Examining what Large Language Models actually do through the lens of stochastic generation rather than reasoning. Using Apple’s research and Google’s Self-Consistency work as bookends to define the boundaries of ’limited capacity stochastic constructors’ and their implications for AI development.
Exploring how Test-Driven Development can serve as a reward function in AI-assisted software development, bridging the gap between documented intentions and measurable outcomes while maintaining human oversight.
A personal account of returning to frontend development after a decade, exploring the complexities of modern JavaScript frameworks like Next.js, and discovering the simplicity of HTMX. This journey highlights the importance of critical thinking in tech choices and the potential pitfalls of blindly following industry trends.
Exploring the complementary approaches of Intention and Chat Based Development in AI-assisted software engineering. This post delves into the evolving landscape of development methodologies, the role of human oversight, and the future of AI in software creation.
A personal account of navigating bureaucratic challenges, highlighting the irreplaceable value of human judgment and empathy in an increasingly automated world. This story underscores the importance of keeping humans in the loop in AI-driven decision-making processes.
Discover the challenges and risks of agent-based AI for businesses. Learn to anticipate and manage the technological, operational, strategic, and ethical challenges related to this revolution.
A personal reflection on the various roles within AI Engineering, from AI-assisted software engineers to specialized AI system engineers, written en route to the AI Engineer World Fair.
Discover agent-based AI, its applications, and revolutionary potential. From intelligent agents to ethical challenges, dive into the future of autonomous AI.
This article bridges the gap between the current Data Science-led Generative AI (GenAI) movement and “good old-fashioned” systems. Tool Augmented Generation is used to desribe using the LLM to choose which tools it should rfetrieve data from. Given the automony planning and oversight are key components to consider.