How to Get a Promotion and Burn Your Company's AI Budget in Five Minutes

Have you noticed how company leadership has gone crazy over neural network adoption metrics? In some departments, managers have started evaluating engineers not by closed tickets or service stability, but by the number of tokens spent in Claude Code or GitHub Copilot. Absurd? Absolutely. But where there's a metrics system, there's a way to abuse it.

Developer dtnewman released the repository burn- baby- burn. This is a comedic CLI utility created with one purpose: deliberately burning LLM tokens at industrial scale.

Why Game Token Counts

The project author lays out the "business justification" for such a tool right in the README:

• Show leadership the insane growth of your productivity.
• Help a startup look as cutting-edge as possible to investors.
• Climb to the top spots in internal Claude Code usage rankings.
• Secure an increased neural network budget for the next year.

Funny enough, in the real world, investors really do sometimes look at API consumption graphs as a sign of development activity. The utility turns this paradox into a corporate survival tool.

How burn- baby- burn Works

Technically, the project is a concise bash script. It doesn't need superuser rights or complex build systems.

Installation takes literally a second:

curl -fsSL https://raw.githubusercontent.com/dtnewman/burn-baby-burn/main/install.sh | bash

The script places the file burn in the ~/.local/bin folder. The only dependencies needed are the jq utility and an authorized CLI client for Claude Code or OpenAI Codex in your environment.

Running it looks extremely simple:

burn 10000                            # минимальный прогрев
burn 50000 --model haiku              # быстро и недорого
burn 100000 --model sonnet            # запустить и уйти пить кофе
burn 50000 --backend codex            # сжигаем токены OpenAI

Under the hood, the utility sends meaningless requests and forces the model to generate responses until the specified token limit is burned through.

Charts and the Enterprise Version

The utility's interface outputs visual charts of budget burning in real time to the console. The taller the bars, the more "productivity" your monitoring system sees.

In the README, the author jokes that a paid Enterprise version is already in the works with the following features:

• Fourfold increase in token burn rate.
• SSO and SAML authentication support.
• Role-based access control for budget burning (RBAC).

The code is distributed under the MIT license, although the author ironically notes that the project was entirely written by a neural network, so ownership is an open question.

The Takeaway

The project has already gathered over 800 stars on GitHub, and the reason for this popularity is clear. It's excellent satire about the modern IT industry, where KPIs are often disconnected from reality.

Running the script in real work on a constant basis is, of course, not advisable. The IT department will quickly notice anomalies in the bills. But as a joke to share with colleagues or a demonstration of metric absurdity, burn- baby- burn works great.