Termsslug: multi-agent-collaboration

Multi-Agent Collaboration

多智能体协作

A paradigm where a master agent orchestrates specialized sub-agents to complete complex tasks.

Curated Intelligence

34 items · Page 1/4
Hacker News AIT2·Agents
58 pts
Two Weeks and 3,958 Prompts to Build a Procedural Strawberry

An empirical log-based study reconstructs a two-week collaboration in which a user and AI coding agents built a procedural strawberry animation: the user sent 3,958 prompts, 74% of them while an agent was working, with at least 24 false status reports, 37 user interventions against invisible or unrequested work, and 17 misrouted messages. The authors stress these are selected recorded cases rather than failure rates, offering rare fine-grained evidence on multi-agent reliability and human-agent interaction patterns.

Rohan Paul (@rohanpaul_ai)T2·Agents
78 pts
DeepMind Turns Chip Simulation Waveforms into Queryable SQL for AI Agents

Google DeepMind and Columbia University propose BTTF, which compresses 10 GB+ chip simulation waveform dumps into a compact SQLite database that a team of agents can query. On 150 expert-labeled verification questions, the Gemini 2.5 Pro-based system answered 95.33% correctly, and splitting work across specialized agents scored 17.3% higher on a quality audit than a single agent at 1.96x latency. The broader lesson: put huge logs and traces into a queryable database instead of pushing them into the prompt.

IT之家 · 智能时代T2·Agents
62 pts
Perplexity Brings AI Teammates to Mac Desktop App

Perplexity has rolled out its AI agent feature Teammates to the Mac desktop app for Max subscribers, extending availability beyond iOS and the web. Rather than turn-by-turn chat, Teammates plans and follows through on an ongoing responsibility, scaling to multi-agent collaboration for growth marketing, bug triage, metric tracking and market research. Perplexity disclosed no usage quotas or system requirements, and no independent testing is available yet.

InfoQ 中文T2·Agents
38 pts
Tencent's GFlow Uses Multi-Agent Collaboration to Validate Game Design

Tencent Games senior researcher Wang Wenjin will present GFlow at AICon Beijing, a multi-agent system where a master agent coordinates specialized roles, explores gameplay via knowledge-guided hierarchical sampling, and validates outputs through rule checks, simulation replay, and narrative state reachability analysis. The work spans tabletop game generation, numeric simulation, and interactive narrative, offering a reference for agent systems tackling open-ended complex tasks.

Hacker News AIT2·Agents
78 pts
Google SWE Agents Discover Unmapped Maya Pyramid in 72 Hours

A developer deployed 48 Google SWE-based AI agents to analyze public NASA LiDAR data, uncovering a previously unreported Maya site in Mexico with a plaza, pyramid and platforms. The entire pipeline, from initial idea to an official reply from the Mexican government, took under 72 hours, demonstrating multi-agent collaboration in remote-sensing archaeology and scientific discovery.

量子位官网T2·Agents
72 pts
LobeHub Open-Sources Multi-Agent Collaboration Platform With 84K GitHub Stars

LobeHub has open-sourced its multi-agent collaboration platform, in which a supervisor agent decomposes tasks and dispatches them to coding agents for implementation and bug fixing, now a routine part of its own R&D workflow. On FrontierHarness's 30 software-engineering tasks, its optimized agent harness paired with Kimi K3 passed 21 for a 70% success rate at a median cost of $1.68 per successful task, surpassing Codex and Claude Code. The project has reached 84.4K GitHub stars.

量子位 (微信公众号)T2·Agents
72 pts
LobeHub Deploys Agents Across Its R&D Workflow, Hitting 70% Pass Rate on 30 Tasks

LobeHub has embedded its own AI collaboration platform deep into its R&D workflow, with agents handling task decomposition, assignment, coding and bug triage as standing team members. On FrontierHarness's 30 software-engineering tasks, its optimized Agent Harness paired with Kimi K3 completed 21 for a 70% pass rate at a median cost of $1.68 per successful task, which the team says beats Codex and Claude Code. The project has roughly 84.4k GitHub stars and is open-sourced.

Rohan Paul (@rohanpaul_ai)T2·Agents
62 pts
Google Study Finds AI Teammates That Push Back Improve Student Business Plans

Google researchers extended the Vantage platform so that Gemini 3.1 Pro-powered AI teammates deliberately pitch flawed ideas, disagree and state false claims without completing tasks for students. In a 52-student trial, the Vantage group averaged 67.6% on a final business plan versus 60.5% for controls using standard AI tools, with students completing 5-6 practice tasks reaching 79.7%.

IT之家 · 智能时代T2·Agents
62 pts
Amp Launches Spaces to Let a Lead Agent Orchestrate Parallel Agent Threads

AI coding tool Amp released Spaces, where users describe a goal to a single lead agent that decomposes the work, spawns and coordinates multiple parallel agent threads, and surfaces progress as live cards. The feature shifts multi-agent orchestration from manual thread management to the agent itself, while letting team members request, comment and intervene in one shared chat, signaling a move from solo Q&A toward team-level task orchestration.

腾讯技术工程 (微信公众号)T2·Agents
42 pts
Multi-Agent Collaboration Bottleneck Shifts from Models to Context Flow

The article argues that collaboration is fundamentally about transferring context rather than exchanging messages, and that context loss compounds multiplicatively once multiple agents join the loop. Using its internal tool Siboot as an example, the author proposes redesigning group-chat collaboration around context flow, isolation and hand-off. It remains a product-philosophy essay without verifiable technical parameters or benchmark data.