Hyacehila's Blog

Generative AI expands not only productivity, but also the boundary of individual action and experimentation. As code gets cheaper, engineers need product thinking, distribution, and validation.

On a rainy night, the Wordless Stele at Qianling reveals to each reader the official history they already believe; an indebted stonemason must seal an old crack with his own name.

A series guide to The Epitaph Protocol: how epitaphs, final letters, tombs, and memorial formulas become mechanisms for sealing, verifying, and delaying historical branches.

AI agents matter less as a way to make games automatically and more as a way to connect models, tools, content generation, and feedback loops to production workflows that can be checked, rolled back, and audited.

Science fiction is not a technical report about new technologies. This post starts from world and Chinese science-fiction history to discuss futures, historical branches, and social science as new material.

During Qingming Festival, Lin Qiming’s digital ancestor service is about to expire, and a family must choose whether to keep maintaining the dead or let him rest a second time.

A Zhihu answer offers a useful reminder: in large-model engineering, clever structures do not always beat simple ones that keep GPUs full and follow matrix-multiplication pipelines.

Word Salad Chopper does more than trim repetitive reasoning text. It suggests that hidden states at newline-boundary tokens may offer a cheap window into generation patterns.

Output tokens are expensive mainly because decoding is serial, KV cache consumes memory, and generation occupies scheduler slots. Agent cost optimization requires controlling output budgets.

Model routing breaks every agent call into cost, latency, tool stability, and escalation decisions. The point is not picking the strongest model, but allocating budget.