The Industrialization of IT

a year ago

Entry-level software engineers in San Francisco make about $190,000 a year, with a total employment cost of around $250,000 when including benefits.
Gemini 2.5 Pro, the best AI model for writing code, costs $1.25 per million input tokens and $10 per million output tokens, making it significantly cheaper than hiring an engineer for high-volume tasks.
A single engineer is likely more productive than an AI model today, but AI models are improving rapidly and will become more cost-effective and capable in the future.
AI models can generate vast amounts of code quickly, but they may produce errors, lack strategic direction, and struggle with security issues.
Future advancements in AI could enable millions of prompts per second, far outpacing human productivity, especially if codebases are optimized for machine parallelization.
Software engineering may become industrialized, with AI agents replacing human engineers in many tasks, leading to cheaper and faster production.
The environmental costs of running large-scale AI models are a growing concern and should be factored into discussions about AI adoption.

Hasty Briefsbeta