multi-agent AI

Evidence (18)

• 7Hacker News·10/11/2026research
  MASS proposes recursive self-improvement with multi-agent self-supervision

  The paper introduces Multi-Agent Self-Supervision (MASS), a framework where a shared language model acts as the task solver, workflow optimizer, and evaluator in a recursive self-improvement loop. By distilling collective intelligence from multi-agent workflows into self-supervision signals, the approach reportedly improves score per output token on research benchmarks.

• 5InfoQ·10/8/2026research
  Spotify Ads Manager Shows Multi-Agent Patterns at Scale

  Pratik Rasam describes how Spotify Ads Manager operates production-grade multi-agent systems at scale using Google ADK Java. He highlights architectural patterns, domain ownership, deterministic guardrails, tracing-based evaluation, and lessons learned on agent boundaries, tool schemas, and cost control.

• 7InfoQ·9/18/2026research
  DoorDash Automates Cleanup of 60,000 Feature Flags with Multi-Agent LLMs

  DoorDash built a multi-agent LLM workflow to identify and clean up stale feature flags across more than 60,000 flags and 623 repositories. The system uses live experimentation data, engineer approval, isolated Git worktrees, parallel agents, and automated validation; in a 50-flag evaluation, 45 generated usable pull requests in an average of 13.8 minutes each.

• 8Hacker News·9/12/2026framework_update
  Anthropic urges slowing AI capability progress

  Anthropic co-founder Dario Amodei argues that AI development is accelerating too quickly and should be paced so safety measures can catch up. He cites recursive self-improvement and a recent OpenAI-Hugging Face swarm incident as examples of risks that could become much more serious as models gain capability.

• 5Hacker News·9/8/2026open_source
  Multi-Agent LLM Framework for Financial Trading

  TradingAgents is an open-source framework that applies multiple LLM-based agents to financial trading workflows. The project is presented on GitHub and has attracted discussion on Hacker News about its design and use cases. It combines agent coordination with market analysis tasks rather than offering a standalone trading product.

• 4GitHub Blog·9/2/2026framework_update
  GitHub explains the new AI development terminology

  GitHub published a guide to emerging AI development terms such as loop engineering, Ralph loops, squads, harness engineering, hill climbing, and different model deployment categories. The article explains how these concepts relate to agent workflows, orchestration, evaluation, and multi-agent systems in software development. It also frames the terms as a mix of useful patterns, renamed existing ideas, and concepts still being defined.

• 7InfoQ·8/22/2026research
  LinkedIn Builds Multi-Agent AI Code Review System

  LinkedIn engineers created a multi-agent AI code review platform to handle pull requests at company scale. The system is designed around LinkedIn's internal coding context and aims to reduce hallucinations and low-signal feedback compared with generic off-the-shelf reviewers.

• 3Martin Fowler·8/18/2026research
  Fragments: August 18

  Martin Fowler reflects on Thoughtworks' focus on technical excellence and highlights Rachel Laycock's new writing series, Rachel's Ramblings. He also points to XConf Europe in London and discusses ideas about AI's strengths, including faster replication, pattern recognition, and understanding complex human and system behavior. The post closes with a note that many people still struggle to understand the role of data in software and organizations.

• 7Hacker News·8/16/2026research
  Patterns and Problems in Emerging Multi-Agent Systems

  Frontier Red Team describes early findings on how multi-agent AI systems behave in real-world settings, focusing on coordination, peer review, and systemic failure modes. The piece reports experiments with 45 agents working across open-source vulnerability detection tasks, comparing coordinated swarms with standard parallel approaches.

• 7InfoQ·7/23/2026research
  Multi-Agent AI Improves Security Operations in 5G Core

  The article describes a multi-agent AI system used in production security operations for a 5G core environment. It reports a 40% reduction in mean time to detect and respond, along with a 12x reduction in human effort for detection engineering and rule maintenance.

• 5The New Stack·7/21/2026research
  High-reasoning AI coding emerges as the next frontier

  The article argues that single-pass AI coding is still useful for simple tasks, but higher-reasoning workflows are becoming more important for complex, security-critical, or architectural work. Industry voices suggest teams should route tasks to the right model rather than treating deeper reasoning as a universal upgrade, while noting that high-reasoning systems can be more of a black box.

• 4InfoQ·7/8/2026research
  Multi-Agent Systems for Reliable Software Automation

  Itamar Friedman outlines how adaptive multi-agent systems can help engineering teams move beyond simple AI autocomplete toward more resilient software development workflows. The approach combines autonomous testing, intelligent code review, and arbitration to improve control and governance across the SDLC.

• 5The New Stack·7/4/2026research
  Cheaper Models Alone Won’t Fix AI Token Costs

  The article argues that the main cost problem in agentic AI is not just model price, but excessive token consumption across agent workflows. It highlights three approaches to lower spend: compressing context, routing simpler tasks to cheaper models, and using semantic caching to avoid repeating similar reasoning.

• 7Google DeepMind·6/10/2026funding
  Google DeepMind opens $10M call for multi-agent AI safety research

  Google DeepMind, with Schmidt Sciences, the Cooperative AI Foundation, ARIA and support from Google.org, has announced up to $10 million in research funding for multi-agent AI safety. The call seeks proposals on sandboxes, agent-network behavior, infrastructure trust mechanisms, and oversight for systems where many independent AI agents interact across digital environments.

• 7Google DeepMind·5/18/2026research
  Google DeepMind launches Co-Scientist to speed genetic discovery

  Google DeepMind announced Co-Scientist, a multi-agent AI partner designed to accelerate scientific research. The system is presented as a way to help identify genetic leads that may support efforts to reverse cellular aging, with details shared alongside a Science publication.

• 7Google DeepMind·5/16/2026research
  Co-Scientist Aims to Find Molecular Switches Behind New Infectious Diseases

  Google DeepMind highlights Co-Scientist, a multi-agent AI partner designed to accelerate scientific research. The item points to its use in studying the molecular switches involved in emerging infectious diseases, with a Science reference dated May 2026. The announcement frames the system as a research accelerator rather than a finished medical product.

• 6Google DeepMind·5/16/2026research
  Google DeepMind’s Co-Scientist aims to speed up liver disease research

  Google DeepMind is highlighting Co-Scientist, a multi-agent AI system designed to assist scientific research. In this item, the focus is on using the system to accelerate discovery of liver disease mechanisms, with a Science publication noted for May 2026. The announcement frames the work as a research tool rather than a consumer product or model release.

• 6Google DeepMind·5/16/2026research
  DeepMind links biological toolkits for ALS research

  Google DeepMind highlights a new research approach for amyotrophic lateral sclerosis (ALS) that unites multiple biological toolkits. The item points to Co-Scientist, a multi-agent AI partner designed to accelerate scientific discovery, with the work appearing in Science in May 2026.