"I don't debug anymore, I just ask ChatGPT 'why is my code sad?' and hope for the best"
What's up builders 👋
Holy ship, what a day. OpenAI just dropped full MCP support in ChatGPT (finally!), Claude is literally making art in Excel files, and Replit's Agent 3 is so autonomous it might file its own tax returns. Plus, we've got $300 billion cloud deals, free AI agent courses from Microsoft, and enough vibe coding content to make your deployment pipeline jealous.
Grab your favorite caffeinated beverage - we're diving into the tools and tricks that'll have you shipping faster than you can say "undefined is not a function."
🔥 Today's Top Story
OpenAI adds full MCP support to ChatGPT - and it's about damn time. Developer Mode now lets you create connectors for write actions, not just search/fetch operations. This means you can finally update Jira tickets, push to databases, and modify external systems directly from ChatGPT conversations.
Why this matters: We've been hacking around ChatGPT's read-only limitations with webhooks and middleware for months. Now you can build proper two-way integrations that actually DO things instead of just fetching data. Think about it - your ChatGPT can now be a full-stack developer that not only writes code but deploys it, updates your project management tools, and probably orders pizza when you're debugging at 2 AM.
🚀 Ships & Launches
• Replit Agent 3 - 10x more autonomous than v2, periodically tests your code without being asked, and handles complex multi-file projects. Warning: burns credits faster but needs fewer corrections
• Incredible Small 1.0 - Agentic AI model that handles 1000+ simultaneous actions and works with massive datasets. Currently in research preview
• DeepMCPAgent Framework - Auto-discovers tools via MCP, builds LangChain/CrewAI agents without boilerplate. Plug-n-play for any MCP-compatible service
• Google Mood Palette Generator - Translates your vibes into color palettes using EmbeddingGemma for semantic matching. Perfect for those "I need a color scheme that feels like debugging at 3 AM"
• Gemini 2.5 Flash Image - Maintains consistent subject identity across any environment. The Past Forward demo lets you remix starter apps with a single prompt
📺 Learn & Build
• Microsoft's Building AI Agents Course - Free 12-lesson course covering what agents are, which frameworks to use, and hands-on coding examples with multi-language support
• Getting Started with Vibecoding - Quick tutorial showing immediate results with vibe-based development. Learn to ship first, optimize later
• ChatGPT Code Reviews Beat Humans - Practical guide showing how ChatGPT caught 312 bugs in 847 PRs that senior devs missed
• Gemini's Nano Banana - Turn any image into miniature action figures with packaging. Prompt: "turn my pet into a plastic action figure next to its packaging" and watch the magic
💬 Builder Conversations
• Claude replicating images in Excel - Alex Albert shows Claude creating profile pics in spreadsheets. We've officially entered the "vibe Excel era" where spreadsheets are canvases
• The deployment reality check - Becky Schmidt reminds us you can vibe your way to working code but environment variables don't care about your vibes
• The Claude commit nightmare - That moment when you realize you asked Claude for multiple updates without committing at a working stage. We've all been there
• Replit Agent v3 credit burn - Early user reports faster credit consumption but fewer corrections needed. The eternal tradeoff of speed vs cost
• AI influencer clothing try-ons - Justine Moore demos image-to-video workflows with Flux Kontext and nano-banana for virtual fashion. The future of e-commerce is here
• Honest AI tool comparisons - Brian's side-by-side of Midjourney, Seedream 4.0, Grok, and Nano Banana with zero paid promotions
📰 Industry Moves
• OpenAI's $300B Oracle Deal - Project Stargate cloud computing deal over 5 years. Your API costs might actually make sense now knowing the infrastructure behind it
• Perplexity hits $20B valuation - Raised $200M, jumping from $18B just two months ago. AI search is printing money faster than your console.log statements
• B2C AI startups hitting 100%+ growth - a16z reports consumer AI companies achieving B2B-level retention rates. The Great Growth Unlock is real
• Mira Murati's Thinking Machines Lab - Former OpenAI CTO's startup working on AI model consistency. Because deterministic outputs are apparently still a dream
• Billions Network GPT - Custom GPT trained on community data for network-specific insights. Clone this pattern for your own community tools
• Stanford CodeMonkeys Paper insights - 40+ page deep dive on parallel AI agents and how they scale. The academic foundation for your next agent swarm
• 5 AI design platforms tested - Alexander Olssen's hands-on review of vibe-design tools worth your time. Skip the hype, use what works
⚡ Quick Links
One More Thing...
Here's a wild thought: We're literally at the point where AI can replicate your face in Excel cells, but we still can't get it to consistently remember environment variables between chat sessions. Maybe the real AGI was the spreadsheet art we made along the way?
The vibe coding revolution isn't just changing how we build - it's changing what we consider "building." When your IDE is a chat window and your debugger is a conversation, the line between coding and creating gets beautifully blurry.
What are you vibe coding this week? Hit reply and show me your wildest AI-assisted creation. Bonus points if it involves Excel art.
Keep shipping,
P.S. - If AI agents become truly autonomous, I'm teaching mine to automatically close Jira tickets as "Won't Fix - Working as Intended." It's not a bug, it's a vibe.