Engineering agents instead of developers: six months of a coding revolution
The author of OpenClaw noted that in six months, agents became mainstream in software development. In May, it was a wow effect; by December, a routine workflow.
Six months ago, Petr Steinberger, creator of OpenClaw (later acquired by OpenAI), published an article about inference-speed development — when an AI agent completes a task in 5 seconds of inference instead of an hour of manual coding. Back then it seemed like a revolution and just early impressions. Now this revolution has become a regular workday in the industry.
From Experiments to Standard
In May 2025, inference speed was an interesting idea. Agents often made mistakes, required custom prompts, worked only on simple tasks. Developers were skeptical: "This isn't real development, it's a toy." By December, the picture had completely changed. Agents learned to work with real projects with thousands of files, complex integrations, legacy code. Accuracy grew from 70% to 90%+. Most importantly, iteration time dropped from hours to minutes. The agent identifies an error, redoes it, puts the result back.
How It Works in Practice
Steinberger discovered a simple truth: delivery speed = inference speed. If the model runs inference in 5 seconds, a feature is delivered in 5 minutes (minus validation). In comparison, this used to take 1-3 hours. But speed isn't the main thing. The main thing is the shift in the developer's role. You no longer write code. You formulate a task for the agent, check the result, approve it or send it back for revision. It's like hiring a junior developer who never gets tired and executes in microseconds.
Methods of Working with Agents
Over six months, practices emerged that Steinberger described and are now standard:
- Minimize clarifications — prepare all context before the request
- Use file-graph and type-hints as training examples for the agent
- Give the agent access to commit history (examples of successful code)
- Don't fix the agent manually — give it an error, and it will redo it itself
- Focus on high-level tasks, not syntax
Engineers now think at the level of features, not lines of code.
The agent handles everything else.
What This Means for the Industry
Development has stopped being about writing code. It's now about task formulation and validation. Those who understood this shift work 3-5x more productively. Those who still write code themselves are gradually losing competitiveness. Steinberger wrote about first impressions, but it turned out — this isn't impressions at all, it's the new normal.