Episodes (Page 2)
✨
Coding agents reinvent common game elements like cameras from scratch.
✨
Most recorded video is never posted due to editing friction.
Ekaterina Deyneka
✨
Large models prioritize consistency, leading to generic outputs.
✨
Code reviews are declining, with AI taking over the process.
✨
Deno uses AI agents with broad production access for incident response.
Ryan Dahl
✨
Sending more tokens can be cheaper and more effective than compaction.
Louis-Francois Bouchard
Omar Solano
Samridhi Vaid
✨
Web infrastructure must handle real-world demands, not just tests.
Patricija Zemaityte
✨
Context-as-a-Service (CaaS) can be more expensive than building your own.
Omer Primor
✨
AI agents struggle with real-world complexities beyond simple games.
Gaurav Mishra
The Dark Arts of Web Automation: Teaching Agents to Use Websites Like Humans — Corey Gallon, Rexmore
✨
Agents can bypass human-like security measures like CAPTCHAs.
Corey Gallon
✨
Browser agents are limited by engineering, not AI models.
Paul Klein IV
✨
Small, screen-agnostic scripts can match frontier model performance.
Pierluca D'Oro
✨
Web agents will drive action, but not necessarily via APIs.
✨
Start LLM knowledge bases with voice dictation for raw material.
✨
Training frontier models at scale is a rare, apprenticeship-based skill.
✨
Intelligence is reasoning on new problems; expertise is applying knowledge.
✨
Models trained on public data collapse when generating private data.
✨
Agent improvement is a data mining problem using traces.
Vivek Trivedy
✨
LLM memory systems can create inaccurate profiles from conversations.
Shlok Khemani
✨
Current agent evaluation methods are outdated for complex tasks.
Ben Hylak