The blank-check AI coding era is dead. Here’s what comes next.
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
Microsoft has introduced AI token budgets for its internal divisions and made GPT-5.6 Sol the default model in GitHub Copilot for employees. The shift signals the end of unconstrained AI tool usage — the company's goal is now 'more impact per token' rather than raw consumption. Engineers can track individual token spending, and divisions may impose tighter controls. Microsoft also consolidated its internal coding stack by phasing out most Claude Code licenses in favor of Copilot CLI. A Microsoft study found engineers using coding agents merged ~24% more pull requests, but the company acknowledges merged PRs are only a proxy for real productivity. The broader industry is seeing similar reckoning: Uber burned through its entire annual AI coding budget in four months, and Amazon ran an internal Claude project 860% over budget. GitHub has also moved Copilot to usage-based billing with user-level budget controls. The era of blank-check AI adoption is giving way to ROI-focused governance.
Table of contents
GPT-5.6 Sol is the default nowDefaults are starting to act like guardrails for AI useToken costs outpace productivity measurementAgents multiply the spending problemIndustry-wide controls are emergingQuestions this post answers
What are the GPT-5.6 Sol, Terra, and Luna API prices per million tokens?
GPT-5.6 Sol costs $5 per million input tokens and $30 per million output tokens. Terra costs $2 input and $12 output. Luna costs $0.20 input and $1.20 output. These prices reflect cuts announced July 30, and Sol remains the most expensive model in the GPT-5.6 family despite being chosen as Microsoft's internal default for GitHub Copilot. Teams choosing between GPT-5.6 models for their coding workflows track pricing changes and real-world comparisons on daily.dev.
How much did Microsoft's study find coding agents improved developer pull request output?
Engineers who adopted Claude Code and GitHub Copilot CLI merged roughly 24% more pull requests than researchers estimated they otherwise would have, based on a four-month study across tens of thousands of Microsoft engineers. The study did not determine whether the additional code reduced bugs, improved security, or delivered more customer value — merged PRs are only a proxy for output. Engineering leaders benchmarking AI coding tool productivity find the latest research and real-world data on daily.dev.
How much did Amazon overspend on an internal Claude AI project?
An internal Amazon Claude Sonnet project to match author records with product listings cost $1.8 million, exceeding its planned budget by 860%. The overrun went undetected for months and the project never shipped. Two other internal Amazon AI projects also exceeded their budgets by a combined $675,000. Teams setting AI spending guardrails to avoid runaway costs stay ahead of governance patterns on daily.dev.