VS Code Leverages Custom AI Models with Copilot

Update for AI workflows in VS Code. ✨ Bring your own language model/API key is now available in GitHub Copilot. More flexibility. More control. More experimentation.👇 https://lnkd.in/eu-WFuuy

This a major announcement for mission and support of coding assistance in air gapped environments. Really recommend you check it out.

BadClaude is a tool in which you can whip Claude to work faster. Built by github user GitFrog1111, you can interrupt Claude running code. Clicking the mouse triggers the whip and pushes prompts to increase the AI's working speed. Code available online, see the comments.   The question is, do you feel getting whipped by AI?

Most AI coding assistants re-read your entire codebase for every task 🧑💻 → More tokens → Higher cost → Slower responses 💡 Came across a smart approach that fixes this using a code graph + incremental context Instead of sending everything to the model, it only sends what actually matters. 📉 Result: → Fewer tokens → Better context → Improved code review quality This is the kind of optimization that will matter a lot as we scale AI workflows. 👇 Comment “Git” and I’ll share the GitHub repo link #AI #AICoding #Claude #LLMOps #DeveloperTools #GenAI

The power shift in AI is happening in the developer IDE, and Anthropic is winning. By December 2025, they held 54% of the enterprise code generation market. OpenAI has been relegated to just 21% of this segment. The real war for enterprise AI isn’t about chat; it’s about writing production-ready code. 🔗 Link in the comment section

Lately, I’ve been noticing a subtle shift in how AI tools are evolving. It’s less about getting better answers, and more about actually getting work done. I was going through the latest updates to OpenAI Codex, and what stood out wasn’t the coding part. It’s that it can now: interact with your computer work across different tools remember context over time and even continue tasks on its own That starts to feel very different from the AI most of us are used to. It’s not just assisting anymore. It’s beginning to take ownership of parts of the workflow. Which also changes how we use it. Instead of thinking “how do I use this tool better”, it becomes more about “what should I delegate, and how do I guide it well”. Feels like a small shift on the surface, but a pretty big one in how work might evolve from here. Curious how others are thinking about this.

LLMs are getting great at writing code. Yet their output still requires careful human review. If AI writes more of our code, how do engineers stay sharp enough to review it? Because strong code reviews do not come from prompts. They come from experience in solving problems and implementing code. As we adopt AI, we need to be just as intentional about preserving that experience, even as more of the code is generated for us.

To build V11, the Evernote team indexed 3 billion notes. This laid the foundation for the product's biggest update since 2020, with three new AI-powered features: 🔎 Semantic Search Find what you mean, no need for exact keywords. Built by indexing 400 TB of data and 1.2 million daily searches. 🔮 AI Assistant (developed in close collaboration with OpenAI) Chat with Evernote to find content, extract key points from a note, generate text, move notebooks, and more. ✍️ AI Meeting Notes Transcribe and summarize your meetings in 100 languages. Watch the video for more on V11. 👇

The future is becoming no-code. It doesn't mean coding is dead. It means code smarter, code faster with AI. Deliver faster.

Everyone’s excited about AI coding agents right now.They can write code, fix bugs, and even run commands . But the real question isn’t capability, it’s safety. How do you let them work without giving them full access to your machine? That’s where Docker Sandbox comes in. It runs AI agents inside an isolated microVM with its own boundaries, so they can build and test things without touching your actual system. We’re moving from AI that suggests to AI that acts and control is going to matter just as much as intelligence. I wrote a quick breakdown : https://lnkd.in/g2qV3JZk