WeAreDevelopers LIVE - SpeculAItions

What if the best way to work with powerful AI is to let it be stupid? Discover a new method for managing agentic AI without micromanagement.

#1about 3 minutes

Using AI agents for large pull request reviews

AI agents can help break down and analyze large pull requests, but the fundamental need for security controls in software engineering remains critical.

#2about 5 minutes

Identifying the formulaic patterns of AI-generated text

AI-generated content often follows predictable formulas, from talk proposals to social media posts, with different models like Claude and ChatGPT exhibiting distinct personalities.

#3about 4 minutes

Questioning the narrative of 10x developer productivity

The push to run multiple AI agents simultaneously creates pressure for unrealistic productivity gains, raising questions about its practical application and the risk of burnout.

#4about 5 minutes

Meta's acqui-hire of Moldbook and the problem of AI slop

Meta's acquisition of the Moldbook creators highlights a trend towards agent-only networks, while the decline of platforms like BuzzFeed signals a broader issue with low-quality, AI-generated content.

#5about 5 minutes

AI offers modest productivity gains, not 10x improvements

Real-world studies suggest AI provides around a 10% productivity boost, not 10x, and can lead to cognitive overload or "brain fry" by shifting work from creation to review.

#6about 7 minutes

Exploring recent AI incidents and creative developer hacks

The AI Incident Database tracks real-world failures, like Chipotle's unsecured chatbot, while developers create clever hacks such as using YouTube as a free file CDN.

#7about 4 minutes

A modern look at HTML tables and JavaScript dates

A review of fundamental web technologies covers the power of semantic HTML tables, the new JavaScript Temporal API for handling dates, and a CSS trick for stable number rendering.

#8about 6 minutes

Playing a game of "Fake or News" with tech headlines

A fun segment challenges the hosts to distinguish between real and fabricated tech news headlines, covering topics from crypto scams to brain-controlled browsers.

#9about 7 minutes

Using the Ralph Wiggum method for agentic coding

The Ralph method involves letting an AI agent iterate naively in a simple loop to solve a problem, offering an alternative to detailed micromanagement and prompt engineering.

#10about 6 minutes

How agentic AI forces a rethink of software engineering

The rise of agentic AI prompts a re-evaluation of the entire software development lifecycle, from team structures to the potential for AI-native programming languages.

#11about 5 minutes

The growing challenge of human-like AI interaction

As AI becomes more human-like for applications such as mental support, it becomes increasingly difficult to detect AI-generated content, leading to the proliferation of fake online personas.

#12about 1 minute

Final advice for critically engaging with AI content

The key to navigating the modern internet is to critically question all content and to always add your own unique voice and perspective to anything generated by AI.