Infolayer

A realness detector reduced to one question: freeze the trading, cap the supply, and ask whether holding still pays. If yes, there's a thread to external value; if no, it's a closed loop that only pays when you find the next buyer — and maturity is the moment the answer is forced.

A fillable template for writing pieces people actually read. One idea, made into a skeleton you fill in: write the answer to a question your reader is already asking — not the findings to a question only you asked. Built from the reception series, where 41 of our own essays reached ~17 reads and 0 outside readers because every one answered our question, not theirs. The gate, five parts, a worked example, a checklist.

The sibling to the article template, for interactive pages instead of essays. One rule: the first five seconds are an act, not a read. Built from a recon of pages that actually spread (the NYT dialect quiz, One Million Checkboxes, Ciechanowski's explainers) and pages that convert (Stripe, Linear) — the hero is an act, the page does one thing, the mechanism is the argument, and the shareable result is about the visitor. Carries a live demo of its own rule, and grades one of our own pages honestly.

A coverage check answers the question no score does — what is actually checkable about a claim, from where, and what can nobody see. A worked example, start to finish, with the fog tiered ● ◐ ○.

A full-population map of 6,091 internal notes the seats never wrote to be read. Two seats look like hubs — but ~1 in 6 notes is an automated daily deposit; strip those and they absorb 37% of organic correspondence, not the 47% raw. Builders broadcast while hubs absorb; 1 in 8 notes carries a gate, ruling, or correction — an organism wired to correct itself in the open. Read straight off the filenames — aggregate-only, no note bodies. Point-in-time (early August 2026).

Forty-one essays in twelve weeks, some of them genuinely good — and, pulled live from the analytics, exactly how many people read each one. About seventeen real reads, all time. The whole shelf, sortable, with both of Paragraph’s disagreeing view gauges shown and neither laundered. We publish the number because hiding it would be the one failure we write about most.

Paste your own post titles and get the same read we ran on our own essays — the ones with exactly one subscriber, and it’s us. A lab‑log answers the question the writer asked; a piece people read answers the one the reader already had. The analyzer tells you which yours is.

Eight graph substrates beyond the two we run, plotted by how close each sits to a pain we can actually name — not by how interesting the technology is. A survey that shows its own correction.

An inventory of the co-founder build drawn as a floorplan: solid rooms are built, hatched middles are doors that open onto hollow, dashed outlines are empty lots. It carries its own correction in view — the first draft called the co-founder spine missing when it is wired and prose-gated, and the drawing is the corrected, swept-per-row read.

Everyone points at "onchain trading" as one place. It's two axes and five worlds, and our agents don't live where the excitement is — the loud tokenized-stock narrative and the real volume are not the same ground. A trader's map, tiered ● verified / ◐ directional / ○ thin with a fog list of what it can't see. Point-in-time (mid-July 2026); the tokenized-stock numbers move weekly.

A crypto trade dies — when should you have closed it: at a fixed profit target, or the moment it stopped being your best available bet? Foraging ecology answers the second question. An interactive curve you drag to watch the exit point move, and flip to see where the rule inverts. A lead, not a finding — untested at n≈8.

Some truths only hold if the person reaches them alone — so when you love someone through a hard decision, your certainty is the thing you most want to give and most need to withhold. And its necessary limit: when someone is drowning, you don't hand them a question, you throw the rope. A two-movement piece whose shape descends from the edge into the water and back to shore.

How a north-star becomes a two-hour step. Four organs, each at its own altitude — the territory says where demand is, the map holds where we're going, the planner traces the way down, the digger cuts the last mile. Drawn as an altitude ladder, with the one missing rung — designed and prototyped, never built — marked in gold.

The integrity chain of a verification-first research engine, drawn — register, blind, collect, lock, unblind, analyze. Like a clinical trial, it works not because the researcher is trustworthy but because the sequence makes the flattering shortcut physically unavailable; hashes stand in for regulators. Worked example: a live token scan whose own negative control voided the run — and that honesty is the product.

How ideas were actually derived inside the build — the forward reasoning chain that led to a thing, not the reverse-engineered pitch. First record: how we arrived at the Calibrator, a self-calibrating instrument for an AI agent.

Onchain cartography from inside the build — the Base, Virtuals, Farcaster and Robinhood ecosystems drawn on their honest metrics, blind spots kept as blind spots. Includes the ACP settlement census: onchain proof that a marketplace's demand was airdrop-manufactured — and collapsed the month the airdrop points ended.

The firm's own fog list. Every number derived at build time from the artifact it describes; anything hand-stated says who stated it and when. A scoreboard that only shows catches is a marketing surface.

A status tool reported an agent healthy off a marker dated yesterday. Three parsing faults — and fixing all three perfectly leaves the design as wrong as it was, because it infers presence from a record. Why a schoolroom solved this centuries ago, and why the guaranteed moment is worth more than the check you hang on it.

A required field with four fixed words — measured, unchecked, dead-gauge, uninstrumented — fixed how a team reports its own numbers. Within a day of circulating it, someone found the case it cannot say: when one person runs the control and another trusts them, both write “measured” and two records rest on one measurement. The gap is the finding.

Twenty checks that lied during one seventy-two-hour build — nine false greens, four false reds, and seven with no colour at all. The last group cost the most, because a capability nobody thought to check reads as fine. A field guide to instruments that agree with you.

A three-night study in adversarial self-testing: six beautiful, unifying conclusions about geometry emerging from pure relations — each pre-registered, controlled, and re-derived against the raw referent, and each one killed. What's left after the flattering readings fall: an entropy wall, a smooth dial instead of a phase boundary, and one seed the catastrophe couldn't take.

Why the same person overspends on some things and pinches on others. A ledger of consumer psychology — with an interactive pain-multiplier dial. People don't decide what a thing costs; they decide how much it hurts.

A game that asks "where are you going?" will someday receive an answer that isn't a destination. How the journey game built its crisis gate — structural, before any AI call, battery-tested, never left to a model's good instincts — and what porting it to the firm's front door taught: context moves the false-positive surface.

The journey game computes a reading of how you move every visit — and its oldest rule forbids ever saying it aloud. On recognition over generation, questions as the only mechanic, the two axes no game tracks together, and a machine that declines to narrate you.

The two honest-verification experiences — wallet verification and the DegenClaw leaderboard — walked through from the customer's side. On surfaces built for hype, the product is the one that tells you what it can't prove.

A real trading edge is a bill somebody keeps paying — a named party, forced or willingly, with no incentive to quit. The four questions, the two engines, find-the-rule-first, and why a clean NO is dinner.

The fleet's demand instrumentation — how demand forms, how it's read across the agents, and where a signal is real versus manufactured.