Building AI Agents

1 Factored cognition

Is factored cognition a field of study, or a company’s marketing term?

Reference: Factored Cognition | Ought.

In this project, we explore whether we can solve difficult problems by composing small and mostly context-free contributions from individual agents who don’t know the big picture.

• Factored Cognition Primer

2 Incoming

• Agency, Intentions, and Artificial Intelligence

  For AI to fulfill its enormous promise, we must ameliorate valid concerns about the risks that these systems pose to humans or even to humanity as they become more autonomous. To do this, it must be recognized that the central concerns are about the intentions of increasingly capable AI systems, especially if AIs develop their own intentions and if those intentions misalign with human values. It is not about AI becoming super-intelligent or conscious. What is more, if AIs acquire intentions, they might have moral responsibility and moral rights.

• Advanced Large Language Model Agents

• Announcing the Agent2Agent Protocol (A2A) - Google Developers Blog

• Introducing smolagents: simple agents that write actions in code.

• Workshop on Agentic AI for Scientific Discovery

• Agent Laboratory: Using LLM Agents as Research Assistants

  Agent Laboratory takes input from a human-produced research idea and outputs a research report and code repository. Agent Laboratory is meant to assist you as the human researcher in implementing your research ideas. You are the pilot. Agent Laboratory provides a structured framework that adapts to your computational resources, whether you’re running it on a MacBook or on a GPU cluster. Agent Laboratory consists of specialised agents driven by large language models to support you through the entire research workflow—from conducting literature reviews and formulating plans to executing experiments and writing comprehensive reports. This system is not designed to replace your creativity but to complement it, enabling you to focus on ideation and critical thinking while automating repetitive and time-intensive tasks like coding and documentation. By accommodating various levels of computational resources and human involvement, Agent Laboratory aims to accelerate scientific discovery and optimise your research productivity.

• CrewAI hosts an AI agent platform and an open-source release:

  • crewAIInc/crewAI: Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI helps agents work together to tackle complex tasks.

  I’ve seen Michael Kuiper do some very impressive things with it.

• Swarm AI

  • kyegomez/swarms — enterprise-grade, production-ready multi-agent orchestration framework (website: https://swarms.ai)
  • Overview - distributeds
  • Swarms AI - Enterprise Multi-Agent Framework

• J-Rosser-UK/AgentBreeder: Mitigating the AI Safety Impact of Multi-Agent Scaffolds (Rosser and Foerster 2025)

  Scaffolding Large Language Models (LLMs) into multi-agent scaffolds often improves performance on complex tasks, but the safety impact of such scaffolds has not been as thoroughly explored. In this paper, we introduce AGENTBREEDER a framework for multi-objective evolutionary search over scaffolds. Our REDAGENTBREEDER evolves scaffolds towards jailbreaking the base LLM while achieving high task success, while BLUEAGENTBREEDER instead aims to combine safety with task reward. We evaluate the scaffolds discovered by the different instances of AGENTBREEDER and popular baselines using widely recognized reasoning, mathematics, and safety benchmarks. Our work highlights and mitigates the safety risks due to multi-agent scaffolding.

• Why Simulator AIs want to be Active Inference AIs

• Announcing the Agent2Agent Protocol (A2A) - Google Developers Blog

• Kimi K2: Open Agentic Intelligence is an open-source AI model for agencey

  Kimi K2 is our latest Mixture-of-Experts model with 32 billion activated parameters and 1 trillion total parameters. It achieves state-of-the-art performance in frontier knowledge, math, and coding among non-thinking models. But it goes further — meticulously optimized for agentic tasks, Kimi K2 does not just answer; it acts.

3 References