AI Blocks › Research a topic properly
Hypothesis tree optimizer
Give it a result you want to improve. Get ranked hypotheses tested one by one, with a record of what worked.
Hypothesis tree optimizer is a free AI skill for the job "Research a topic properly". Give the instructions to an AI assistant that supports the required files and tools, then review its output.
Free · No account · MIT licence · Made by K-Dense · Level: expert · Thinking: high
Made by K-Dense. See the original on GitHub; we copied it as it is, with its licence.
What it does
A steady way to improve a piece of work by testing many ideas and remembering every result. Give it something you can change, a goal and a way to score it, and it runs experiments one after another while keeping a ranked tree of ideas. It first pins down four things: the starting material, the goal, a fast scoring command for the search, and a separate held-out test used only before accepting a change. It writes each idea as a node, tests it in a fresh copy so files stay clean, records the evidence and prunes dead ends. A bundled Python tool does the bookkeeping. You need Python 3.10 and Git. It suits model training, agent tuning and prompt improvement.
You give it
- The project or file to improve, under Git
- The goal and which direction is better
- A command that scores a candidate
- A separate held-out test set (optional)
You get
- A tree of ranked hypotheses
- Results of each experiment with scores
- A log of what worked and what failed
- A tested version of your project
Use it when
- Your agent or prompt passes only some tests and you want it to improve step by step.
- You are tuning a training recipe and have lost track of what you tried.
- You have a script that scores results and want many experiments run for you.
- You need an honest record of what worked and what failed.
How to use it
- Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
- Ask for the job in your own words. Your AI opens the skill when it fits.
- It works best when your AI can run code and keep files between steps.