Introducing Thyself

A note to the small group of people who joined the private test. Thank you for being here early.

On curation

Curation, in its original sense, is an act of care. The Latin root is curare: to tend, to look after, to take responsibility for. Walter Benjamin, writing about his own library in 1931, understood the collector's deepest impulse as a struggle against Zerstreuung, dispersion.^[1] Every object in the world exists in a state of scatter, available to anyone, belonging to no particular order. The collector's work is to interrupt that scatter by placing things in relation to one another — to give them a context that is also a commitment. The collection becomes an argument about what matters and why.

The word migrated to the internet and shed most of that weight. It came to mean filtered content: a feed narrowed by an algorithm toward some optimization target. The human sense of it, where someone takes genuine responsibility for what gets through and why, largely disappeared.

Thyself is an attempt to recover that sense. The premise is that a person's reading life is something worth tending, that the articles you spend time with, the ideas you return to, and the topics you allow to develop over months constitute something real about who you are and who you are becoming. A curator, in the way we are using the word, is someone who takes that seriously.

The problem we kept returning to

Herbert Simon, writing in 1971, observed that "a wealth of information creates a poverty of attention."^[2] He was describing a tradeoff that has only sharpened since: the more content there is, the more costly it becomes to find what is worth your time. The systems built to address this have tended to resolve the tradeoff in a particular direction, optimizing for engagement rather than meaning.

The psychologist Mihaly Csikszentmihalyi spent decades studying what makes human experience feel worthwhile. His research consistently pointed toward activities with clear boundaries, immediate feedback, and a balance between challenge and skill, the conditions he called flow.^[3] Infinite scroll satisfies none of these. It has no natural endpoint, offers no feedback about accumulation, and calibrates difficulty only toward the path of least resistance.

We found ourselves asking what a reading environment would look like if it were designed around the curator's sensibility: bounded, accountable, and oriented toward accumulation.

A bounded daily panel

Every day the Salon surfaces eight items, drawn from a catalog we tend: essays, long reads, podcasts, visual pieces. The panel is finite. Once you have seen it, there is nothing more to load. The constraint is structural and intentional.

Barry Schwartz's research on the paradox of choice suggests that reducing optionality often improves both the quality of decisions and the satisfaction people feel afterward.^[4] Eight items is enough to offer genuine variety across media types and subject areas. It is few enough that you can hold the full panel in working memory and make a considered choice about where to spend an hour.

The visual design reflects this. Cards sit idle showing only a single typographic keyword, the semantic core of the piece. Hover to see the title, abstract, source, and estimated reading time. The interface withholds detail until you ask for it, which tends to produce a different quality of attention than a feed that front-loads everything simultaneously.

How personalization works here

The recommendation pipeline scores catalog items against a vector representation of your interests, built from your intake answers and whatever you have explicitly marked as read or saved. Relevance to your long-term interest profile accounts for 60% of the score; recency for 30%; and a small day-of-week context signal for the remaining 10%.

Three of the eight slots are reserved for exploratory picks: content adjacent to your established interests but outside your current categories. This is the part of the design we feel most strongly about.

There is a well-documented phenomenon in recommendation systems sometimes called the filter bubble, the tendency of relevance-ranked systems to progressively narrow the range of content a person sees.^[5] Eli Pariser described it as a kind of epistemic enclosure: the system learns what you engage with and serves more of it, which shapes what you engage with next, in a loop that is self-reinforcing and largely invisible. The three exploratory slots are a structural interrupt in that loop. A good curator widens taste as well as confirms it. The system is built to do the same.

The persona itself, the model of your interests the system works from, is transparent and editable. Your Persona page shows you the current interest weights, how they have shifted over time, and how the media types you have consumed compare to what the system has been recommending. If the model has drifted somewhere you do not recognize, you can correct it. Changes take effect the following day.

On privacy

We use only explicit signals: articles you mark as read, save for later, or annotate with a reflection. Hover time, scroll depth, and click-through patterns are not collected. This is a deliberate constraint on what the system is allowed to know about you.

The philosopher Helen Nissenbaum has argued that privacy is best understood as contextual integrity — information flows appropriately when it matches the norms of the context in which it was shared.^[6] When you tell us you are interested in ecology and architecture, you are sharing something in the context of building a reading profile. That information surfaces relevant articles. It is not used to infer mood, model behavior, or build a profile for any other purpose.

What this period is for

The catalog is seeded and the recommendation engine is running. What we are still learning is whether the experience feels right: whether the panel size is calibrated correctly, whether the exploratory slots land well, whether the Persona page gives you a useful picture of how the system sees you.

If something feels off, our Discord is the right place to tell us. We read everything.