Breaking down new LLM releases: the architecture choices, the benchmarks, and what the numbers mean.
CS enrollment is falling as AI writes the code. Why a computer science degree is still worth four years, if you train to judge software rather than type it.
Read full dispatch
Coding has never been the hard part. AI made writing code cheap, so why haven't delivery dates moved? What actually slows software shipping, with evidence.
Read full dispatch
Harnesses tend to agree on features but differ in behavior. That's why having fluency in a single open-source AI coding harness can be more valuable than constantly chasing the latest and greatest. It's also important to consider the costs associated with locking into one option.
Read full dispatch
Context, harness, loop, and graph engineering are fresh labels for traditional system design elements like control loops, state machines, and test harnesses. Take a moment to explore the changes and see what's new.
Read full dispatch
Here's a straightforward overview of 9 advanced RAG retrieval techniques, including hybrid search, HyDE, query routing, and others, each designed to solve specific challenges. It's a friendly guide to help you understand how each method works and what problems they address.
Read full dispatch
Amazon's new AI code review process means that humans are checking the AI's output, rather than the other way around. This comes amidst outages, public denials, and the policies that followed, highlighting the ongoing efforts to ensure safety and accountability.
Read full dispatch