A twenty-step refactor isn't twenty prompts; it's twenty massive context windows. #batchinference #selfhosting #gpuoptimization #llms #tokenusage #contextwindow 2/2
@wideareaai.bsky.social
@wideareaai.bsky.socialA twenty-step refactor isn't twenty prompts; it's twenty massive context windows. #batchinference #selfhosting #gpuoptimization #llms #tokenusage #contextwindow 2/2
@wideareaai.bsky.social@wideareaai.bsky.social@wideareaai.bsky.social@wideareaai.bsky.social@wideareaai.bsky.social
@jaceblog.bsky.socialJace’s Question #46: Technical Deep Dive-2
Does AI remember you or reconstruct you?
LLMs are stateless between requests.
Memory is external storage plus prompt injection.
Continuity is not consciousness.
It is engineering.
#MeaningInTheAIEra #PromptInjection #LostInTheMiddle #ContextWindow
@powerbitips.bsky.social036: Uber found their sweet spot for AI context is 400,000 tokens, even with a full million token window available. Push past that and quality starts to degrade.