What changed in August 2026

Two more maps join the hub. The walks map draws every route we hold — all 185 — from the GPX recorded on the day rather than a line drawn between two summits, coloured by length: 41 easy, 73 moderate, 49 hard and 22 severe. Blank ground means we have not recorded a walk there, not that there is nothing to walk. The relief map shows the shape of the ground itself, hachured from a 30-metre terrain model across the whole National Park, with the seventeen lakes and sixteen settlements marked. Neither is a navigation map — take OS Explorer OL4 to OL7 onto the hill — they are for the planning you do before you leave. The walks map →

There is now one place for every map here, and two new maps to open it with. The bus map answers a question this site had been answering badly: 238 of the 598 fells have a checked bus stop within a kilometre, and 128 more within five, with the walking distance to each given rather than left to guess at. The fells map plots all 544 Lakeland tops by height with the 214 Wainwrights picked out, so it shows which list a summit belongs to and not only where it sits. The hub also gathers the parking and access maps, the Howgills and the perfect round, which until now were scattered. Maps of the Lake District →

Fourteen more routes carry written directions rather than only a distance and a climb: what each leg is actually like underfoot, the waypoints to tick off, the decision points, the places people get it wrong, and how to get off if the day turns. That takes it to 86 of our 185 routes, up from 72. Among them are Allen Crags and Glaramara, Bannisdale Fell, Dodd and Whiteside, and the Birketts around Devoke Water, Grisedale Pike, Carrock Fell, the Dunnerdale Fells and the Kentmere Horseshoe. None of it is computed — it is written from having walked it.

Wetherlam, Mardale Ill Bell, Ill Bell, Hart Side, Dale Head, Robinson and Seat Sandal now carry a walker’s note alongside the summary — the things a map cannot tell you. Where the sitting is dry and where it is not. Which side of Ill Bell is kinder on the knees on the way down. Why Seat Sandal is an easy bag for an adult and, on one occasion, a mutiny for two children. That is 80 fells with a note, up from 73.

A new piece answering the question everyone asks and nobody has data for: when should you actually come? April averages 16.7 dry days and 16.0 sunny ones. August averages 10.5 and 10.2 — level with February, and only December is poorer. That is eleven years of daily weather records for a point in the central Lakes. Separately, the national park's own parking figures show eleven of twelve car parks at their busiest in August and not one in April. Two sets of numbers, collected by different people for different reasons, pointing the same way. It also makes the case against April, which is real: colder than June, less daylight, and snow on the tops well into the month. April, not August →

The month block on the front page now covers all twelve months rather than eight, so there is a suggestion whatever day you arrive. May to August were the gap, and they were the hardest to choose: in high summer every walk fits the daylight, so the thing that separates one from another is not the light but the crowds. Those four months now lean on the parking figures instead — Threlkeld at 1.5 times its January traffic in June, against Brockhole at more than fourteen. Same four walks a month, chosen for a different reason than the winter ones.

Bakestall, Whiteside, White Side, Grey Friar, Green Gable, Grike, Gray Crag, Middle Fell, Fellbarrow, Brandreth, Long Side, Haycock, Caw Fell, Rosthwaite Fell, Graystones, Watson's Dodd, Great Rigg, Wether Hill, Arthur's Pike and Sergeant's Crag. Each answers in a sentence, then gives our own measured distance, ascent and grade, then the catch — which is rarely the climbing. On Watson's Dodd it is finding the summit at all; on Caw Fell it is how far you are from help; on Middle Fell it is that you must go up and down at the ends of the ridge and never off its sides. Every figure comes from our own route data, so it changes when the route does.

Forty-nine fell pages now say when the last bus leaves the stop nearest the walk — and which town it is heading for. That second half matters more than it sounds. Our figure is the last departure in any direction, so at Cow Bridge in Hartsop the answer is 19:51, which is right if you want Windermere and no use at all if you want Penrith: that one went at 17:45, from the stop across the road. Naming the town lets you see at a glance whether it is your bus. Twelve stops so far, each confirmed against the operator's timetable. Where a time cannot be pointed at a named place, nothing is published.

A hundred and twenty-two things we knew about places to stay were not being shown. The rule is that a fact only appears with the sentence from the property's own site that proves it, which is right — but the check for a yes wanted the word on its own, and these answers were written with their detail attached: "yes, shared" for a drying room, "yes, six spaces, plus unrestricted on-street" for parking. All true, all quoted, none of it printed. Parking gains thirty-four, dogs and family rooms twenty-four each. Four are still held back, because they begin with yes and then say the opposite — a tumble dryer is not a drying room, and we would rather say nothing.

