5thDevBlogCurated Outside the Norm, Our Updates and Observations
ResearchThe Agreeable Agent
Models are trained to agree with us, and new research shows why that's riskier in coding agents than in chatbots and why observing agents beats trusting them
NewsWhen the Model Moves Into the Chip
AMD's deal for World Labs is a bet that the next wave of agents will need a model of the physical world, and here's what that means for tech leaders
EducationWhat We Choose to Keep
Sean Goedecke's essay shows AI breaking a genuine signal, the link between solving a hard puzzle and having produced real insight, and reaches for a narrow, fie
NewsBiological Computing and the Race to AGI
A growing, funded slice of the AI industry is betting that the path to more capable and more efficient intelligence isn't a bigger GPU cluster, it's literal bio
ResearchThe Other Half of Trustworthy AI
A new paper shows you can train language models to know what they don't know, closing much of the gap between expressed confidence and real uncertainty
ResearchThe jump and the record
DeepMind researcher Tom Zahavy argues LLMs structurally lack abduction, the inventive jump behind discoveries like general relativity
NewsWhat it actually takes to own your intelligence
Garry Tan's Startup School talk named the era 5thDev is built for: personal AGI, agents on infrastructure you own that compound your knowledge
Development5thDev: Reliability for Agents
Reliability, not intelligence, is the bottleneck for agentic AI