Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Over the past year, I’ve focused heavily on AI Evals, both in my consulting work and teaching. A question I get constantly is, “What’s the best tool for evals?”. I’ve always resisted answering directly for two reasons. First, people focus too much on tools instead of the…
MCP UI extends the Model Context Protocol to enable AI agents to return fully interactive UI components. It solves the critical challenge that commerce experiences require visual and interactive elements like product selectors, image galleries, and cart flows. This open-source…
Outtake uses GPT-4.1 and OpenAI o3 agents to detect security threats 100x faster.
Commerce is a dynamic ecosystem where our mission is to empower every merchant to succeed. We optimize each step of their journey—from product creation to customer delivery—using advanced tools, infrastructure, and partnerships to solve a complex optimization challenge.
What makes a good leader? What do good leaders do? And commando, soldier, and police leadership.
Sam Altman commentary on Intelligence Age and AI's role in solving major societal problems.
Google releases Veo 2 video generation in Gemini and Whisk, creating eight-second high-resolution videos from text or image prompts.
Claude introduces extended thinking feature enabling longer internal reasoning chains for complex problem-solving.
OpenAI solicits infrastructure partnerships across power, data centers, construction, and equipment for AGI compute buildout.
OpenAI o1 reasoning models demonstrate performance gains in coding, strategy, and research tasks.
Quantum physicist Mario Krenn uses OpenAI o1 for complex physics problem-solving.
OpenAI released SWE-bench Verified, human-validated benchmark for code-solving model evaluation.
Anthropic expands model safety bug bounty program to solicit vulnerability reports and strengthen adversarial defenses.
Anthropic partners with AWS and Accenture to deliver enterprise AI solutions built on Claude.
We explore large-scale training of generative models on video data. Specifically, we train text-conditional diffusion models jointly on videos and images of variable durations, resolutions and aspect ratios. We leverage a transformer architecture that operates on spacetime…
New vision models are now available: LLaVA 1.6, in 7B, 13B and 34B parameter sizes. These models support higher resolution images, improved text recognition and logical reasoning.
WHOOP delivers personalized fitness and health coaching with GPT-4.
Guest post by Weights & Biases Weights & Biases (W&B) provides AI developers with powerful tools to build better models faster and is used by almost every top LLM research lab today. To help support internal and external users, W&B built WandBot – a question-answering bot built…
We've trained a model to achieve a new state-of-the-art in mathematical problem solving by rewarding each correct step of reasoning (“process supervision”) instead of simply rewarding the correct final answer (“outcome supervision”). In addition to boosting performance relative…
We are improving our AI systems’ ability to learn from human feedback and to assist humans at evaluating AI. Our goal is to build a sufficiently aligned AI system that can help us solve all other alignment problems.
We built a neural theorem prover for Lean that learned to solve a variety of challenging high-school olympiad problems, including problems from the AMC12 and AIME competitions, as well as two problems adapted from the IMO.
We’ve trained a system that solves grade school math problems with nearly twice the accuracy of a fine-tuned GPT-3 model. It solves about 90% as many problems as real kids: a small sample of 9-12 year olds scored 60% on a test from our dataset, while our system scored 55% on…
What questions do they answer? How do they compare? What open-source solutions are available?
Should I switch from a regex-based to ML-based solution on my application?
Should I join a start-up? Which offer should I accept? A simple metaphor to guide your decisions.
We’ve trained a pair of neural networks to solve the Rubik’s Cube with a human-like robot hand. The neural networks are trained entirely in simulation, using the same reinforcement learning code as OpenAI Five paired with a new technique called Automatic Domain Randomization…