actions into a single decision step to shorten trajectories.
- Built-in safety layer refuses illegal or high-risk requests outright and pauses at sensitive operations (payments, data deletion, privacy grants) for explicit user confirmation.
- Trained via online reinforcement learning on trajectories exceeding 100 steps, paired with adaptive curriculum learning to progressively tackle longer tasks.
Nous Research has bundled Bot Mode as default in Hermes Agent v0.20.3 for Hermes Desktop, replacing the single-agent session list with a roster of named bots that each have their own chat, memory, skills and pinned model for multi-agent workflows on the desktop.
Key technical details:
* Standalone repo archived; active development continues in-tree at `apps/desktop/src/plugins/hermes-bots/`, first shipped as a one-day public beta plugin by co-founder Teknium
* Bots are stored as isolated profiles under `~/.hermes/profiles//` with own config, memory, skills, credentials and history; managed via existing `profiles.*` gateway RPCs `list, create, describe, configure`
* Avatars generated via the `image.generate` RPC; options include geometric faces in seven shapes/ten colors or uploaded/generated portraits/pixel pets
* Routines are ordinary Hermes cron jobs namespaced ` bot: » `; bot-to-bot messaging uses a persistent Agent Inbox with CLI handoffs e.g. `hermes -p chat -c "Agent Inbox" -q "..."` and `@mention` handoffs
* Per-bot config via New Agent dialog supports cloning a profile, pinning provider/model, custom SOUL.md and skills; duplicates clone config, skills, SOUL.md, memory and appearance
* Bundled version adds Groups for roster sections synced across machines, Group chats for 2-6 bots with up to three serial rounds, and a multi-source roster pulling bots from every connection under Settings → Connections
Key technical details:
* Standalone repo archived; active development continues in-tree at `apps/desktop/src/plugins/hermes-bots/`, first shipped as a one-day public beta plugin by co-founder Teknium
* Bots are stored as isolated profiles under `~/.hermes/profiles//` with own config, memory, skills, credentials and history; managed via existing `profiles.*` gateway RPCs `list, create, describe, configure`
* Avatars generated via the `image.generate` RPC; options include geometric faces in seven shapes/ten colors or uploaded/generated portraits/pixel pets
* Routines are ordinary Hermes cron jobs namespaced ` bot: » `; bot-to-bot messaging uses a persistent Agent Inbox with CLI handoffs e.g. `hermes -p chat -c "Agent Inbox" -q "..."` and `@mention` handoffs
* Per-bot config via New Agent dialog supports cloning a profile, pinning provider/model, custom SOUL.md and skills; duplicates clone config, skills, SOUL.md, memory and appearance
* Bundled version adds Groups for roster sections synced across machines, Group chats for 2-6 bots with up to three serial rounds, and a multi-source roster pulling bots from every connection under Settings → Connections
Michael Larabel writes that Meta Superintelligence Labs announced the release of Muse Glimmer, a 30-billion-parameter open model for always-on local agent workflows with weights released under Apache 2.0. The model is sized to run on a single consumer GPU and targets local coding agents, LLM-as-a-judge evaluation and similar uses, having been trained and evaluated for end-to-end agentic task completion, multi-step reasoning and optimized local deployment. Details are posted on research.meta.ai and the model is available on Hugging Face, with Ollama 0.32.7 already adding support.
-
-
Ashwini Sinha explains how ESP-Claw turns an ESP32 into a self-programming agent through natural language instructions. By integrating LLMs, the system can automatically generate code and flash firmware to perform tasks such as managing sensor data or controlling peripherals without manual coding. This setup effectively transforms conventional microcontrollers into intelligent embedded systems capable of autonomous updates and performance optimization.
- Minimum hardware requirements include 8 MB PSRAM and 8 MB Flash memory on the ESP chip.
- The agent can access information via web search capabilities to answer specific queries.
- Setup requires a USB-C cable for initial firmware flashing and programming.
- Minimum hardware requirements include 8 MB PSRAM and 8 MB Flash memory on the ESP chip.
- The agent can access information via web search capabilities to answer specific queries.
- Setup requires a USB-C cable for initial firmware flashing and programming.
Skill Recorder converts screen activity into repeatable agent tasks by recording user sessions and reconstructing intent via GitHub Copilot. Captured data includes clicks, window changes, URLs, and optional narration to produce structured Skills or Automations.
- Built with Electron.
- Performs on-device transcription using Whisper for 99 languages.
- Limits cloud uploads to specific metadata during the analysis phase.
- Built with Electron.
- Performs on-device transcription using Whisper for 99 languages.
- Limits cloud uploads to specific metadata during the analysis phase.
Anurag Singh writes that Cursor’s strength is its capacity for deep project understanding through efficient context management rather than simple model speed. Instead of pushing a whole repository into an LLM's window, the IDE employs search tools to provide only relevant code segments as tasks arise. This curated approach ensures models receive high-quality information and instructions tailored to their specific capabilities, allowing them to maintain existing architectural patterns more effectively.
* The tool uses Instant Grep via a local text index instead of scanning every file from scratch.
