Six things shipped this week that change what you can build. Here’s the short version, and why each one matters for your projects.
Laravel 13.14 ships a first-party AI SDK
Laravel 13.14 is out, and the big one for us is the new first-party AI SDK — a single API covering text generation, tool-calling agents, embeddings, audio, images, and vector stores. The release pushes PHP attributes further too, so configuration sits right next to the code it governs, and it raises the minimum PHP version to 8.3.
Why it matters: PHP and Laravel run most of the projects here. An official AI SDK means you can drop a chatbot or smart search into a capstone app without wiring up some third-party library. Source: Laravel.
Python 3.15 hits late beta, and free-threading is real now
Python 3.15 reached its second-to-last beta on June 2, with the release candidate due July 28. The longer arc matters more: free-threaded, no-GIL Python has finally grown up this year, and the uv toolchain has taken over installs.
Why it matters: start a Python project today knowing what 3.15 changes — and that real multithreading is now practical — and you skip a painful rewrite later. Source: Python.org / InfoWorld.
VS Code 1.123 adds 1M-token context and side-by-side agents
The June 3 update is hefty. You get 1M-token context windows for Anthropic and OpenAI models, several agent sessions running side by side, a research agent that hands back a cited Markdown report, and native Mermaid and YAML previews. There’s a smart safety touch, too: extensions now wait two hours before auto-updating, so a compromised release can’t hit you the second it ships.
Why it matters: it’s the free editor most students already run, and these turn it into a real AI workspace at no extra cost. Source: Visual Studio Code.
Gemini 3.5 Flash and GPT-5.5 land, and prices fall
Two model releases worth noting. Gemini 3.5 Flash claims flagship-level quality at about four times the speed of the last generation. GPT-5.5 is now the default in ChatGPT, tuned for long, multi-file agentic coding. Google also cut its top plan from $250 to $200 and added a $100 developer tier.
Why it matters: cheaper, faster models stretch a student budget — more you can actually demo per dollar. Source: LLM-Stats / Augusto Digital.
Claude Code passes $2.5B as AI-assisted coding goes mainstream
Fresh survey data puts 90% of developers using at least one AI tool at work, and Anthropic’s Claude Code has reportedly hit a $2.5B annual run rate about nine months after launch. Awareness climbs every quarter.
Why it matters: AI-assisted coding isn’t a side experiment anymore — it’s the baseline employers assume, so get fluent with one tool before you graduate. Want the deeper look at how these are packaged? Read our explainer on Claude skills. Source: JetBrains Research.
Google’s Colab CLI puts free GPU/TPU in your terminal
Google shipped a Colab CLI that lets you — and AI agents — run local code on remote Colab GPU and TPU runtimes. Translation: serious hardware, borrowed straight from your command line.
Why it matters: training or testing an ML model for a thesis without owning a GPU just got cheaper. Source: Google / LLM-Stats.
Building with any of this? Grab a ready-to-run system or a capstone guide from our source code projects and AI guides, and ship faster.