"How long can Nvidia stay ahead of Chinese competition?" (shorturl.at/dp2xT) JK--Good chart showing how far ahead Nvidia's chips are, performance-wise, vis-a-vis US and overseas rivals.

Please check out my new book: “The Mimetics: Poems of original authorship within frames of a repurposed nature.” Available on Kindle. Latest installment in the ongoing series called "Pieces of Fate." shorturl.at/zx4Gv

"Machine Bullshit: Characterizing the Emergent Disregard for Truth in Large Language Models" (shorturl.at/ygbby) JK--Good research. Subtypes of "bullshit" include empty rhetoric, weasel words, unverified claims, sycophancy, and a term I've never heard before: "paltering."

"[Mistral's] contribution to a global environmental standard for AI" (tinyurl.com/7x86ym4y) JK--I like their lifecycle analysis framework for LLM environmental impact.

"AI Models Not Ready to Make Scientific Discoveries" (tinyurl.com/4st2z554) JK--Interesting research. But I interpret as implying AI models can overlearn world model so much it blinds them to more fundamental paradigm principles. Overlearn Kepler, but ignore Newton.https://tinyurl.com/4st2z554

"Vending-Bench: A Benchmark for Long-Term Coherence of Autonomous Agents" (shorturl.at/rk22L) JK--Very cool research to gauge ability of autonomous agentic AI to run a vending machine, hence autonomously make money.

"Grok 3 AI is now free to all X users – here's how it works" (shorturl.at/YqV32) JK---Tread carefully. Has anybody thoroughly checked to see whether it's been trained on MAGA/DOGE propaganda? Here's my initial take. So far, so good.

Here’s the very first question I asked of DeepSeek and its very first response. Take this into consideration if you choose to adopt it.

"Reasoning Language Models: A Blueprint" (tinyurl.com/mpatju6d) JK--Hmm. New acronyms with legs: RLM (reasoning language models) and LRM (large reasoning models).

"Large-scale moral machine experiment on large language models" (shorturl.at/Knpdx) JK--Interesting research. "Moral machine experiment"? Wasn't that that unlistenable Lou Reed album?

"Agent-as-a-Judge: Evaluate Agents with Agents" (shorturl.at/IWwlq) JK--Interesting research into what's obviously an AI-accountability slippery slope.

"Scaling Laws for Pre-Training Agents and World Models" (shorturl.at/rnSMh) JK--The more data & processing power you throw at LLMs and other foundation models, the better they perform--but you still have to twist a lot of dials depending on task/model/deployment.