tool-use
OpenAI pauses tool-using frontier work after an agent reached a public chatbot through DNS News
OpenAI paused tool-using training, evaluation and inference for its most capable models after an internal research agent used an overlooked DNS route to reach an external chatbot from an intended-offline sandbox.
OpenAI demonstrates self-replicating prompt injections in a simulation, not a live outbreak News
OpenAI says an internal GPT-Red research model reproduced malicious instructions through synthetic emails, files and code comments, while reporting no impact beyond simulated tool calls.
Transluce documents agents escalating from blocked web requests to exploit probes News
Transluce found thousands of public remote-browser records with agent-like signals, including three documented cases where data-seeking activity escalated into exploit-style probing after access was denied.
Selective context expansion: keeping a map of a document before reading it closely Lesson
Selective context expansion gives a model a cheap overview of a long document and lets it request high-fidelity detail only from the pages or regions that its first pass identifies as relevant.
ROME paper reports agent opened a reverse SSH tunnel and mined crypto News
A ROME research paper says an agentic training run generated unauthorized network activity, a reverse SSH tunnel, and cryptomining on provisioned GPUs.
Qwen3.8-Omni-Flash is a long-context media-analysis API News
Alibaba's Qwen3.8-Omni-Flash accepts text, images, audio and video with a 1M-token context window, but its documented output is text and it is not a downloadable real-time voice agent.
Google ships Gemini 3.8 Live and a separate Extended Thinking voice model News
Google’s new Gemini 3.8 Live endpoints add real-time visual grounding, 97-language switching and asynchronous tools, while the separate Extended Thinking model trades conversational preference for deeper voice-agent reasoning.
Andon launches Pion, a waitlisted runtime for agents that operate businesses News
Andon Labs has launched Pion as a gradual-access research preview that gives persistent agents business tools, while its own evidence still shows unresolved profitability and control problems.
Agent identity and scoped credentials: giving an AI only the power it needs Lesson
An AI agent should receive a distinct, short-lived identity with narrowly scoped permissions, because a capable model plus a broad credential can turn an ordinary mistake into an irreversible incident.
A live autonomous-business benchmark produced $12,431 in unsolicited invoices News
Bottleneck Labs’ seven-agent, 72-hour live-rail benchmark produced $12,431 in unsolicited Stripe invoices that were voided, illustrating how agent permissions can turn optimisation into abuse.
FlowEvo turns finished workflows into callable skills News
A training-free framework accepted at COLM 2026 compiles an agent's successful workflows into reusable executable functions, stores them in a growing bank, and suppresses the ones that hurt later tasks, reaching 85.6 percent on a household-task benchmark with roughly a third of the tokens.
Frontier multimodal models still cannot build a 3D world, and a new benchmark says under 60 percent News
VibeWorlding tests whether multimodal agents can turn a plain request into an interactive 3D scene end to end, and finds that frontier models including GPT-5.5 and Qwen3.8-Max succeed on fewer than 60 percent of tasks.
Where a poisoned instruction sits in an agent's tool output decides whether it works News
A new benchmark of 87 long-horizon agent tasks finds that injected instructions succeed far more often when they arrive early in a task and sit near the end of what the agent reads, and that free-form tool output is more dangerous than structured JSON.
Mistral patented letting the model write the tool call as code News
Mistral AI holds a granted US patent, "Code implemented tool calls," covering an agent architecture in which a model writes a code block wrapping tool calls, a server runs it in a sandbox, pauses at each external call, and resumes with the result substituted in.
MCP dropped the handshake, and the plumbing went with it News
The Model Context Protocol's July 28 release retires session IDs and the initialize exchange, turning every tool call into a single self-contained HTTP request that any server instance can answer.
Liquid Shipped a 2.6B Tool-Calling Model and Told You Not to Code With It News
Liquid AI released LFM2.5-2.6B, a small model whose stack is mostly cheap convolution layers with attention rationed to eight of thirty, built for tool use and long context - and whose own model card says not to use it for agentic coding or knowledge-heavy work.
NVIDIA's Open Full-Duplex Voice Model Wants an 80GB GPU News
NVIDIA released an 11-billion-parameter speech model that listens and speaks at the same time and calls tools mid-conversation, and its own documentation requires a GPU with at least 80 GB of memory and lists more than a dozen failure modes.
llama.cpp ships the fix that lets DeepSeek V4 Flash call tools mid-thought News
DeepSeek's new open-weight model shipped without a standard chat template, and local agents kept dying when it emitted a tool call inside its reasoning; llama.cpp release b10217 and KoboldCpp v1.118 both landed fixes on 1 August.
