On this week's AI Tripod Podcast, we discuss Salesforce's shocking research exposing the reality behind AI agent hype - their testing revealed only 30% of MCP tools actually work in real-world scenarios. We break down what this means for the future of AI automation and cover the week's biggest developments.Also discussed: DeepSeek 3.1 now available, ChatGPT's new pricing in different countries at just $4/month compared to $20 in the US, some token optimisation strategies, and Quinn image edit revolutionary open-source image editor model that's challenging Adobe with precision text editing and a challenger to Gemini new image editing model.Plus: the new agents.md standard streamlining AI development workflows across different IDEs.Topics Covered:- Salesforce MCP study: 70% failure rate- DeepSeek 3.1 vs Claude performance - ChatGPT pricing hacks- Quinn 3 image editor demo- New agents.md developer standardAdditional sources:Salesforce AI research paper: https://huggingface.co/papers/2508.14704 Quinn model: https://qwenlm.github.io/blog/qwen-image-edit/#AI #ChatGPT #SalesforceResearch #AITools #TechNews #OSSImageEditorAIModel