Why we teach Rust the way Google does

When we designed the engineering track of the Fellowship, the most debated decision was Rust. Three weeks is a large share of a four-month programme. Rust has a reputation for a steep learning curve, and plenty of engineers have excellent careers without writing a line of it. So why make it central?

The short answer: learning Rust properly makes you a better engineer in every language, the industry is moving towards memory-safe systems code, and there is a proven way to teach it that we could build on.

Rust forces understanding

Most languages let you get a surprising amount done without understanding where your data lives, who owns it or what happens when two threads touch it at once. That is a feature, until it isn't. The bugs that cost the most, in production and in interviews, tend to live exactly there.

Rust does not let you skip those questions. The borrow checker makes ownership and lifetimes explicit. The type system makes you say what can be shared between threads. The first weeks are frustrating for most people, and the frustration is the lesson. Engineers who come through it write better Python, Go and TypeScript afterwards, because they have learned to see the things those languages hide.

The borrow checker is a teacher that never gets tired of explaining ownership to you.

The industry has moved

This is no longer a niche bet. Google has written publicly about shifting new Android code to memory-safe languages, Rust among them, and reported in September 2024 that memory-safety issues fell from 76% of Android's vulnerabilities in 2019 to 24% in 2024. The Linux kernel added support for Rust in version 6.1, released at the end of 2022. Major cloud and infrastructure companies now run Rust in performance- and security-critical paths.

AI has added another reason. A lot of the fast, low-level code around machine learning, from tokenisers to data-frame libraries to inference servers, is written in Rust. Engineers who can read and write it have an advantage in exactly the part of the stack that is growing fastest.

Comprehensive Rust, and why we follow it

Google's Android team developed and published a free course called Comprehensive Rust. Its core is four days of Rust Fundamentals: types and values, control flow, references and user-defined types; pattern matching, traits, generics and closures; memory management, smart pointers, borrowing and lifetimes; iterators, modules, testing, error handling and unsafe Rust. On top of that sit deep dives on concurrency, idiomatic Rust, unsafe code, bare-metal work, and using Rust inside Android and Chromium.

We did not want to invent our own sequence when a well-designed one already existed and had been used inside a company that ships a great deal of Rust. So Module 2 of our engineering track follows the Comprehensive Rust structure closely. To be clear, we are not affiliated with Google; the course is openly licensed, and what we add is our own.

What we add

A four-day course is designed for professional engineers who will go back to their teams and keep practising. Our Fellows need the practice built in, so we made three changes.

• Time. We spread the material over three weeks, with the fundamentals in weeks four and five and the concurrency, async and unsafe deep dives in week six.
• Labs. Each segment has graded exercises and a real project: porting the HTTP server Fellows built in Module 1 to async Rust and measuring the difference, writing a safe wrapper over a C library and fuzzing it, and building a concurrent key-value store.
• Continuity. Rust does not disappear after week six. Fellows implement Raft in Rust in Module 3, can use it for serving in the AI module, and every capstone includes at least one Rust service.

Is it too hard for freshers?

It is hard. It is not too hard, provided the foundations come first. That is why Rust sits in weeks four to six, after a module on computer systems where Fellows learn about memory, caches, threads and the cost of a context switch. By the time they meet the borrow checker, they know what it is protecting them from. We also run a borrow-checker clinic in week five, because everyone gets stuck in the same few places.

What it does for your career

Very few job descriptions for new engineers require Rust. That is part of the point. An engineer who can explain ownership, reason about concurrency and point to a Raft implementation and a fuzzed FFI wrapper they wrote stands out in any backend, systems or infrastructure interview, whatever language the company uses. The companies that do use Rust find it hard to hire for, and they notice.

If you want to see the full module, it is on the Software Engineering + AI/ML curriculum page. Cohort 03 starts on 26 October.