KersonAI retweeted
When people compare Claude to Gemini in their coding skills, they are comparing apples and oranges.
Claude isn't an LLM. It's a coding application: it has multiple "agents" responsible for calling various tools, such as searching files containing specific keywords and extracting snippets around the matches, or creating a test script and running it.
Gemini is a pure LLM with reasoning steps. Pure reasoning, while it significantly improves an LLM's skill compared to an LLM answering right away, is not as effective for solving complex coding problems because complex coding requires not just thinking step by step, but also obtaining the actual step outcome when the step involves applying a function to some input and observing the output.
The thinking in LLMs tries to simulate the execution of the code, but a simulation is never perfect. It鈥檚 like an LLM can pretend to be a Linux terminal, but it cannot *be* a Linux terminal.
I hope Google is working on an agentic system for coding, and hopefully, it will be part of the Gemini UI. But right now, if you use Gemini, you use one LLM in a vanilla way, while when you use Claude, you use a multi-agent system capable of using coding-specific tools.
KersonAI retweeted
OpenAI quietly dropped a 34-page technical manual on building AI agents that 99% of people will never read.
I spent 3 days coding every single pattern they revealed.
Here's the practical guide to autonomous AI agents:馃У
KersonAI retweeted
ChatGPT-o3 is insanely powerful.
But 95% of users have no clue how to unlock its full potential.
Here are 10 ways to use it like a pro:
KersonAI retweeted
The musical legend @Jcrudess testing out the new Styles feature 馃憖
Have you tried it yet?
KersonAI retweeted
Thanks for sharing! Love seeing these use cases!








