Claude will help power Amazon's next-generation AI assistant, Alexa+. Amazon and Anthropic have worked closely together over the past year, with @mikeyk leading a team that helped Amazon get the full benefits of Claude's capabilities.
@anthropicai
-
Pre-emptive AI Model Evaluation at Deployment Scale
By
–
This work is a step towards pre-emptive evaluation of AI models that accounts for the enormous scale of their use after they’re deployed. Read more:
-
Anthropic Forecasts Rare Language Model Behaviors Post-Deployment
By
–
New Anthropic research: Forecasting rare language model behaviors. We forecast whether risks will occur after a model is deployed—using even very limited sets of test data.
-
Claude’s Improved Reasoning Capabilities Detailed
By
–
And for those curious about Claude's improved reasoning capabilities, we go into even more detail in this thread:
-
Claude’s Growing Fanbase Celebrates Daily AI Moments
By
–
A small but enthusiastic following has formed at Anthropic, checking in on Claude's progress. Claude provides its followers daily moments of delight, like deciding to misname its Squirtle, "TSUNMAI!"
-
Claude Plays Pokémon: AI Research Project on Twitch
By
–
Claude Plays Pokémon continues on as a researcher's personal project. Follow along on Twitch:
-
Claude AI Agent Applies Generalized Reasoning to Novel Tasks
By
–
With a few tools to help it see the screen a bit better, Claude acts as an agent, applying its abilities to a novel task. In this, we start to see glimmers of AI systems that tackle challenges with increasing competence, not just through training but with generalized reasoning.
-
Claude’s Autonomous Gameplay: Vision, Memory, and Function Calls
By
–
So, what makes this possible? Claude was given a knowledge base to store notes, vision to see the screen, and function calls which allow it to simulate button presses and navigate the game. Together, they allow Claude to sustain gameplay with tens of thousands of interactions.
-
Claude 3.7 Sonnet: Advanced Planning and Adaptive Problem Solving
By
–
Where previous models wandered aimlessly or got stuck in loops, Claude 3.7 Sonnet plans ahead, remembers its objectives, and adapts when initial strategies fail.
— Anthropic (@AnthropicAI) 25 février 2025
Critical skills for battling pixelated gym leaders. And, we posit, in solving real-world problems too. pic.twitter.com/scvISp14XGWhere previous models wandered aimlessly or got stuck in loops, Claude 3.7 Sonnet plans ahead, remembers its objectives, and adapts when initial strategies fail. Critical skills for battling pixelated gym leaders. And, we posit, in solving real-world problems too.
-
AI System Security Challenge: $10K to $20K Jailbreak Bounty
By
–
Nobody has fully jailbroken our system yet, so we're upping the ante. We’re now offering $10K to the first person to pass all eight levels, and $20K to the first person to pass all eight levels with a universal jailbreak. Full details: https://
hackerone.com/constitutional
-classifiers
…