Promptimizer2.0

Optimize your prompts
with a visual decision tree

Promptimizer explores mutations of your prompt — shortening, clarifying, adding examples, tightening constraints — and scores each candidate against your rubric to find the best version automatically.

Beam search treeGPT-4o scoringToken penaltyLive visualizationShareable results

Never stored — lives only in memory for this session.

Each box is a separate test input. Add more to enable train/test splitting.

Uses GPT-4o to convert your free-text rubric into structured scoring criteria.

Run Configuration

3

Tree depth. Deeper = more mutations, higher cost.

20

Total nodes to explore before stopping.

0 — no penalty1 — heavy penalty

How much to penalize prompts that are longer than the parent.

25%
10% — fewest test calls50% — most coverage

After the optimization run, the top-scoring nodes are evaluated on held-out test inputs to check for overfitting. Only applies when 2+ eval inputs are provided.

0.00
0 — random 60%1 — full UCB bandit

UCB bandit tracks which mutation types consistently beat their parent and allocates more budget to them. Mirrors multi-armed bandit explore/exploit tradeoff.

0.00
0 — greedy top-K1 — high initial temperature

Adds decaying noise to beam selection so lower-scoring nodes can survive early generations — helps escape local optima. Temperature cools 35% per generation.

0.10
0 — never stop early0.5 — stop quickly

Stop the run if the best score improves by less than this between depths. Lower values let the run go deeper before giving up. Default 0.1.