AI Coding Agents Hit Product-Market Fit — What It Means For Developers

May 2026 marks an inflection point: Anthropic is about to report its first profitable quarter, and OpenAI now allocates over 32% of its workforce to enterprise sales. Simon Willison, a well-known voice in the AI developer community, just published a deep analysis: coding agents have truly found product-market fit.

No longer experimental

Until recently, companies paid a flat per-seat fee for Claude Code or Codex. But as of November 2025 (Anthropic) and April 2026 (OpenAI), both switched to API token-based pricing. In other words: enterprises now pay full API prices for every agent interaction.

This is an aggressive move — and it only works if companies are willing to pay, because agents have become indispensable.

According to data from Simon Willison, a "moderately heavy" user of Claude Code and Codex can burn through roughly $2,180 in API tokens per month (at API prices). At $200/month for the Max + Pro plans, individual developers still come out ahead. But for enterprises signing year-long contracts, the costs are rising fast.

Why this is a big signal

Product-market fit is not "many people are trying it." It means customers are willing to pay and stick around. The numbers back this up:

• Anthropic has 390 open positions; OpenAI has 703
• 32.6% of OpenAI's openings are in enterprise sales and support
• The latest models (GPT-5.5, Opus 4.7) have higher API prices than their predecessors

Notably, prices are not dropping over time — they are rising. GPT-5.5 costs 2x GPT-5.4. Opus 4.7 costs about 1.4x Opus 4.6 (adjusted for the new tokenizer).

What this means for developers

When coding agents hit product-market fit, everything accelerates:

1. Agent skills are now mandatory

This is no longer optional. Companies are spending real money on agents. Developers who use agents effectively will outperform those who don't — by a wide margin.

2. Agent costs become a line item

Like AWS bills or GitHub licenses, API costs for coding agents will appear in monthly team budgets. Understanding how to optimize token usage — prompt caching, choosing the right model for each task — becomes a practical skill.

3. Agents go beyond code

A coding agent can automate anything you can do by typing commands into a computer. As Simon Willison points out: from software engineers to knowledge workers, everyone stands to benefit.

4. Prices will keep rising

When both market leaders raise API prices for their latest models, that's a trend, not an exception. Individual developers should lock in fixed-price plans ($100-$200/month) while they still can.

Bottom line

Coding agents have moved past the "experimental" phase. When enterprises are ready to pay full API rates for thousands of employees using agents daily, the signal is impossible to ignore.

The question is no longer "should I use an AI agent?" — it's "how do I use it effectively?"