Viktor in London

My notes

Voice memos recorded while walking, transcribed, geocoded and scored from −5 (not for me) to +5 (home) — shown as colored dots. Entirely subjective, deliberately so.

Heat map (liveability)

A computed 0–1 score blending how reachable four things are on foot, with distance decay and diminishing returns (the 2nd nearby café matters far more than the 12th):

score = 0.35·connectivity + 0.30·coffee + 0.20·food + 0.15·good life

Hub stations count triple; ★★★ cafes and 9.0+ restaurants double; Aesop is excluded from good life while full supermarkets weigh extra. Shown as five grades (0.2-wide score bands): grades 1–3 are pale washes, grades 4–5 stand out in deep teal. Full methodology lives in the project's docs/livability.md. Note it measures amenity access only — Soho maxes out precisely because it's Soho; peace and safety are what the walk notes and crime layer are for.

Coffee

Two sources. Independent cafes come from propercoffee.app — only their ★★★ Purist and ★★ Specialty tiers (the "Everyday" tier is skipped). Chain locations (Kiss the Hippo, WatchHouse, Rosslyn, Hagen, Origin, Ozone, Qima) come from each chain's official location pages. Bold-ringed badges are the ★★★ destination-grade places.

Food 8.0+

Restaurants from The Infatuation's London reviews (~2,500 reviews scanned, Aug 2026). Kept: rating ≥ 8.0, price £–£££, minus obvious fine-dining, tasting-menu and special-occasion-only places — this is a casual-eating map. Bold ring = rated 9.0 or higher.

Stations

Every Underground, Elizabeth line, Overground, DLR and rail station in fare Zones 1–2, from the official TfL API. Bold ring = well-connected: three or more lines, an Elizabeth line stop, or a major interchange.

Good life 🥑

The "Waitrose & GAIL's" toggle on the map. A tongue-in-cheek gentrification index: branches of Waitrose, GAIL's, Aesop and Whole Foods, from their official store finders (Sep 2026). Where these cluster, the flat whites are reliable and the rent is not.

Gyms 🏋

Branches of the gyms and class studios from my own gym shortlist (Oct 2026), from each brand's official club pages. Each brand carries a mark out of 10 for fit for me — training quality, atmosphere, facilities and member feedback from Reddit, Trustpilot and branch reviews — not a review-site average. Bold ring = the shortlist (Third Space, BXR, Equinox, Gymbox, plus BLOK among studios). Mass-market chains (Virgin Active, Fitness First, Nuffield, Better, Anytime, PureGym, The Gym Group) are on the map too, muted and tagged "Chain". Prices were checked on 11 Oct 2026 and exclude joining fees.

New-builds

Reviewed new-build and major-conversion developments from a parallel research project, london-newbuilds, which swept planning records, property portals, architecture directories and ~40 developer portfolios across Zones 1–2 into a catalog of 1,422 developments backed by claim-level evidence. Inside the dotted area the map shows three reviewed tiers:

🏢 B — investigate (531): credible candidates — upscale, architecturally significant, or a genuine conversion. A navy badge means worth a look, not good: the research rates no building "great" until comfort, operations and fire-safety evidence is verified, and none has cleared that bar yet. Each popup lists the open due-diligence gates as "Verify: …" tags (heating, cooling, acoustics, external walls, …) plus the catalog's known-issues note.

🏗 C — probably-ok holds (58): mostly schemes not yet delivered (dashed badge, "Future"), plus a few with unresolved safety status (amber badge — check the current FRAEW/EWS1 position).

C — probably-bad holds (39, greyed out): buildings with resident-reported operational failures or a confirmed fire-safety defect on record. They stay on the map because they're often famous and attractive-looking — the flag explains specifically why to be careful.

Tags name the thing to check rather than pass a verdict, and each popup is dated — safety and operations claims go stale, so recheck close to any decision. Not shown at all: reviewed holds that are simply not relevant to this hunt (affordable/student product, serviced stays, ordinary stock) and the unreviewed discovery pipeline. Each popup carries the catalog id; the full record, classification reasoning and evidence trail live in the research repo (pinned commit, regenerated here by scripts/build-newbuilds.py).

Crime heat

Recorded crime from the Metropolitan Police via the London Datastore (LSOA-level, 12 months to Jul 2026). Each small statistical area gets a severity score:

score = 0.7 × violent crime density + 0.3 × theft density

where violent crime = violence with injury + homicide + personal robbery + residential burglary (threats, harassment and no-injury assault are deliberately excluded — about 74% of the raw "violence" bucket involves no physical harm), and theft = theft from the person + other theft + bicycle theft (no shoplifting). Densities are per km² of each area's true size, each component normalized to its own 95th percentile, and the color ramp saturates at the 99th percentile of the score. Positions are area centroids (ONS), so the surface has ~250 m resolution — read it street-corridor by street-corridor, not door by door.