A retrieval-augmented generation (RAG) system is built to answer strictly from the documents it retrieves. But when engineers optimize these AI pipelines end-to-end, the reader module can learn a shortcut: instead of relying on retrieved evidence, it starts answering from its own internal memory — while the system’s overall accuracy keeps climbing. This is the hidden challenge of “role drift,” a failure mode in compound AI systems where individual modules learn to bypass their assigned tasks even as end-to-end performance improves.
To address this, researchers at MIT and Harvard introduce Role Anchor, a technique that forces modules to stay in their lanes during training. When applied, the technique mitigates role drift. For example, it forces the RAG reader to rely on retrieved evidence instead of answering based on its internal knowledge.
The primary takeaway for practitioners is that end-to-end accuracy alone can overstate how much a compound AI system has genuinely learned. — Read More
Daily Archives: August 18, 2026
Hollywood strikes first AI copyright truce with ByteDance
The Motion Picture Assn. has struck its first agreement with an artificial intelligence company, saying Monday that ByteDance has strengthened copyright guardrails into its AI video and image tools that drew legal threats from every major Hollywood studio in February.
… The MPA, the trade association that lobbies for major studios such as Disney, Netflix and Sony Pictures Entertainment, said it has had constructive engagement with ByteDance over the last several months to rein in Seedance, a text-to-video generation tool, and Seedream, which produces images using text commands. — Read More
Roadmap decisions rather than dates.
In 1975, Fred Brooks’s The Mythical Man-Month warned us that adding individuals to a late project makes that project even later rather than speeding it up. However well-understood this is, it remains extremely common for leaders to repeat this mistake, even today. My personal theory is that most leaders performing the “ask for more staffing” gambit intuitively know they are optimizing to “get stuck” in a way that shifts responsibility off them because they didn’t get the requested funding, rather than that there is no way to accomplish the task at hand.
Even if teams avoid the simplest versions of the mythical man-month’s trap, organizations that operate through strict ownership of defined boundaries introduce the same problem. Well, of course your team can’t launch that without approval from the architecture team and the security team and performing a user research study. … Even with strong AI-enhanced workflows, I still find human review of load-bearing technical decisions (e.g. a decision that will meaningfully impact subsequent decisions) to be extremely valuable. — Read More
Software Engineering fundamentals matter more than ever
The manifestation of my imposter syndrome, for me and today, is what does it mean to be a software engineer. There’s a lot more noise than signal on the Internet about agentic engineering, what can be accomplished, and its implications for the future. The title I chose rather gives it away; it’s about choosing — carefully — all the things you need to choose when you’re solving the puzzles of software and systems development.
… In the past year, agent harnesses crossed the “can it be done” rubicon. … The economic models for the large models aren’t viable from any report that I’ve seen, but the capability isn’t going away. Instead it’s shrinking (fast!). Open weight models are making (beefy) personal computers quite capable of doing the same. They’re not quite as effective, but the delta in time and capability isn’t large. — Read More
Z.ai Delays GLM-5.3 Weights Two Weeks After Cyber Score Beats Mythos 5
GLM-5.3 scored 84.5% on CyberGym, edging Anthropic’s restricted Mythos 5, and Z.ai responded by holding its downloadable weights until around August 28. The lead vanishes on exploitation benchmarks, and every figure came from Z.ai’s own harness.
Beijing-based Z.ai released GLM-5.3 on Friday while holding back the model’s downloadable weights and gating its most sensitive cybersecurity functions. In launch tests, GLM-5.3 scored 84.5% on CyberGym and edged Anthropic’s restricted Mythos 5. Z.ai, which published downloadable weights for its previous GLM models, is gating this one with the kind of control American labs have used on their strongest cyber models. — Read More
Rick's Cafe AI 