Behind Martin: Building an AI assistant from the ground up

Martin’s developers genuinely care about your experience—how you use it, what you gain from it, and how it fits into your...

You can now edit videos using Meta AI

Users can choose from 50 preset options to edit videos that are up to 10 seconds long.

OpenAI strikes cloud services deal with Google despite rivalry

Having OpenAI as a customer could significantly increase Google Cloud’s numbers.

OpenAI launches o3-pro with better reasoning and accuracy

OpenAI says people testing the model say o3-pro is better than the earlier o3 in almost every category.

OpenAI’s open model delayed until later this summer

In the past, the company has been criticized for not sharing its models openly with the research community. 

Anthropic Claude’s AI-written blog is no more

Anthropic framed Claude Explains as a way to demonstrate how human editors and AI can work together.

X (Twitter) locks down data, no more free AI training from tweets

Third-party collaborators can still train AI models using X’s data, unless users specifically opt out.

Claude Gov: Anthropic’s AI secret weapon for U.S. security

Anthropic says Claude Gov models went through the same intense safety checks as their other AI systems. 

Reddit is suing Anthropic for allegedly using its data to train AI

Reddit says it told Anthropic not to use its content, but claims that the company ignored that warning.

Anthropic’s Claude AI is writing its own blog with human help

Anthropic says the whole point is to show how AI can support human work, not replace it.

DeepSeek accused of using Google Gemini to train new AI model

Since the internet is filled with AI-generated content, it can be hard to tell where training data originally came from.

Meta plans to use AI for its product risk assessments

The goal of this AI-based approach is to make the review process faster and more efficient.

Google’s AI Edge Gallery can save and run AI models offline

The app runs these AI models locally using the phone's processor.

DeepSeek debuts lighter R1 AI model with better math reasoning

DeepSeek R1's lighter version beats Google’s Gemini 2.5 Flash on a tough math test called AIME 2025.

Perplexity’s new tool can craft reports, spreadsheets, and more

Perplexity Labs is available on iOS and Android Apps and on web browsers.

Research shows AI is already cutting entry-level jobs in tech

These roles involve tasks like writing code, researching data, or organizing files, all of which AI can now do quite well.

Anthropic rolls out a voice mode for its Claude chatbot apps

It’s designed to make conversations feel more natural and human-like.

OpenAI may soon let you sign in to third-party apps with ChatGPT

The Sign in with ChatGPT feature is still in the early stages.

OpenAI’s operator AI agent gets smarter with new O3 model

It act like a digital assistant that can work mostly on its own, without needing much input once a request is...

Anthropic’s Claude goes off the rails, blackmails developers

Claude threatened the engineers to expose their affair if the company went ahead with replacing it.