🤖 Pioneer of the new Agents-as-a-Service (AaaS) model: agents.vrsen.ai/ 🐝 Building agencii.ai
Joined April 2021
- Tweets 47
- Following 77
- Followers 1,256
- Likes 49
Watch the full video and find out how to prepare: piped.video/emiZkLvBzhQ?si=Qg_o…
7) And this has HUGE implications for the world of AI agents:
- It’s EXPONENTIAL from here.
- Agents will finally become RELIABLE.
- Agents will tell YOU what to do.
- Workflow automation platforms are DONE
- Agents will EVOLVE overtime.
- AI Agents will be able to run for DAYS.
6) This means that these AI SYSTEMS are finally **bigger than the sum of their parts**.
Which is why they can call up to **600 tool calls in a sequence**... 🤯
5) But with o3 and o4-mini it’s finally there!
As quoted from OpenAI’s blog post, “ were specifically trained to use tools through reinforcement learning, teaching them not just how to use tools, but to reason about when to use them.”
5) All previous AI models were missing this crucial feedback loop.
Previously, models had never been trained to reason over tool calls; they were trained to reason over messages, but not over tool calls. Even Gemini-2.5 and o3-mini struggled with this. They would call the tools, see the outputs, but would still fail to correct their course accordingly.
This has always caused trouble in our AI agency, because our agents would often stop short and fail to complete their tasks - even when they were on the right track.
4) Feedback loops are crucial in systems.
Through feedback loops systems can either self-correct or amplify its behavior over time.
3) For example,
A thermostat in your home is a system that has a few elements like the AC, temperature sensor, and the windows outside. Whenever the temperate in the room goes up (like it’s 100+ degrees outside in Dubai😓), the AC kicks in to bring the temperature down to a comfortable level.
This is an example of a **feedback loop** in action
2) Each system has 3 core components:
- Elements
- Interconnections
- Purpose
Each system is a **set of elements connected together in a way that achieves something**.
(Understanding this will give you a powerful new lens through which you can better understand the world.)
1) First, let’s understand what a system is.
“A system is a set of interconnected things that together produce a specific behavior over time” - Donella Meadows.
A key thing to note is that **a system is bigger than the sum of its parts**.
(Remember this for later)
Everyone MISSED this:
"o3 and o4-mini are no longer even AI models; they’re AI SYSTEMS."
This is HUGE for AI agents.
(a thread) 🧵
100%.
The real game changer, however, is the fact that it's being adopted as a **standard** in the industry.
Just as OpenAPI allowed developers to use any third-party software, MCPs allow Agents to use external tools.
Standard means collaboration. 🤝🤝
In a world where you can build anything from a single prompt, it's the people who take initiative that will make the largest impact.
This Revolutionary AI Business Model Will Make Millionaires in 2025
piped.video/ED4SUWgoAhw
Not impressed by Deepseek. Not even worth a video.
Their reasoning model doesn't support function calling at all, and even the chat model still hallucinates on every function call.
This model is useless for building AI agents.
Big updates coming to Agency Swarm “soon”
Operator is based on a new model we’re calling “computer-using agent” (CUA).
CUA combines GPT-4o's vision capabilities with advanced reasoning through reinforcement learning. It’s trained to control a computer in the same way a human would—it looks at the screen, and uses a mouse and keyboard.
The model still has limitations and will continue to evolve based on feedback. We plan to bring CUA to the API for developers soon. openai.com/index/computer-us…