@petergyang: "The really big models can do really dumb things for minuscule tasks." From Micky (@Rasmic): "There's no need to use st…
Summary
Peter Yang 和 Micky 讨论大模型在小任务上的浪费问题，主张让较小的模型处理具体任务，而用最强模型做编排与审查，并分享了四步 agentic 工作流。
View Cached Full Text
Cached at: 10/12/26, 07:27 AM
“The really big models can do really dumb things for minuscule tasks.”
From Micky (@Rasmic):
“There’s no need to use state of the art models for everything. I had some tax work to do with a lot of numbers and expenses to crunch. Luna just blew through that, and I used Sol as a subagent to track Luna’s work.
State of the art should probably stick to orchestrating and reviewing, because a lot of the smaller models are getting really, really good.“
📌 Watch Micky break down which models he uses here: https://youtu.be/8rEAkEvhhcQ
@petergyang: “The really big models can do really dumb things for minuscule tasks.” From Micky (@Rasmic): “There’s no need to use st…
Channel: @petergyang Source: https://www.youtube.com/watch?v=8rEAkEvhhcQ&feature=youtu.be
Peter Yang (@petergyang): “One wrong assumption in [your software factory] just destroys everything.”
Here’s my new episode with @Rasmic, where he walked me through his 4-step agentic engineering workflow to ship production apps:
- Isolate: Give each agent its own worktree
- Build: Create reusable
Similar Articles
@mylifcc: This is not an ordinary large model, but a Multi-Agent Orchestration System—a small model itself that intelligently and dynamically coordinates multiple cutting-edge models such as GPT, Claude, and Gemini, autonomously assigning roles, decomposing tasks, and completing comp...
Sakana AI has released a Multi-Agent Orchestration System that uses a small model to intelligently coordinate cutting-edge large models like GPT, Claude, and Gemini to autonomously assign tasks and handle complex workloads.
When working on agent workflows or autonomous agent loops, a major bottleneck affecting experience is: every small branch decision, tool routing, task completion check, or safety interception, if all handled by expensive reasoning LLMs, not only brings seconds of delay per call, but token costs also...
This article introduces the open-source list awesome-jev-projects, which decouples System 1 fast decision models from System 2 reasoning LLMs to optimize AI agent workflows, reduce latency and costs, and provide code-level references.
@yibie: Recommended article: Eugene Yan from Anthropic (former Amazon/Alibaba ML team lead) writes a practical guide to his personal AI workflow. Not abstract ideas, but concrete methods you can replicate tomorrow: how to organize directories for easier model retrieval, how to write C…
This is a tweet recommending Eugene Yan's AI workflow guide, detailing how to efficiently collaborate with AI and achieve compound work gains by organizing context, configuring CLAUDE.md, creating skills, and more.
@Aoyi21: https://x.com/Aoyi21/status/2108474440904495566
A long-form Chinese opinion piece drawing on Andrej Karpathy's tips for interacting with large language models (such as using the ASD-STE100 simplified language standard) to explain how AI can be used to clarify concepts, visualize them, and turn them into interactive mini-tools — enabling truly personalized learning tailored to each individual.
@MaximeRivest: this is such a wonderful example of the jagged intelligence / capability frontier. pointing at things in and image seem…
讨论指出，图像中的指向（pointing）看似是基础能力，但若模型未在训练中学习，往往做不好；作者质疑大模型的「涌现能力」说法，并认为更大模型未必总是更好，便宜/本地/专用模型可能有相当表现。