A printable tick-list: every Wainwright in his seven books, highest first within each, with a box to tick and the height in metres. It is laid out for A4 and prints fine in black and white — nothing on it depends on colour — and it is an ordinary page rather than a download, so you print it or save it as a PDF straight from the browser. Built for the rucksack lid and the kitchen wall rather than the screen. The Wainwrights — a tick-list →

Three routes carried my own photographs yesterday; fifty-six do now, three hundred and forty-seven pictures in all. Each is placed by matching where it was taken against the line on the map, so the picture beside a leg is a picture of that leg. Several turned out to belong to a different walk over the same ground — one was taken twenty kilometres away — and those were moved to the walk they came from rather than published where they did not belong. A long descent can carry several; a short hop between two tops carries one, or none at all. Legs without a photograph read exactly as they did before.

The panel naming the summits that wait for the sun led with the two equinoxes and then named each top's slowest morning. But slowest means the worst of four dates, so wherever midsummer happened to be worse, the December figure never appeared at all. On thirty-four walks a summit that waits a serious length of time in late December said nothing about it: Castle Crag waits thirty-five minutes on the shortest day, and the page talked about June. Every material midwinter wait is now named, with the time the sun actually clears the skyline. Same modelling, same caveat — this is the skyline, never the weather.

The band showing how a day looks on the tops was a row of twenty-four pale blocks in three colours so close together that outdoors, on a phone, they were one colour — and a single hour was ten pixels wide, too narrow for anything but colour to live in. It is now eight blocks of three hours, each wide enough to carry an icon: sun, rain, snowflake, cloud or wind, so the shape tells you what the limit is and the colour only tells you how serious. Three forecast models still run behind it, and where they disagree the block is drawn fainter — faint means less agreed, never less serious. A key underneath names all of it.

The transport panel used to decide whether a bus served a fell from a list typed against each car park, and on 141 fells it was wrong — telling people to drive when a stop was a walk away. Latrigg said drive; it is 2.6 km and 300 m of climbing from a stop in Keswick. Every fell is now measured along real paths from the nearest stop, and lands in one of three shapes: a short step from the stop, a quantified walk in, or honestly out of reach. Where we cannot measure it — a bus whose stops we don't hold, a valley our path data doesn't join up — the page says so instead of telling you to drive. Last buses, Sundays and winter are still to come.

Three routes now carry my own photographs against the legs they belong to — the Fusedale Circuit, the Whinlatter Loop, and Lonscale Fell to Bakestall, twenty-six pictures in all. Each one is placed by matching where it was taken to the line on the map, so the picture beside a leg is a picture of that leg. A long descent can carry several; a short hop between two tops carries one, or none. Where there is no photograph the leg simply reads as it did before — an empty frame would say less than nothing.

Eleven routes now show their confirmed escape routes as a marker on the route map, not only as a flag on the elevation profile. The flag tells you how far into the walk an escape is; the marker shows you the ground you would be descending, which is the thing you actually want when you are deciding whether to carry on. Each one names where it comes out. Only escapes I have walked or checked appear — the rest stay off the map until they are.

Thirty routes now carry a stage-by-stage account of the walk rather than distances alone: what the leg is, the waypoints to tick off, the decisions, the places it is easy to go wrong, and the way off. The eight added this week are High Rigg, Great Mell Fell, Stone Arthur, Gowbarrow Fell and Little Mell Fell, Castle Crag and Brund Fell, Hartsop Dodd, the Sour Howes to Troutbeck Tongue round, and Blencathra to Mungrisdale Common.

A hundred route pages now say when the sun actually clears the skyline from each summit on the walk — not sunrise, which is a different thing on a fell with high ground to its east. Both equinoxes are shown because they fall in the walking season, and each summit also names the morning it waits longest. Rounded to the quarter hour and meant as a general idea rather than a schedule. Modelled from a 30 m elevation model against each summit’s own skyline, never observed.

Keswick, Ambleside and Grasmere now carry the full list of places to stay near them — every eligible property, with a full card where we have quoted a drying room, dogs, family rooms or parking, and a name-and-link line where we have not. Route and fell pages that used to repeat those lists now show a single line naming the trailhead and the town, and link across. The same hotel was being printed on as many as 58 pages; the walk itself gets the room back.

