Interesting review, thx. In my opinion, The issue with AI agents isn’t about frameworks or syntax, it’s deeper. 1/ The definition of "agents" is unclear, with no standard. Some even call a simple LLM call an agent, which isn’t accurate, 2/ Language models weren’t designed for automation; we use them as components, which creates challenges, 3/ Most of applications aren’t ready for automation via agents, unless they start and create an standard interface, like how businesses adapted the RESTFull api. For example, Stripe’s new tools for agents are a step forward, but broader readiness is lacking. The real problem is redefining agents, adapting models for automation, and preparing applications, not frameworks. Most new libraries these days are just tools or facades to attract funding during the AI hype. It’s a smart move: they gain support, build something quickly for the public, and then work on tackling the real issues. There’s a competition to solve the core problem, so the funding goes to addressing that, not just the platform itself. One day, we’ll see a "ChatGPT moment" for this "Agent" concept.
Even though I haven't understood the critics, I would be happy if you dive more into this. I think that getting to know more about agents programming and frameworks really gives you the tool to tailor for specific needs that can be heavily creative. This would be very valuable!
That dynamic calling of agents is what langgraph addresses, it provides more control on the logic flow. I could see pydantic agents acting as nodes in langgrapgh!!! Combining 2 powerful agent frameworks !!!
Great video. I fail to see specific use cases where to use pydantic AI as other Agentic frameworks offer data validation which is already built on Pydantic. I do like the idea of using more vanilla python for the flow.
It's interesting, but really, waiting for more intelligent models will solve most of what you were showing. We won't need to build scripts for AI to traverse.
Interesting review, thx. In my opinion, The issue with AI agents isn’t about frameworks or syntax, it’s deeper. 1/ The definition of "agents" is unclear, with no standard. Some even call a simple LLM call an agent, which isn’t accurate, 2/ Language models weren’t designed for automation; we use them as components, which creates challenges, 3/ Most of applications aren’t ready for automation via agents, unless they start and create an standard interface, like how businesses adapted the RESTFull api. For example, Stripe’s new tools for agents are a step forward, but broader readiness is lacking.
The real problem is redefining agents, adapting models for automation, and preparing applications, not frameworks. Most new libraries these days are just tools or facades to attract funding during the AI hype. It’s a smart move: they gain support, build something quickly for the public, and then work on tackling the real issues. There’s a competition to solve the core problem, so the funding goes to addressing that, not just the platform itself. One day, we’ll see a "ChatGPT moment" for this "Agent" concept.
Even though I haven't understood the critics, I would be happy if you dive more into this. I think that getting to know more about agents programming and frameworks really gives you the tool to tailor for specific needs that can be heavily creative. This would be very valuable!
Nice, I just happened to need this today. Thanks
one thing though, this is quite at an early stage!
That dynamic calling of agents is what langgraph addresses, it provides more control on the logic flow.
I could see pydantic agents acting as nodes in langgrapgh!!! Combining 2 powerful agent frameworks !!!
I think there could be a lot of use cases for this but I have to explore it more to see it's true capabilities.
Great overview! 👏
The power of AI: thinking fast, acting faster ✨
Great video. I fail to see specific use cases where to use pydantic AI as other Agentic frameworks offer data validation which is already built on Pydantic.
I do like the idea of using more vanilla python for the flow.
Thank you for Sharing bro
which one to use and when lang graph vs llama agents vs magnetic-one vs phi-data-agent?
is it compatible with langgraph?
On the first glance, I preferred BrainBlend-AI/atomic-agents.
It's interesting, but really, waiting for more intelligent models will solve most of what you were showing.
We won't need to build scripts for AI to traverse.
Tbh, I expected something better.
From me or from Pydantic ?