"Debugging used to mean finding bugs in your code. Now it means explaining to your AI assistant why its vibe generated function is somehow both syntactically perfect and completely wrong."

Hello Builders! 👋

🔥 Today's Top Story

Google just partnered with Replit, this matters because Google's throwing its weight behind the "just describe what you want" approach to building software, directly competing with Claude Code, Cursor and the VS Code AI extensions crowd.

Replit's browser-based IDE now gets Google's Gemini models baked in, the skeptic in me wonders if this is Google playing catch-up after missing the boat. But for builders? More competition means better tools, faster. The "I'll just build it myself this weekend" era is getting very real.

🚀 Ships & Launches

Gemini Deep Think Goes Live - Google's answer to o1 drops for Ultra subscribers with extended reasoning mode baked into the app.

Capsule Adds Claude Sonnet/Opus 4.5 - Pick your poison between the two newest Claude models for coding, finally some model flexibility in the wild.

NaryAI Shows Brand Perception - New tool reveals how AI models actually see your brand, which is either fascinating or terrifying depending on your SEO.

📺 Learn & Build

Copilot's Secret Goldeneye Model - Ve Sharma breaks down the mysterious "Raptor Mini" that appeared in VS Code and what it means for your workflow.

Run AI Coding Models Locally - Abid Ali Awan rounds up five lightweight models that won't send your code to the cloud or murder your RAM.

Video Segmentation for Precise Editing - Justine Moore shows how to target specific objects in video (like turning one dog into a mini pony) when text prompts fail.

Google's 30+ Free AI Agents - Google's surprise drop of production-ready agents you can actually study, fork, and ship without hitting paywalls.

Vibe-Coded Data Pipelines - Python beats DSLs for data pipelines when you're just vibing through the implementation instead of wrestling YAML configs.

💬 Builder Conversations

Testing Vibe-Coded Projects - Cosmo asks the question we're all thinking: how do you automate testing when AI wrote half your codebase?

Best Model for iOS - Vedanth tried Gemini 3 for vibe coding an iOS app and wasn't impressed, wants recommendations for better alternatives.

Video Models Gone Viral - Justine Moore points out that video AI has finally crossed into mainstream viral territory, with Grok-generated content spreading beyond tech Twitter.

Will Smith Spaghetti Redux - The iconic AI benchmark gets remade with Kling 2.6, showing just how far video generation has come in under three years.

📰 Industry Moves

Harvey Raises Series F - a16z Growth leads another round for the legal AI platform that's actually getting lawyers to use AI daily.

Micro1 Hits $100M ARR - Jumped from $7M to $100M ARR in under a year competing with Scale AI on data labeling and training.

Chicago Tribune Sues Perplexity - Another copyright lawsuit drops, this time targeting Perplexity's RAG system for allegedly lifting content without attribution or payment.

Anthropic CEO Calls Out - Dario says competitors are "YOLO-ing" on spending amid bubble concerns, which is rich coming from a company burning billions.

Meta Slashes Metaverse Budget - Planning 30% cuts to Reality Labs as Horizon Worlds remains a ghost town nobody asked for.

Netflix Acquiring Warner Bros - Massive M&A move that would reshape streaming and content production if regulators don't kill it first.

💎 Hidden Gems

Claude's Plugin System Unlocked - Bundle custom slash commands, subagents, and automations into shareable plugins that actually make agentic workflows feel less janky.

100 Trillion Token Deep Dive - a16z analyzes reasoning model usage patterns across massive scale, revealing how developers actually use these tools in production.

One More Thing...

Wild that Google's making moves with both Replit and dropping Deep Think mode in the same week, feels like they're finally waking up to the vibe coding reality that builders have been living in for months. Meanwhile GitHub's out here with secret Goldeneye models we're all reverse-engineering.

What are you shipping this weekend? Testing any of these new AI coding tools?

Keep shipping,

P.S. If you haven't tried prompting with vibe coding energy yet, you're missing out. Specificity is overrated when the models can read your mind anyway. 😏

Keep Reading

No posts found