Every route page now draws its own elevation ridgeline in the hero, and the profile below the map is shaded by steepness — using the same four bands as the route’s steepness rating, so the chart is the evidence for the grade rather than a second opinion. Summits are numbered once, and those numbers run through the ridgeline, the profile, the stage notes, the pacing table and the summits list. The stage notes that change what you do on the hill — what is easy to get wrong, and how to get off — now stand apart from the ones that don’t. And the “Places to stay” recommendations no longer stop mid-sentence behind a “More on why” toggle.

The 509 and the 517 now appear on fell and route pages. The 517 is the Little Langdale Shuttle the National Park names without a number, and it is the only bus that reaches Little Langdale, so Lingmoor Fell, Wetherlam from the east, Holme Fell and Black Fell all gain a car-free option. The 509 runs the Ullswater side of the Penrith to Keswick corridor, calling at Pooley Bridge, Glenridding, Patterdale, Hartsop and Dockray, so there are now two options where the site showed only the 508. Neither entry states a frequency, because we have not verified a timetable and inventing one for a bus somebody plans a day around is not something we will do; both send you to the operator instead. Separately, 134 more car parks now say which buses stop at them, taking the total to 314. The threshold for attaching a bus to a car park moved from 50 to 150 metres, which is roughly a two-minute walk. Place pages now run their text at the full width of the card, as the feature pages do, with more space between the lines to suit the longer measure. And the events section on those pages has its panel back.

Twenty route pages around Borrowdale, Buttermere, Whinlatter and the Coledale fells now say whether the walk fits the bus: which service, when it reaches the start, when the last one leaves, and the time to be off the last summit to make it. That turnaround is deliberately 30 minutes earlier than the arithmetic allows. The 77, 77A and 78 run daily including public holidays, not just weekends like the Wasdale bus, so a car-free day in Borrowdale is a Tuesday in October. Two of the twenty finish in Keswick, so they say the better thing: there is a bus out and no last bus to miss. Times are from the operator's summer 2026 timetable, read on 21 August. Fell pages also carried 85 claims that a fell is 'also known as' something else, and 75 were wrong — Great Crag is not also called Stonethwaite; Stonethwaite is the hamlet telling you which Great Crag. Ten genuine ones remain. And the Coniston page has lost its one commercial link, so it ends on the FAQ rather than a place to buy a map.

Route and fell pages around Keswick, Ambleside, Grasmere and Wasdale now carry a “Places to stay” panel — 74 route pages and 172 fell pages, from 66 properties. On a fell page they are grouped by which start you would walk from. Everything each place offers is quoted from its own website — drying room, dogs, family rooms, packed lunch, step-free access. If we could not find a business saying it in its own words, we do not print it. No prices, no booking, no affiliate links, every link nofollow, and we have not stayed at any of them. Distances are straight lines, never walking times. Where a place states it does not take children we say so, because finding out at the door is worse. This is four valleys, not the whole Lake District: 257 of 779 collected entries have been checked against the page they were quoted from, and the rest stay hidden until they have been. Town and village pages do not carry the panel yet.

We told readers that Wasdale had no public transport at all, and that the fells above it were one for the car. That was wrong. The Wasdale Explorer runs from Ravenglass to Wasdale Head on Saturdays, Sundays and bank holiday Mondays through the summer — 2 May to 31 August this year — for £3 all day, with children free and well behaved dogs allowed. Sixteen fell pages carried the wrong advice, Scafell Pike among them. The first bus reaches Wasdale Head at 09:00 and the last one back leaves at 19:00, and both run the whole season, so a full day on the fell works without a car from the start of May. It begins at Ravenglass on the Cumbrian Coast line, so the journey can be done by train and bus throughout. Every affected page now names the service and links its timetable.

Opening the routes page used to fetch the full GPS track of every route on the site before you had chosen anything — 169 files, 41 MB, with browser caching deliberately switched off, so a reload paid for it all again. On a phone in the valley on the morning of a walk, that is a serious chunk of a data plan. The map now loads one small file instead, 281 KB, and not until you scroll to it. The drawn lines are simplified for an overview map; every route's full GPX download is unchanged. The page also has a search box now, so you can find a walk by where it starts rather than scrolling past 168 cards.

The email signup in the footer was styled for a light background and then placed on the dark green one, so its heading and small print sat at about a fifth of the contrast they needed and the button all but vanished into the background. On the home page the styling was missing altogether, leaving a plain browser input and button too small to tap comfortably. Both are fixed everywhere the box appears.

