`claude-prompt-engineer` is a constrained agent — meaning its behavior is highly-controlled, leading to better results than open-ended agents. It chains together lots of Claude 3 calls that work together to find the best possible prompt.
AGENTS
-
claude-prompt-engineer: AI Agent Creates Optimal Claude Prompts
By
–
Introducing `claude-prompt-engineer` ✍️
— Matt Shumer (@mattshumer_) 20 mars 2024
An agent that creates optimal Claude 3 prompts.
Just describe a task, and a chain of AIs will:
– Generate many possible prompts
– Test them in a ranked tournament
– Return the best one
And it's open-source: https://t.co/nrivU2BWmn pic.twitter.com/ruxDCrq1DLIntroducing `claude-prompt-engineer` An agent that creates optimal Claude 3 prompts. Just describe a task, and a chain of AIs will:
– Generate many possible prompts
– Test them in a ranked tournament
– Return the best one And it's open-source: https://
github.com/mshumer/gpt-pr
ompt-engineer
… -
Learning LLM Skills with AI Tutor Cosmo
By
–
2. Engineering Output Size with LLMs
— Rowan Cheung (@rowancheung) 20 mars 2024
Being able to run prompts and get feedback directly from Cosmo (the AI tutor Corgi) makes learning really fun.
You have an LLM helping you learn how to work better with LLMs! 🤯 pic.twitter.com/KYUOnlklS02. Engineering Output Size with LLMs Being able to run prompts and get feedback directly from Cosmo (the AI tutor Corgi) makes learning really fun. You have an LLM helping you learn how to work better with LLMs!
-
Palantir Shares Digital Agency Team Health Strategies at Midwest Camp
By
–
Ditch the "traditional agency model" for greater team health and work quality. Palantir invites you to join us for "Unburdening Teams, Unleashing Potential: New Strategies for Digital Agency Success" at @midwestcamp 2024. #midcamp2024
Wednesday 10 am, Room 314A -

GPT-5 Summer Release Could Reshape AI Agent Development
By
–
A well-reported piece with multiple sources suggests GPT-5 this summer (if it is called GPT-5). If you are planning a business around older models, take note. We will learn a lot about the future of AI from how much it beats GPT-4. Agents are an obvious use if it is much better.
-
GPTs as Mini-Agents: Sharing, Education, and Monetization Strategy
By
–
GPTs are good ways of sharing mini-agents, and I think they solve a lot of problems for teams and organizations. When they can be shared to non ChatGPT Plus users they will have a role in education as well. But OpenAI suggested a way to monetize them, and there has been no news.
-
Personal Superintelligent AI Enabling True Democracy
By
–
In one of my favorite near term AI timelines, everyone is in control of a personal superintelligent AI interface to the world. Reality will become transparent to all, and we coordinate based on the undeniable outcomes of our actions. This will be the first true democracy.
-
Enhancing Haiku Subagents with Advanced Tools and Capabilities
By
–
Of course, this is just the start. You can imagine each of the Haiku subagents could be connected with tools/functions calling/image recognition. And improve this framework even more.
-
Maestro: Claude Opus Framework Orchestrating Haiku Subagents
By
–
Introducing Maestro ✨
— Pietro Schirano (@skirano) 19 mars 2024
A framework for Claude Opus to orchestrate subagents.
Simply ask for a goal, and Opus will break it down and intelligently orchestrate instances of Haiku to execute subtasks, which Opus will review at the end. 🧙♂️
The output is saved as markdown 👇 pic.twitter.com/yCxkw8ECwxIntroducing Maestro A framework for Claude Opus to orchestrate subagents. Simply ask for a goal, and Opus will break it down and intelligently orchestrate instances of Haiku to execute subtasks, which Opus will review at the end. The output is saved as markdown
-
GPT-4 Limitations in AI Agents and Future Upgrades
By
–
The brains here are GPT-4, I believe, and Devin has GPT-4 style limitations on what it can accomplish. I assume that the brains will be upgraded when GPT-5 class models come out, and there will be many other agents on the market soon. A thing to watch for.