Aug 27, 2025

Moore’s Law of LLMs: AI is Moving Faster Than You Think

Meet the expert

Serge Haziyev

Advanced Technologies CTO, SoftServe

What’s trending

As GPT-5 rolls out, the bigger story isn’t just what it can do today. It’s what the “Moore’s Law” suggests AI will be capable of by 2030.

The Moore’s Law of LLMs: AI capabilities — especially large language models (LLMs) — are improving at an exponential rate, with the duration of reliably executed tasks (that would take for humans) doubling every seven months.

This trend is tracked by METR.org and suggests that by 2030, AI agents will be able to autonomously complete week-long software development tasks.

Market disruption or hype?

Disruption: The exponential trajectory is backed by empirical benchmarks from the past six years. If GPT-5 today can execute a task with 80% reliability that would take a human expert 26 minutes, then in four years, AI would perform a task equivalent to 50 hours of human work.

Why it matters: We can forecast when AI will develop new capabilities. If the trend of the past six years continues, frontier AI systems will be capable of autonomously carrying out week-long and eventually month-long projects.

What it means for your business

Sprint-level autonomy: By 2030, AI agents will reliably deliver end-to-end software features that meet defined acceptance criteria.

Human roles evolve: Team members will focus more on software architecture, agent orchestration, and domain expertise.

Project-level autonomy: Still out of reach — humans will coordinate dependencies, manage integration, and adapt to the real-world feedback.

Opportunities and hurdles

Opportunities

Much faster software project delivery
Major cost savings in reducing technical debt and modernizing software
More time for humans to focus on strategy and creativity

Hurdles

Making sure AI is used safely and correctly
Training teams to work with AI agents
Managing complex projects that AI can’t handle yet

SoftServe’s approach

SoftServe already uses AI agents for real-world tasks, such as software development, modernization, and testing. We also collaborate with NVIDIA, AWS, Google Cloud, and Microsoft to integrate agents into enterprise workflows for greater operational efficiency. We help clients across industries:

Transform businesses with agentic AI
Build products with state-of-the-art AI capabilities
Accelerate software delivery with agentic engineering

Start a conversation with us