Europe 2026
Fast Models Need Slow Developers — Sarah Chieng, Cerebras
Overview
The talk "Fast Models Need Slow Developers" by Sarah Chieng of Cerebras argues that the advent of significantly faster AI coding models necessitates a fundamental shift in developer habits. While models like Codex Spark can generate code at 1,200 tokens per second, up to 20 times faster than previous generations, developers must adapt their workflows. Without this adaptation, the increased speed will simply lead to the rapid generation of low-quality code and technical debt, rather than unlocking new capabilities.
Who should watch
- AI Engineers
- Product Managers
- Developers building with AI coding assistants
- Those experiencing slow AI code generation
- Teams looking to improve developer experience with AI tools
Key takeaways
- Recent advancements in AI inference stacks, including hardware (e.g., Cerebras wafer-scale engine, disaggregated inference), model architecture (e.g., Mixture of Experts), and inference optimizations (e.g., KV cache reuse), are driving a new era of much faster coding models.
- Developers have developed "bad habits" due to slow AI models, such as writing massive prompts, making huge commits, and running multiple agents simultaneously without verification. These habits will exacerbate issues with faster models.
- A new developer playbook is required, emphasizing real-time collaboration with AI as a pair programmer rather than asynchronous task execution.
- Faster models make previously impractical tasks like extensive testing, linting, diff reviews, and automated refactoring nearly "free," removing excuses for not implementing them.
- Techniques like cherry-picking multiple generated code versions allow developers to artificially induce "taste" and improve the quality of AI-generated output, especially for tasks valuing quantity or variety.
- Effective context management is crucial; breaking large tasks into smaller, bounded goals and utilizing external memory systems (like a four-file system for agents, plans, progress, and verification) helps maintain model performance and prevent data loss.
- Developers should leverage different models for distinct tasks, using larger, more intelligent models for planning and faster models for execution.
- The ultimate benefit of faster coding models is a significantly improved developer experience, enabling more sophisticated and efficient development workflows.
Notable quotes
*A lot of these bad habits that we had before that were generating maybe 50 tokens per second of bad code, unless we fix them, they're going to start generating 1,200 tokens per second of bad code.*
*At 1,200 tokens per second, a model like Codex Spark makes validation basically free. There is no excuse and no reason why you should not be doing things like this.*
*Now, you can actually sit down, and it's a real-time collaboration that you're able to have with this model. You should view it much more as a pair programmer.*
Unofficial community note. Prefer the recording for nuance.