Blog

Leaderboards are useful signals, but they rarely match your real prompts, risk tolerance, budget, or latency needs. Build a small personal eval set so model choice becomes evidence, not vibes.

A practical guide to local multimodal AI workflows: CLIP embeddings, FFmpeg-style media processing, private notes search, Apple Silicon and mobile inference, plus when local AI is worth the tradeoff.

A practical guide to evaluating 3D-aware generative fill, text-to-video, sprite generation, and conversational 3D editing when the goal is production work rather than a viral demo.

A practical LLM evals workflow for product teams: build golden datasets, compare prompts and models, add regression gates to CI/CD, use human review loops, and know when open-ended game-world evals are worth the extra effort.

A practical guide to voice-to-workflow AI for knowledge workers, founders, and ops teams: capture messy thoughts, triage meetings and email, extract tasks, hand work to calendars and project tools, and build a habit that respects privacy.