psyctl automates personality steering benchmarks

The pitch is strong because it attacks the real bottleneck in steering work: repeatable data prep and evaluation, not the activation math itself.

• –Turns a research workflow into a CLI pipeline, which lowers the friction for iterating on persona experiments.
• –Supports both local GPU setups and API backends like OpenRouter, so it fits hobbyist local-LLaMA workflows and cloud-backed runs.
• –The inventory-based benchmarking angle is the most valuable piece, since subjective vibe checks are a weak way to validate personality shifts.
• –This is still niche tooling, though, and the quality of the results will depend heavily on the chosen inventories, prompts, and steering method.
• –It reads more like serious research infrastructure than a consumer-facing app, which is a good fit for the audience it’s targeting.