Rated by different models instead of users
GENERATIVE AI
-

Veo 2 Video Generation AI Integrated Into Google Gemini
By
–
Veo 2 is super fun to play with, and people have been creating some amazing videos with it. Its implicit understanding of the physics of the world is kind of mindblowing. Looking forward to seeing more people enjoy it now that it’s part of @GeminiApp! https://t.co/bMqLCmNHmd
— Demis Hassabis (@demishassabis) 15 avril 2025Veo 2 is super fun to play with, and people have been creating some amazing videos with it. Its implicit understanding of the physics of the world is kind of mindblowing. Looking forward to seeing more people enjoy it now that it’s part of @GeminiApp
! -
Research Beta and Google Workspace Integration Now Available
By
–
Research is available in beta for Max, Team, and Enterprise plans in the US, Japan, and Brazil. The Google Workspace integration is now available for all paid plans. Go try them out: https://
claude.ai -

Claude.ai Launches Google Integration and Research Features
By
–
We've added two major features to claude dot ai today that have been 10x productivity boosts for me: – Google Docs, Calendar, and Gmail integration
– Research – our first step toward an agentic researching agent Most of today's AI research tools fall into two extremes: instant -
Provide examples for complex outputs across apps
By
–
10/ For complex outputs across apps, provide examples. GPT-4.1 follows examples precisely. Show exactly what you want:
"Answer in this format:
SUMMARY: One sentence
ANALYSIS: Three paragraphs
RECOMMENDATION: Bullet points" -
Provide clear format for creative tasks with GPT-4.1
By
–
8/ For creative tasks, give it a clear FORMAT. GPT-4.1 excels at following structural guides: "Write a story with:
Setting: [x]
Characters: [x]
Plot points: [x]
Theme: [x]
Length: [x]" Using this approach in tools like Windsurf / Cursor gives better creative output. -
Use chain of thought for research and analysis tasks
By
–
7/ Use "chain of thought" for research or analysis tasks. Ask GPT-4.1 to:
– Break down what you're asking
– Consider key factors
– Analyze options
– Reach a conclusion This structure produces dramatically better results than just asking for an answer. -
Prompt “think step by step” improves GPT-4.1 accuracy
By
–
4/ Ask it to "think step by step" before answering complex questions. GPT-4.1 needs explicit permission to show its reasoning. Adding this phrase improves accuracy by ~4%. Example prompt:
"First, think carefully step by step about what documents are needed to answer the query. -

Structure of system prompts: persistence, tools, planning
By
–

2/ Structure system prompts with these 3 main elements – Persistence: Tell the model it's in a multi-turn conversation. It shouldn’t stop until the user's question is fully answered. – Tool Use: Encourage the model to use tools instead of guessing. – Planning: Let the model
-
Be direct and specific with GPT-4.1 for better responses
By
–
1/ Be direct and specific GPT-4.1 follows instructions literally now.
Instead of: "Write something about climate change" Try: "Write a 300-word email explaining 3 climate actions my small business can take" The difference is night and day.