* Cursor is significantly more effective in complex or large projects where it can identify and reuse established implementations.
* Its search ability helps prevent the common problem of generating redundant helper functions that already exist within the codebase.
* The tool uses Instant Grep via a local text index instead of scanning every file from scratch.
* Cursor is significantly more effective in complex or large projects where it can identify and reuse established implementations.
* Its search ability helps prevent the common problem of generating redundant helper functions that already exist within the codebase.
This article discusses how the newest generation of Claude models necessitates a shift in context engineering. Anthropic found that they could remove over 80% of their system prompts without losing performance on coding evaluations by moving away from rigid, often conflicting instructions and instead allowing the model's inherent judgment to guide its behavior based on surrounding context.
Key shifts in methodology include:
- Moving from strict rules to letting models use judgement for nuance such as documentation style or intent interpretation.
- Prioritizing intuitive tool interface design over providing restrictive examples that limit exploration.
- Implementing progressive disclosure by using skills and deferred loading to manage large context windows efficiently.
- Replacing repetitive instructions with streamlined, high-fidelity descriptions directly within tool definitions.
- Transitioning from manual memory management in files toward auto-memory and rich references like HTML artifacts or code snippets.
Key shifts in methodology include:
- Moving from strict rules to letting models use judgement for nuance such as documentation style or intent interpretation.
- Prioritizing intuitive tool interface design over providing restrictive examples that limit exploration.
- Implementing progressive disclosure by using skills and deferred loading to manage large context windows efficiently.
- Replacing repetitive instructions with streamlined, high-fidelity descriptions directly within tool definitions.
- Transitioning from manual memory management in files toward auto-memory and rich references like HTML artifacts or code snippets.
A distillation of the Claude Fable 5 workflow into actionable skills designed to guide AI agents through a systematic think, act, and prove methodology. The framework improves agentic reliability by enforcing specific sequences like classifying tasks, gathering parallel evidence from primary sources, making surgical edits rather than broad changes, and undergoing adversarial verification via an automated judge. It includes specialized domain adapters for sectors such as coding, research, marketing, and DevOps to tailor the reasoning process to specific professional requirements. The method is specifically designed to mitigate common LLM failures like reward hacking, silent code errors in tests, and false claims of task completion.
>"Before Fable 5 was deprecated, it wrote down its own problem-solving method. Step by step. How it classifies a task, defines "done," gathers evidence, commits to one recommendation, makes the smallest correct change, verifies by observation, and reports the outcome honestly."
- Core skills: fable-method (thinking), fable-loop (orchestration), fable-judge (verification), and fable-domain (adapter generation)
- Focuses on preventing unauthorized staging or "reward hacking" through strict observation cycles
- Validated against 260+ agent runs using blind LLM judges to verify results via code execution rather than reports
>"Before Fable 5 was deprecated, it wrote down its own problem-solving method. Step by step. How it classifies a task, defines "done," gathers evidence, commits to one recommendation, makes the smallest correct change, verifies by observation, and reports the outcome honestly."
- Core skills: fable-method (thinking), fable-loop (orchestration), fable-judge (verification), and fable-domain (adapter generation)
- Focuses on preventing unauthorized staging or "reward hacking" through strict observation cycles
- Validated against 260+ agent runs using blind LLM judges to verify results via code execution rather than reports
ESP-Claw is an AI agent framework by Espressif that enables ESP32 series chips to function as active decision-making centers rather than passive executors. It implements a full intelligence loop involving sensing, reasoning, and execution locally on edge hardware. Through conversational programming via chat interfaces, users can define device behaviors using dynamic Lua loading without traditional coding expertise.
- Event-driven architecture with millisecond response times
- Structured memory management that prioritizes privacy by keeping data off the cloud
- MCP communication support allowing devices to act as both a client and a server
- Broad compatibility with various LLM APIs including OpenAI, Anthropic, Qwen, and DeepSeek
- Support for multiple development boards via web-based flashing or local builds
- Event-driven architecture with millisecond response times
- Structured memory management that prioritizes privacy by keeping data off the cloud
- MCP communication support allowing devices to act as both a client and a server
- Broad compatibility with various LLM APIs including OpenAI, Anthropic, Qwen, and DeepSeek
- Support for multiple development boards via web-based flashing or local builds
OpenTag is an open-source, self-hosted alternative to Claude for Slack. It enables users to run AI agents directly within Slack threads that can read conversations, answer queries, execute tools, and render rich, generative UI elements like tables and bar charts. Built on the CopilotKit SDK, it allows for complete ownership of the runtime and model without per-seat pricing or vendor lock-in.
- Open-source Slack AI agent
- Generative UI for rich message rendering
- Human-in-the-loop approval gates
- Multi-platform support via adapters (Discord, Telegram, WhatsApp)
- Self-hosted architecture for privacy and control
- Open-source Slack AI agent
- Generative UI for rich message rendering
- Human-in-the-loop approval gates
- Multi-platform support via adapters (Discord, Telegram, WhatsApp)
- Self-hosted architecture for privacy and control
