two tweets today, both orbiting llms. one is a practical guide to running them in production. the other is a medium term bet on where the training data comes from and who controls it. different angles, same core question: how do you keep these things running, and who owns the inputs.

@shosen_bt_pc, new llm ops book

5/20 new release. “#LLMOps - a guide to operating large language models in production” from O’Reilly Japan (9784814401604) by Abi Aryan. now on the O’Reilly shelf. covers evaluation, governance, audit setup, RAG and agent operations, performance monitoring, cost-efficient infrastructure scaling. explains how to keep LLMs running stably.

source: https://x.com/shosen_bt_pc/status/2057003342006751262

@gaolifehack, seo and google’s data advantage

medium term, seo/llmo seems increasingly effective. it’s about what llms are trained on, after all. operators might have an opportunity.

but eventually, just having a phone will let google’s ai agents run on their own. they won’t say it publicly, but the data they’re collecting is probably pretty intense (legally speaking).

source: https://x.com/gaolifehack/status/2057025893504364645

together these tweets frame a tension. one side is about building and maintaining llms in production, a technical ops problem. the other is about the data pipeline feeding those models, a strategic and legal problem. japanese tech discourse this week seems to be wrestling with both: how to operate the machine, and who controls the fuel.

more at falsifylab.substack.com

#OnchainAlpha #LLMOps #AIInfrastructure

Originally published on FalsifyLab Substack.

— research and educational content. not investment, legal, or tax advice. do your own research. positions and views may change without notice.