AI agents have real potential. I went back to Devin, the AI coder agent, after not using it for a day & completely forgot that I assigned it to create a visualization of the economics of startup dilution. It plugged away and I came back to a solid draft that could be iterated on
AGENTS
-
Centaurs and Cyborgs: Human-AI Collaboration Framework
By
–
Write up of our paper: https://
oneusefulthing.org/p/centaurs-and
-cyborgs-on-the-jagged
… -
Claude 3 GPT-4 Gemini guardrails comparison analysis
By
–
What a weird set of guardrails from Claude 3. GPT-4 is the only one willing to run with it. Gemini tries to be helpful, suggesting I have come across a crime scene.
-
Building Code Agents: Programming AI Software Engineers
By
–
Webinar Wednesday 3/20: Building Code Agents! Sign up here https://
crowdcast.io/c/codeagents Agents that write and run code are powerful, as Cognition Labs showed with their recent release of Devin, the "AI SWE". But they are complex to program, hard to deploy, and even harder to -
Analyzing the architectural approach of AI agents like Devin
By
–
Wondering if “Devin” approach will make it work better smith like a combination of LLMs with an access to search which can keep enhancing its own knowledge base more precisely to the context
-
Smaller Models Handling Smaller Tasks in AI Systems
By
–
Yeah that's why the smaller model is doing the smaller tasks
-
Claude Opus Subagents Build Drawing App Autonomously
By
–
Claude Opus's ability to orchestrate subagents is absolutely insane and deserves more attention.
— Pietro Schirano (@skirano) 18 mars 2024
Watch Claude direct subagents to build an entire drawing app. 🧙♂️
If you like, I could share this code, which can basically solve any goals you present, step by step.
Like Devin! pic.twitter.com/sFs8i22B9GClaude Opus's ability to orchestrate subagents is absolutely insane and deserves more attention. Watch Claude direct subagents to build an entire drawing app. If you like, I could share this code, which can basically solve any goals you present, step by step. Like Devin!
-
Claude Opus Deploys Haiku Instances for Multi-Task Problem Solving
By
–
Claude Opus breaks down the task into subtasks, then deploys instances of Haiku to solve each task. Once these tasks are solved, the output goes back to Opus for review. Pretty incredible. When I tried to build things like this with GPT4, it was very hard for the model to know
-
Frontier Models, RAG, and AI Agents in Enterprise
By
–
In addition to reviewing the positives & negatives of the three frontier models, this post has some of my more provocative takes for people using AI inside organizations, including why RAG may not be the future, and why agents like Devin the Coder might be