# Garry Tan: AI, Agency, And Desired Futures Status: deepened milieu note ## Source - Date captured: 2026-06-19 - Source published date: 2026-06-18 - Source type: `X screenshot` - Source title: `AI can be a liberator if you choose agency` - Source URL: - Local source file: `external_material/archive/processed/https___x.com_garrytan_status_2067606805459714229_s=46&t=DEtAIpP3vBwwioECC_VTAw.docx` ## People / Organizations - Garry Tan - X / Twitter - AI / agency discourse ## Neutral Summary The screenshot argues that AI affects people differently depending on whether they have agency and a clear sense of the future they want. Without that orientation, AI can become another control mechanism. With agency, it becomes an amplifying tool that helps a person do things they could not do alone. ## Theme Theory Relation This is a compact outside expression of a point that matters deeply for Theme Theory and What To Build: ```text AI is not direction by itself. AI amplifies direction once a desired state is clear enough. ``` That maps directly to the [object of interest](../../core/object-of-interest.md). In Theme Theory, the object of interest is not merely an inspirational phrase. It is the desired real-life state that gives action, creative, tools, and support a direction. The tweet is useful because it states the negative case. If a person has no clear future they are trying to realize, AI does not automatically give them one. The model can supply output, options, automation, and persuasion, but those may become externally guided rather than agentic. The positive case is close to the WTB thread in the corpus. If a person or organization has a clear desired future, AI can multiply their ability to specify, build, coordinate, and iterate toward it. The scarce layer becomes not raw generation, but judgment about what state is worth moving toward. ## Deep Corpus Comparison This source is compact but central to the WTB side. The tweet's deepest alignment with Theme Theory is: ```text AI amplifies agency only after direction exists. ``` The object of interest is one way to make direction explicit. It does not just name a goal. It names a desired real-life state around which actions, creative, tools, and support can coordinate. This note also helps distinguish agency from automation. A person can use AI to generate outputs while becoming less agentic if the direction is supplied by platforms, feeds, models, or other people's defaults. The TT version of agency requires the person or organization to know what state they are trying to make more real. That applies to this project itself. Codex is useful here because the user has a developing desired future for the artifact. Without that, the same agentic capacity would likely produce generic organization rather than faithful structure. ## Core Links - [Object Of Interest](../../core/object-of-interest.md) - [Creators, Builders, And Audience](../../core/creators-builders-and-audience.md) - [Corpus Digestion Pass 1](../../project/corpus-digestion-pass-1.md) ## Candidate Concepts / Edges - clear desired future -> object of interest - agency -> movement toward desired state - AI -> capability multiplier, not direction source - lack of orientation -> susceptibility to control - WTB -> specification before execution ## Promotion Judgment - Promote to core? `maybe` - Reason: strong short example for the WTB claim that AI makes the object, direction, and specification more important, not less important.