Stories by gronky_

Don't trust AI agents

Early science acceleration experiments with GPT-5 [pdf]

GPT-5 has a hidden system prompt

OpenAI Testing Agent Demo

The First LLM

OpenAI adds MCP support to Agents SDK

Security on the Path to AGI

OpenAI Is Adopting MCP

A new quest to save the "doomsday glacier"

Calling a Wrap on My Weeknotes

Building Agentic Flows with LangGraph and Model Context Protocol

Reasoning Best Practices

What will AI do to (p)research?

Mitochondrial swap from cancer to immune cells thwarts anti-tumour defences

The biggest AI flops of 2024

Claude 3.5 helped me fight off a $10k rental car damage claim – and won

LLMs as Method Actors: A Model for Prompt Engineering and Architecture

Rediscover the Joy of Early Web Development with XSLT

1