Tag
The author read an article about Dify's decline and reflected on how Dify, once a hot topic two years ago, is no longer discussed, feeling that it seems to have fallen behind.
该论文提出了一种两阶段的 competence-aware 框架，用于 LLM 多智能体系统的自动编排，根据模型在当前查询上的能力动态调整角色激活与通信图，在六个基准上平均达到 93.81%，同时在高能力骨干上降低 36.26% 的 token 成本。
LumaBrowser, an all-in-one local AI assistant bundling LLM chat, image/music/voice models, agent scheduling, clustering, and IDE integration, is now open source and hosted on GitHub after a year in development.
一位有五年实时AI助手构建经验的工程师总结了生产级Agent系统的四条关键经验：可视化流程图、预设人工审批环节、集中管理模型选择、以及预先设计模型出错时的处理机制，并附带了一个公开的参考实现项目。
Anthropic为Claude Code推出工业级多智能体协作架构：Haiku 5.5作为并行高速子代理处理机械任务，Sonnet 5.5作为主力生成代码，Opus 5.5仅在关键决策点以--advisor身份介入，通过'三权分立'式分层显著降低旗舰模型的上下文与token消耗。
Aster by AsterWise launches an intelligent model routing service that dynamically routes tasks across models, with aster-code for coding agents and aster-work for custom agents and workflows, offered on a paid, prepaid basis.
Earendil and the Pi community shipped Pi 1.0 alongside an experimental new package, Pi Durable, a minimalistic framework for building long-running, malleable agents that can run anywhere with a JavaScript runtime and survive failures via pluggable storage backends.
MoFlow formulates agentic workflow generation as a multi-objective MDP and solves it with Convex-Hull Monte Carlo Tree Search with set-valued backups, covering the Pareto front so workflows can be retrieved for any preference without retraining. It outperforms six baselines on math, code, and QA benchmarks in terms of hypervolume.
DataMagic introduces a declarative multi-agent orchestration framework (DVSpec) that turns raw tabular data into narrated data videos, improving quality from 2.13/5 to 3.89/5 over GPT-5 baselines and cutting task time by 79.7% in user studies.
The author describes building 'Lunar Citadel,' an experimental AI civilization with persistent state and social continuity, and has created a small Discord community for AI context architects to collaborate and share ideas.
Opti-Q is a database-inspired optimizer for multi-LLM question answering that plans execution DAGs to optimize answer quality under cost, latency, and energy constraints, achieving significant improvements on benchmarks.
SPL (Structured Prompt Language) is a declarative language that unifies deterministic and probabilistic computation modes for LLM workflows, enabling specification of computations that run across local and cloud nodes. It includes an extensive cookbook and experimental validation.
PRECEDE is a precedent-guided co-scientist for side-effect-aware drug redesign that frames drug modification as evidence-grounded reasoning over drug–side-effect associations and biomedical knowledge graphs, coordinated by an LLM orchestrator with human oversight.
The author argues that chat-first AI tools are insufficient for building autonomous agent workflows and describes alternative primitives like scheduled triggers and sub-agent delegation, advocating for a paradigm shift from 'chat with tools' to 'autonomous processes that use LLMs'.
The author shares a practical 4-tier LLM routing stack for agent work, where a fast orchestrator handles most requests and only escalates to expensive models when deep reasoning is required, significantly improving cost and interactivity.
A technical teardown of how multi-agent frameworks like CrewAI and AutoGen actually route information under the hood, revealing that they are essentially automated prompt-chaining loops. The article explains why agents get stuck in infinite loops due to context window inflation and missing deterministic stop conditions, offering practical advice for builders to treat agents as functional programming functions instead of human collaborators.
SciOrch presents an 8B vision-language model trained with MCTS to coordinate multiple expert LLMs for multimodal scientific reasoning, achieving superior performance while reducing API costs.
InfraMind introduces an infrastructure-aware multi-agent LLM orchestration framework that uses reinforcement learning to dynamically select models and topologies based on real-time system load, achieving up to 7x lower latency and 99.9% SLO compliance under high load.
BotCircuits Agent is an open-source framework that introduces a Workflow-native Agent Loop architecture, splitting deterministic state-machine navigation from targeted LLM execution to reduce deviations and token costs.
Anthropic quietly shipped /workflows in Claude Code, replacing LLM orchestrators with code-based control flow to avoid token tax and context sloppiness.