Google’s new model lets you steer AI-generated worlds

Welcome back. Google wants you to build entire virtual worlds using AI and its latest model promises photorealistic, interactive environments you can explore in real time. Meanwhile, Apple just dropped a new MacBook Pro packed with serious AI power. 

Also: How to keep your Discord messages off Google Search, a browser running local AI models, and why a senior dev believes domain knowledge remains irreplaceable in the AI era. 

Google’s new model lets you steer AI-generated worlds in real time: The search giant just unveiled Genie 4, its most realistic world model yet. Gemini 3.8 Flash runs alongside it, planning what happens next and checking frames as Genie renders them. It’s pitched for games, interactive learning, and robotics, where teams can test agents in simulation before real-world deployment. Scenes reportedly stay consistent for around two minutes, though longer runs still drift. API and AI Studio partners get access soon. You can join the waitlist here.

Apple’s new laptop gives local AI a speed boost: The iPhone maker just unveiled a 14-inch MacBook Pro with its M6 chip, and local AI is a big part of the pitch. Apple says Xcode builds compile up to 1.3x faster than on M5. Each GPU core also packs more Neural Accelerators which are dedicated AI circuits that speed up local LLM prompt processing by as much as 25%. Memory still tops out at 32GB, so stick with smaller models. Prices start at $1,999 and you can pre-order it here.

Liquid AI’s agent-ready models now fit on your phone: The AI startup just released LFM3, six open-weight models covering text, vision, audio, and decision-making. Each has around 2B parameters and can be fine-tuned on a single GPU, so you can build a specialized agent and run it on-device. The text model takes up about 1 GB and, per the company, beats the larger Qwen3.5-9B on browser and business-app tasks. Agentic coding and general knowledge are still weak spots. Explore the docs and use cases.

Not because you don't care. Because after 6 hours of building, typing paragraphs about what you built feels like punishment.

So the PR description stays vague. The README gets "TODO." The architecture decision lives in your head until you leave the company.

Wispr Flow: syntax-aware voice dictation that works in every app. Speak your PR descriptions, commit messages, and docs. camelCase, snake_case, and acronyms stay intact.

Engineers at OpenAI and Vercel use it daily. Available on Mac, Windows, iPhone, and Android. The best docs are the ones that actually get written.

Try free

Claude got the job. Coding interviews might finally be dead, or at least they are at Anthropic. Engineering manager Felix Rieseberg just announced that he’s scrapped them from his team's hiring pipeline, and his logic is straightforward. Claude already breezes through the React challenges and assembly questions candidates face. So why should he bother evaluating developers on tasks that AI clearly handles better? 

Out with the old. Now coding interviews were never flawless, and though Rieseberg admits they once predicted team success fairly well, that has changed. Engineers who quickly broke down problems and found elegant fixes once thrived. Not anymore. Today, some of his top-performing engineers shine no matter their technical scores, rendering the conventional coding interview essentially obsolete. 

Hunger beats tenure. Relatively new engineers are racing past colleagues who are probably better coders and have been around much longer. For example, two newer Anthropic engineers built a functional graphical interface for Claude Code over a single weekend. Just eight days later, the team shipped the first public version. 

The new way. Rieseberg hasn’t figured out the perfect replacement yet, but he’s already looking for different signals. Instead of coding puzzles, he’s scanning resumes and portfolios for small, working prototypes that show taste, empathy, and thoughtful decision-making. If you are trying to stand out, his advice to junior engineers applies here as well. You should try building each prototype around a single solid idea instead of burying a few strong ones inside multiple half-baked attempts. You can check out his full essay here.

How to learn Terraform (the most important tool in DevOps): A senior engineer explains why Terraform still matters even when AI can write the config for you, then walks through how it actually works. You’ll cover providers, variables, state, and outputs, plus the mistakes AI-generated infrastructure can make if you don’t understand what it’s changing.

Appllama Skills (3.4k ⭐): Give your coding agent a better eye for mobile UI. The repo helps Claude Code, Cursor, Codex, and other agents study patterns from top-grossing apps, then use them to build or improve Expo and React Native screens with better navigation, motion, accessibility, and simulator-tested polish.

How to stop AI code reviews from spiraling (by Anthropic): This playbook shows how to keep agent-written PRs from getting stuck in endless review loops. You’ll learn how to verify work before the PR, use REVIEW.md to catch recurring issues, cap review rounds, switch later passes to diff-only review, and turn incidents into evals so the reviewer improves over time.

Logs pile up and nobody reads them. Bugs reach you through users, not log files. A former Meta L8 who also did stints at Microsoft and Atlassian shared a two-prompt setup that turns those logs into a daily improvement report.

Each morning a ranked fix list lands in “reports/”

Two key engineers just left Anthropic ahead of its IPO. Their farewell posts hint at why they’re walking away.

If you enjoyed this newsletter, you should definitely check out our IT and cybersecurity newsletters. Subscribe using the links below:

We put together 20+ hands-on guides covering coding agents, agentic engineering, and leadership frameworks to help you level up your career. You can get them here.

Advertise with us: Join companies like Google, IBM, and Datadog. Showcase your product to our 350K+ engineers and 150K+ followers on socials. Get in touch.

You can also reply directly to this email if you have suggestions, feedback, or questions.

Until next time — The Code team