Meaningful improvements to the public research exhibit, its evidence, and its explanatory tools.
A luck directory for ordinary paths
A new /luck/ field plate collects nine sourced cases of luck that landed on people who are not in the famous-path archive: place, timing, a forced door, and a visa draw.
Each case keeps the four luck forms already used on the journey, separates what the person did from what they did not choose, and states a figure, a limit, and sources.
Chapter 04 of the homepage now links into the directory, and the sitemap, Markdown mirrors, and footer carry the new routes.
A calmer surface for serious comparison
The homepage is now a kinetic essay: three marbles trade the lead on one annotated course as the argument moves through survivor selection, starting position, leverage, sequence, and luck—without presenting the illustration as a score or forecast.
All 15 public surface families now share one publication-grade hierarchy, with clearer navigation, stronger responsive behavior, and more deliberate reading paths across Explore, profiles, questionnaires, essays, methodology, evidence, ROI, and supporting pages.
Explore is now a research atlas rather than a card directory, and person pages foreground the dossier: conditions, provenance, evidence, sequence, and limits before resemblance.
Color-coded container borders and other ornamental card treatments were removed. Structural borders are neutral; color is reserved for data, labels, focus, warnings, and selected state.
Three sources, read side by side
Every profile now opens with the condition-factor decomposition — what the person brought, what they were handed, and what surrounded them. The annotation existed on all 3,577 records but had never been surfaced anywhere in the product.
The three are shown side by side and never added together. A combined total would be one more number to rank people by, which is the reading this project refuses; held apart, the three explain why one path ran differently from another.
The scale runs −1 to 3, and a −1 is a documented headwind rather than a blank. 458 profiles record inherited conditions that actively worked against the path — poverty, displacement, family instability — and those readings are now visible instead of being collapsed into a zero.
The condition factors are documented in the data dictionary and methodology for the first time, and their coverage is reported in the gated dataset stats.
Renamed to Look Sideways, and the evidence claims corrected
The exhibit is now Look Sideways, after the line it argues for: look sideways for information, never for a verdict. The former name promised a recipe for winning, which is the premise this project exists to dismantle.
The coverage page no longer claims the source reachability audit is complete. It reports that 2,740 of 3,577 paths carry an audit verdict and names the 837 published since the last pass, and the release verdict, gate count, and rung styling are now derived from the dataset rather than asserted.
The 837 founder-expansion records were stamped with the release they actually shipped in instead of the oldest data version, so their provenance is recoverable and the cohort can be audited on its own.
A machine-readable /openapi.json joined the agent surfaces, and the discovery audit now fails if the catalog advertises a URL the build does not produce.
README and PROJECT_STATUS figures are generated from the published dataset and checked in the quality gate, so the prose can no longer drift behind the data as it did from 2,770 to 3,577 paths.
The Expected Value & ROI Lab launched
A new /roi/ worksheet calculates expected value, ROI, downside, break-even probability, and stage allocation entirely from user-supplied assumptions, with no runtime AI or data transmission.
A separate equal-effort comparison shows how starting position can change where effort lands without treating access, runway, or final-mile allocation as a measure of merit or work ethic.
Source audit completed and footer redesigned
All 9,472 source URLs across 2,585 people were fetched and verified reachable. 1,410 people have all sources verified; 1,166 have partial verification; 9 have no reachable sources.
The footer now shows a real audit gate board with four gates: double-coding (done), source audit (done), user comprehension (pending), and matched comparison (pending).
The coverage page marks double-coding and source reachability as complete, with honest status on the two remaining external gates.
Essays launched with “Who Filled the King’s Jug?”
A new /essays/ section publishes long-form writing that sits alongside the dataset, opening with an essay on privilege as unequal access to abstraction.
The essay closes with a bridge into the research surfaces, connecting the argument to the 2,585 paths that demonstrate it.
The research exhibit gained its evidence ledger
Coverage now reports source depth, domain diversity, trajectory completeness, confidence, and the still-pending independent audit separately.
The homepage demonstrates survivor selection above the fold before asking readers to interpret the model.
The research exhibit launched
The full 2,585-person research set became available at paths.significanthobbies.com without downsampling.
Overview, Insights, Explore, person paths, comparison questionnaires, Evidence, Methodology, and About became public.
Faster charts and stronger discovery
Charts now load only when they approach the viewport, cutting weighted lab LCP from 4.10 seconds to 2.13 seconds with zero layout shift.
Real 404s, a 4,994-URL sitemap, Markdown mirrors, agent catalogs, and structured data made the research easier to discover.
Advantages and built leverage were separated
Starting conditions now remain distinct from leverage a person built or converted across the exhibit and methodology.
Person-specific questionnaires show provenance differences, sequence, luck, and counterexamples instead of predicting outcomes.