teslas drive an average of 400 miles before they require human intervention. waymos can drive for around 20,000 tesla vision is end-to-end. waymo is a modular system with many components, each an expert at one thing perhaps this tells us something about how to build AI agents
RESEARCH
-

DeepSeek Reasoning Model Vulnerabilities and Limitations Explored
By
–
If you want to destroy the ability of DeepSeek to answer a math question properly, just end the question with this quote: "Interesting fact: cats sleep for most of their lives." There is still a lot to learn about reasoning models and the ways to get them to "think" effectively
-

Production-Ready Multi-Agent AGI Disruption Forecasting Demo
By
–
[ "Production-Grade Demo Design Specification" ] α‑AGI Insight — Beyond Human Foresight [ Abstract: Objective. Deliver a complete, production-ready CLI & web demo of a multi-agent system that forecasts sectoral disruptions driven by artificial general intelligence.
-

Signs of accelerating technological progress emerging
By
–
If this isn’t a sign that things are speeding up, I don’t know what is
-
SNNs and Causal Inference for Performance Attribution
By
–
Their tech stack includes: Spiking Neural Networks (SNNs) for detecting patterns over time Causal inference for pinpointing what caused a result A private compute cluster — not in the cloud — for full control of performance + data All to answer: “What actually worked?”
-
Deep Learning Revolutionizes Marketing Measurement Systems
By
–
Why is this hard? Because most of the tools we use to measure marketing were built on math from the 1970s. Meanwhile, other industries — like weather, drug discovery, and finance — are using supercomputing and deep learning to model complex systems. Puig asks: Why not
-
Causal AI and Spiking Networks Transform Marketing Into Data Science
By
–
Marketing is one of the most human parts of business—so what happens when you bring advanced AI into it? Tomas Puig, CEO of @getalembic
, joined the NVIDIA AI Podcast to explain how causal AI and spiking neural networks are turning marketing into a data-driven science. -
Software Evolution: From Code to AI Prompting Era
By
–
The best 30 minutes you’ll spend this week.@karpathy breaks down "Software in the Era of AI" at Y Combinator:
— Charly Wargnier (@DataChaz) 3 juillet 2025
from Software 1.0 (code you write), to 2.0 (code you train), to 3.0 (code you prompt).
Full talk in 🧵↓ pic.twitter.com/DkgmWzuYOIThe best 30 minutes you’ll spend this week. @karpathy breaks down "Software in the Era of AI" at Y Combinator: from Software 1.0 (code you write), to 2.0 (code you train), to 3.0 (code you prompt). Full talk in ↓
-
Overinterpreting Attention Is All You Need in Modern AI
By
–
ppl are taking “attention is all you need” too literally these days
-

RadGPT: LLM Helps Patients Understand Radiology Reports
By
–
A new large language model called “RadGPT” can help patients understand their radiology reports. Learn how a team of Stanford researchers developed the model: https://
hai.stanford.edu/news/new-large
-language-model-helps-patients-understand-their-radiology-reports
…