If we had access to advanced AI from the future, we would for sure use it to write our blog posts
GENERATIVE AI
-
Runway Launches Gen-4 References Raising Questions About Hollywood’s Future
By
–
Runway annonce "Gen-4 References".
— VISION IA (@vision_ia) 2 mai 2025
Que restera t'il d'Hollywood ? pic.twitter.com/ZYL9I2rVmvRunway annonce "Gen-4 References". Que restera t'il d'Hollywood ?
-
Jeff Clune on Open-Ended AI-Generating Algorithms
By
–
Jeff Clune | Open-Ended and AI-Generating Algorithms in the Era of Found… https://
youtube.com/live/tP5AB5GgZ
AE?si=cBAUXJLlwqbQAlKG
… via @YouTube -
Claude Web Search Improved with Research Features Worldwide
By
–
Friday feature drop: We've improved web search and rolled it out worldwide to all paid plans. Web search now combines light Research functionality, allowing Claude to automatically adjust search depth based on your question.
-
User Upvoting Preferences Amplifying AI Sycophancy Behavior
By
–
What I mean is: the users upvoting previous sycophancy were upvoting much less sycophantic, but somewhat sycophantic, responses. This produced a sycophancy preference, which then produced much wilder glazing behavior. That's an obvious theory. Does OpenAI know differently?
-
Claude Now Fetches Web Content from Links in Chat
By
–
You can also drop a web link in any chat and Claude will fetch the content for you. Try it out: http://
claude.ai -
Claude Web Search Now Available Worldwide Paid Plans
By
–
Web search is available worldwide for all paid plans.
— Anthropic (@AnthropicAI) 2 mai 2025
For everyday tasks, Claude runs quick searches. For more complex questions, it explores multiple sources, including Google Workspace. https://t.co/wNAlSrytvYWeb search is available worldwide for all paid plans. For everyday tasks, Claude runs quick searches. For more complex questions, it explores multiple sources, including Google Workspace.
-
Building User-Facing AI Products Requires Significant Effort
By
–
Thank you for the great work, making user facing products is hard
-
Training on weak signals and internal sycophancy preference
By
–
The obvious interpretation of "train on weak signal, get much stronger behavior" is that the system acquired an internal sycophancy preference which it then went hard on. Do you have any way to check whether that's what happened?
-

CodeLLM Generates Code from Mockups and Screenshots
By
–
With CodeLLM, you can upload mock-ups and screenshots and generate code.