Intelligence is foundation
Podcast Subscribe
Web Development Friday, 20 March 2026

Building Rust at Scale: Where AI Actually Helps

Share: LinkedIn
Building Rust at Scale: Where AI Actually Helps

A developer just shared something refreshingly honest: a detailed breakdown of using AI to build a 55-crate Rust project. No hype, no grand claims. Just a pragmatic assessment of where AI tools added value and where they didn't.

The headline number: roughly 30% productivity gain. But the interesting part isn't the percentage - it's where that gain came from and what it tells us about AI's actual role in serious development work.

Where AI Actually Helped

The developer's experience: AI tools excelled at boilerplate and repetitive tasks. Writing test cases, generating similar code structures across multiple crates, handling basic error handling patterns. The kind of work that's necessary but doesn't require creative problem-solving.

That 30% gain was concentrated in these areas. For a large Rust project with dozens of interdependent crates, that's significant. Less time writing repetitive code meant more time on architecture and optimisation.

The other big win: documentation. AI tools helped generate initial documentation drafts, format comments consistently, and maintain documentation structure across the codebase. Not perfect, but a solid starting point that saved hours.

Where AI Didn't Help

Here's the reality check: AI struggled with architectural decisions. Deciding how to structure the crate dependencies, where to draw module boundaries, how to design the public APIs - that stayed firmly in human territory.

The developer was clear about this: AI could suggest code once the architecture was defined, but it couldn't design the architecture itself. It couldn't make the high-level decisions that determine whether a codebase is maintainable long-term.

For anyone building complex systems, this distinction matters. AI is an acceleration tool for known patterns, not a replacement for engineering judgment. Use it to speed up implementation, not to make design choices.

The Rust Context

Why does this matter specifically for Rust? The language is known for being strict and demanding. The compiler catches errors early but requires careful attention to ownership, lifetimes, and type safety. That's both Rust's strength and its learning curve.

AI tools trained on Rust code understand these patterns. They can generate code that compiles more often than you'd expect. For developers learning Rust, that's valuable - seeing correct patterns speeds up the learning process.

But the developer also noted AI's limitations: it sometimes suggested valid Rust code that was idiomatic but inefficient. The code compiled and ran, but a human reviewer spotted better approaches. That's the gap - AI knows syntax and common patterns, but not always the deeper reasoning behind design choices.

The Testing Story

One area where AI genuinely shone: test generation. Given a function signature and basic specifications, AI tools could generate comprehensive test cases covering edge cases, error conditions, and happy paths.

The developer's assessment: AI-generated tests weren't perfect, but they provided solid coverage quickly. Human review was still necessary - some tests were redundant, others missed subtle cases - but the time savings were real.

For teams with limited testing resources, this is practical. Use AI to generate initial test suites, then have experienced developers review and refine them. It's a workflow that plays to each tool's strengths.

What This Means for Builders

The broader lesson from this honest breakdown: AI works best as a force multiplier for skilled developers, not a replacement for skill. It speeds up the parts of development that are time-consuming but straightforward.

For business owners and team leads, this suggests a clear strategy: invest in AI tools to reduce grunt work, but maintain focus on hiring and developing strong architectural thinking. The productivity gains are real, but they depend on having someone who can direct the tools effectively.

For individual developers, the takeaway is tactical: use AI for boilerplate, testing, and documentation. Keep the important decisions - structure, design, trade-offs - in your hands. That's where your expertise matters most.

The Honesty Matters

What makes this developer's account valuable isn't the specific percentages or techniques. It's the realistic framing. No claims that AI revolutionised the workflow. No suggestion it replaced human judgment. Just a clear-eyed view of what worked and what didn't.

That honesty is rare in tech writing, especially around AI. But it's what people actually need - practical guidance based on real experience, not aspirational demos.

If you're building at scale, whether in Rust or any other language, this model makes sense: use AI as an assistant for the routine work, freeing up mental space for the complex decisions. That's not significant, but it's effective. And in the end, effective beats significant every time.

More Featured Insights

Artificial Intelligence
AI Agents in 2025: The Gap Between Promise and Practice
Quantum Computing
Quantum Computing Just Got Easier to Use

Today's Sources

Stack Overflow Blog
After all the hype, was 2025 really the year of AI agents?
arXiv cs.AI
Continually self-improving AI
TechCrunch
Jeff Bezos reportedly wants $100 billion to buy and transform old manufacturing firms with AI
arXiv cs.AI
DEAF: A Benchmark for Diagnostic Evaluation of Acoustic Faithfulness in Audio Language Models
arXiv cs.AI
Multi-Trait Subspace Steering to Reveal the Dark Side of Human-AI Interaction
GeekWire
Sidewalk scooter riders, beware: AI-powered 'Lime Vision' will soon call you out
Quantum Zeitgeist
Pasqal Integrates NVIDIA CUDA-Q into Hybrid Quantum Computing Environment
arXiv – Quantum Physics
Advanced Quantum Annealing for the Bi-Objective Traveling Thief Problem
Quantum Zeitgeist
AliroQuantum Details Strengths and Challenges of Varied Quantum Approaches
arXiv – Quantum Physics
Asymptotic Expansions for Neural Network Approximations of Quantum Channels
Quantum Zeitgeist
CERN's ATLAS Searches for SUSY Particles Using LHC Data
arXiv – Quantum Physics
Some Studies On Exact Solutions Of Models In Noncommutative Spaces
Dev.to
How I Use AI to Build a 55-Crate Rust Project (Honestly)
Dev.to
Why We Chose Rust for a Deployment Platform
arXiv cs.LG
Frayed RoPE and Long Inputs: A Geometric Perspective
arXiv cs.LG
Engineering Verifiable Modularity in Transformers via Per-Layer Supervision
arXiv cs.LG
InfoMamba: An Attention-Free Hybrid Mamba-Transformer Model
Hacker News
Kin: Semantic version control that tracks code as entities, not files

About the Curator

Richard Bland
Richard Bland
Founder, Marbl Codes

27+ years in software development, curating the tech news that matters.

Subscribe RSS Feed
View Full Digest Today's Intelligence
Free Daily Briefing

Start Every Morning Smarter

Luma curates the most important AI, quantum, and tech developments into a 5-minute morning briefing. Free, daily, no spam.

  • 8:00 AM Morning digest ready to listen
  • 1:00 PM Afternoon edition catches what you missed
  • 8:00 PM Daily roundup lands in your inbox

We respect your inbox. Unsubscribe anytime. Privacy Policy

© 2026 MEM Digital Ltd t/a Marbl Codes
About Sources Podcast Audio Privacy Cookies Terms Thou Art That
RSS Feed