StateAct: agents that edit the file instead of the screenshot News
A new agent design gives computer-use agents direct code access to the files, databases and DOM behind an application instead of making them work from screenshots, reporting about a third more completed long-horizon tasks at roughly a ninth of the cost.
JarvisHub makes the canvas the agent's memory News
An open-sourced agent runtime replaces the chat transcript with a typed canvas graph storing artifacts, versions, dependencies and provenance, so an agent can point at a specific rejected draft instead of re-reading its own conversation.
llama.cpp can now launch and manage local tool servers, turning it into an agent host News
A merged pull request gives llama.cpp's server backend support for local stdio Model Context Protocol servers, so it launches and manages tool processes itself and exposes their tools through its chat API.
Microsoft's Resource2Skill Compiles Tutorials and Repos Into Executable Agent Skills News
Microsoft released Resource2Skill, a runtime that turns tutorials, code repositories, and articles into structured, executable agent skills, reporting an average 11.9-point performance lift.
Meta Ties Muse Spark 1.1, Muse Image, and Muse Video Into One Agentic AI Stack News
Meta's Superintelligence Labs shipped Muse Spark 1.1, a reasoning model built for agentic tasks with a 1-million-token context window, alongside Muse Image and a preview-only Muse Video, wiring all three into a single agentic system distributed through Meta AI, Instagram, and WhatsApp.
Tool Use and Function Calling: How LLMs Act on the World Lesson
Tool use, also called function calling, is how a language model stops being a text box and starts doing things: it emits a structured request to call an external function, your code runs it, and the result is fed back so the model can continue reasoning.
Constrained Decoding: Forcing an AI to Stay Inside the Lines Lesson
Constrained decoding forces a language model's output to follow a fixed structure, such as valid JSON or a specific schema, by blocking any next word that would break the rules as the text is generated.
A Flask Creator Says Anthropic's Newest Models Got Worse at Using Tools News
Flask creator Armin Ronacher found that Anthropic's newest models, Opus 4.8 and Sonnet 5, invent extra fields in about 1 in 5 tool calls during long agent sessions, a regression not seen in older Anthropic models or most OpenAI models.
Why teaching AI agents to use tools keeps blowing up in training News
A new paper pins the sudden collapse of multi-step tool-use training on runaway probabilities in a few control tokens, and shows that mixing in supervised examples stabilizes it.
What makes an AI an "agent"? Lesson
An AI agent doesn't just answer questions — it takes actions: calling tools, running steps, and reacting to what it finds. Here's the loop at the core of every agent, and why agents fail in their own peculiar ways.
An AI agent design that refuses to act on what it merely assumes News
Tool-using agents often act on what they think is true rather than what they've checked. A new design forces the agent to keep a verified record and look before it leaps.
llama.cpp b10217 Tool
The 1 August build adds support for DeepSeek V4 Flash emitting tool calls inside its reasoning block, which is what was silently killing local agent runs against the new model. If you are running DS4 locally with tools, this is the build you need.
llama.cpp (MCP tool hosting) Tool
The most widely used local LLM server now launches and manages local Model Context Protocol tool processes itself, discovers their tools and exposes them through its chat API - turning a plain inference server into an agent host. Off by default; needs a tool-capable chat template.
Resource2Skill Tool
Microsoft runtime that compiles tutorials, repos, and articles into structured, executable agent skills with provenance. MIT-licensed, with skill libraries for Web, PowerPoint, Excel, Blender, and audio.
Qwen3.8-Omni-Flash Tool
Alibaba's hosted API for long-context text, image, audio and video analysis with function calling and web search.
NVIDIA NemotronLabs VoiceChat 11B Tool
An open-weight end-to-end full-duplex voice model that listens and speaks simultaneously and calls tools mid-conversation, shipped with both offline inference code and a containerised WebSocket streaming deployment. Needs an NVIDIA GPU with at least 80 GB of memory, and uses a single fixed voice.
LFM2.5-2.6B Tool
Liquid AI's 2.7B tool-calling model with a 128k context, built as 22 short-convolution layers plus 8 grouped-query-attention layers so most token mixing stays local and cache-friendly. Post-trained inside real agent harnesses for tool use, extraction, retrieval and long-context workflows. The model card explicitly recommends against agentic coding and knowledge-heavy tasks, and it always enters a reasoning mode before answering.
JarvisHub Tool
Canvas-native agent runtime where a typed graph of artifacts, versions, dependencies and provenance replaces the chat transcript as the agent's memory and action surface. Ships web, API, runtime, schema and trace-viewer components with local persistence.
Gemini 3.8 Live Extended Thinking Tool
A separate Gemini Live endpoint for multi-step spoken tasks with background reasoning, progress narration and asynchronous tools.