I’ve been diving into GitHub’s upgraded Copilot (spoiler: it’s way more than autocomplete now), tracking OpenAI’s shifting release plans, and watching Alibaba quietly drop a big signal with Qwen 3.
If you’re building with LLMs, this one’s worth your time!
🚀 GitHub Supercharges Copilot: Smarter, More Autonomous, and Packed with New Models
GitHub just gave Copilot a serious upgrade, and it's a big deal for devs.
First up, there's a new “Agent Mode” in VS Code. Basically, you can ask Copilot to handle multi-step tasks, and it’ll take over, updating files, writing tests, running commands, even suggesting changes across your whole project.
Think of it as your AI pair programmer with initiative.
Even better, Copilot now supports multiple AI models, not just OpenAI’s.
You can switch between Claude 3.5, Gemini 1.5, and OpenAI’s newest models. So depending on your project, you pick the brain that works best.
Oh, and GitHub added MCP (Model Context Protocol) support, meaning Copilot can plug into external tools and databases. More context, smarter results.
Bottom line: Copilot’s growing from autocomplete helper into a full-on coding assistant.
If you haven’t tried it lately, now’s the time.
Read more here!
🔥 Alibaba Teases “Qwen 3” – Its Next-Gen Open-Source LLM
Alibaba is teasing the release of Qwen 3, its next-gen open-source large language model, and it's looking like a major step forward in the AI arms race.
Slated to drop as early as April, this model builds on the success of the previous Qwen family. If you’ve been following open-source LLMs, you’ll know Qwen has already made serious waves on leaderboards like Hugging Face’s Open LLM rankings.
What sets Qwen 3 apart, at least from early hints, is a push toward efficiency and accessibility. Alibaba is reportedly optimizing it to run across a range of devices, from desktops to mobile phones. That’s a bold move aimed at bringing powerful AI into more real-world applications, beyond just research labs and cloud servers.
It’s also a smart strategic play. By keeping Qwen 3 open-source, Alibaba is positioning itself as a global contender not just in AI performance, but in openness and collaboration, directly challenging the closed-model approach of players like OpenAI and even China-based rival DeepSeek.
You can read more here: reuters, supplychain, qwen3
🤖 GPT-5 Is Coming (But Not Just Yet)
Sam Altman just shared an update on OpenAI’s release roadmap, and there’s a shift in plans.
First up, o3 and o4-mini are comming in the next few weeks. These are likely evolutionary steps, building on the o-model family that's been powering tools like Copilot.
But the big one, GPT-5, is now a few months away.
Altman says the delay is all about quality: the team wants to make it “much better”, ensure seamless integration across systems, and build enough infra to handle what’s expected to be massive demand.
It’s a reminder that building truly next-gen models isn’t just about the weights, it’s about scaling, tooling, and getting the experience right.
Looking forward to seeing what GPT-5 brings when it lands.
That’s it for today’s roundup! I’ll be back in couple days with more updates that matter to devs, builders, and curious minds.
👋 In the meantime —> stay curious!