This week we feature articles from Stephan Rabanser, Sayash Kapoor, & Arvind Narayanan, and Alberto Romero.
Additional reading comes from Ethan Mollick, JA Westenberg, Daniel Kocot, and Steve Hedden.
News comes from Graphwise, Dataiku, Krisp, and Cockroach Labs.
Our next issue will be published on 3/11/26
All previous issues are available at https://gilbane.com/gilbane-advisor-index
Opinion / Analysis
Towards a science of AI agent reliability
Quantifying the capability-reliability gap
“Surprisingly, even though the lack of reliability of AI agents is well known, right now the AI industry doesn’t have good tools for measuring reliability, or even a good definition of reliability.”
Stephan Rabanser, Sayash Kapoor, and Arvind Narayanan decided to take a stab at a “comprehensive measurement of reliability”. Together with some additional researchers have they’ve made a sensible and very useful start and published a draft, which they summarize in this post. The complete draft can also be downloaded. (9 min)
https://www.normaltech.ai/p/new-paper-towards-a-science-of-ai
A new Wharton study on AI warns of a growing problem: cognitive surrender
Casual users should pay special attention
Alberto Romero: “The study introduces the concept of “cognitive surrender,” our tendency to adopt AI outputs with “minimal scrutiny,” overriding “both intuition and deliberation.” I’ve read it, and the findings, although unsurprising, are still quite scary.”
You can download the full paper, Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender, with the included link. (7 min)
https://www.thealgorithmicbridge.com/p/a-new-wharton-study-on-ai-warns-of
More Reading
- A guide to which AI to use in the Agentic Era — It’s not just chatbots anymore. via One Useful Thing
- Everyone in AI is building the wrong thing for the same reason via Westenberg
- What other industries can learn from healthcare’s knowledge graphsvia towards data science
- Why AI transformation is a software engineering discipline problem – Part 1 via Architectural Bytes
Content Technology News
Graphwise announced the immediate availability of GraphRAG
Beyond vector-only RAG, knowledge graphs provide context and common sense to AI. The AI-workflow engine turns “Python prototypes” into production-grade systems.
https://graphwise.ai/news/new-graphrag-solution-moves-beyond-vector-only-rag-knowledge-graphs-provide-context-and-common-sense-to-ai/
Dataiku launches 575 Lab, its open source initiative for responsible AI
The 575 Lab will focus on delivering deployable tools that strengthen explainability, privacy, and governance across modern AI and agentic systems.
https://www.dataiku.com/press-releases/dataiku-launches-575-lab/
Krisp launches real-time Voice Translation SDK
Enables CX (customer experience) platform developers to embed real-time multilingual voice-to-voice translation into live customer conversations.
https://krisp.ai/blog/real-time-voice-translation-sdk/
300-node clusters now supported in CockroachDB
CockroachDB v25.4 supports 300-node clusters and 1PB of data per cluster at 2.2M tpmC on transactional workloads to improve scalability for AI-era applications.
https://www.cockroachlabs.com/blog/300-node-clusters-supported-cockroachdb/
The Gilbane Advisor is authored by Frank Gilbane and is ad-free, cost-free, and curated for content, computing, data, web, and digital experience technology and information professionals. We publish recommended articles and content technology news most Wednesdays. We do not sell or share personal data.
Subscribe | View online | Editorial policy | Privacy policy | Contact