A new guide gathers all 1,043 parking places behind the fell and route pages into a single map and list — 464 free, 197 that charge, 542 car parks, 398 lay-bys and 103 verges — filterable by charge, kind, bus route, facilities and whether we know the space count, and it filters without JavaScript. It is the first place on the site to surface the fallback chains: where a car park has known alternatives, the page says which and how far. Provenance is stated rather than implied, including the 244 entries we hold a position for and no details at all. It carries no prices and no risk ratings — the page explains why on both counts. Marked beta while the data is validated in the field; corrections go to parking@hikes.guide.

Fell, route and place pages now carry one short, calm note beside the parking block: vehicles are left entirely at their owner's risk, we list places people park but cannot vouch for any of them, and the practical advice is to take valuables with you, leave nothing on show, and park somewhere else if a spot feels wrong when you get there. It is deliberately a standing condition of parking anywhere rather than a warning about any particular site — we publish no safety rating, no incident counts and no claim that any place is safe, because the police data behind such claims is anonymised to approximate points and would overstate the risk at named car parks. It appears once per page, not once per car park.

Every data table on the feature guides — 54 of them across 21 pages — now scrolls inside its own box. Before this, a table wider than a phone screen could drag the whole page sideways as you read, or, worse on one page, push its contents links off the edge where they could not be reached at all. The tables themselves are unchanged; only where the scrolling happens has. Found by checking the last-petrol page on a 375-pixel screen, which is the width of a fairly ordinary phone.

The last-petrol page now draws each drive in as a strip-map: a single road-line from where you leave the fast road to the valley head, every filling station marked at its true distance along that road, the last one flagged, and the gap beyond it shaded with its length written into it. Nine of them, one per corridor, drawn to scale — so the picture is the measurement rather than an illustration of it. The longest gap is 24.5 km, from Bridge End Garage in Egremont to Wasdale Head coming down the coast; the shortest is 9.9 km into Mardale. Two of the drives share every road but the last and are drawn as one fork, because the shared stretch is genuinely one drive and only the gap at the end differs. Penrith Truckstop is drawn on its line and struck through: it is fuel-card only, no petrol and no cash, and your map app will offer it to you at one in the morning. Alongside it, a new panel answers the other half of the question — if you are setting off before the shops open, which is the first place on your road that is actually open, last station first. Every number is the one already published in the tables below it; the drawings measure nothing new. Also fixed: on a phone the tables were pushing the whole page sideways.

Where the last filling station, shop and cash machine actually are before each of the seven dead-end valleys — measured along the road rather than as a straight line, and verified against the operator's own source or the government's statutory fuel register rather than off a map. The page leads with what you PASS coming in, not what is nearest, because the two turn out to differ by up to 16 km: Eskdale's nearest fuel is 12.0 km away, but arriving from the M6 the last station on that road leaves 28.2 km still to drive. Fuel is the boring half — every valley sits 9.9 to 18.2 km from a pump, a factor of 1.8. The last shop ranges from 0.2 km at Glenridding to 13.3 km at Seathwaite, a factor of 66. Every corridor names the roads it uses, so a route over Wrynose and Hardknott says so. Checked 8 August 2026, and the page declares its own age.

The sunrise page has always argued that a summit's horizon is made of other fells and that the sun has to clear them. It now draws it: four summits, each with its whole 360-degree skyline as a solid silhouette, every sunrise and sunset of the year marked at the bearing it crosses. Carl Side has Skiddaw standing 13.7 degrees up across its north-east, so the midsummer sun climbs the shoulder of it and arrives 95 minutes late, where the midwinter sun clears open ground and is 2 minutes late. Helvellyn has nothing above the horizontal anywhere. All four share one vertical scale so they can be read against each other, exaggerated about four times, and the caption says so.

The sunrise page has always answered "when does my fell get the sun?". It now answers it the other way round: pick one of the sixteen dates we model and see the five summits that catch the light first and the five that hold it longest. Often there is no single winner - seven summits share first light on the shortest day, ten share it on New Year's Day - so the page names the count rather than crowning a fell, and equal times get equal ranks. It covers the sixteen dates the dataset actually solved, not any date you like. It is a geometry table, not a recommendation: we still publish no time to set off or start down.

We have held last-light figures for all 214 summits since the start of the month, and the page that explains them was still telling only the morning story. It now tells both ends: the fells that lose the sun first, those that hold it longest, and the 48 that keep it past the valley's own sunset. Castle Crag loses the sun at 19:30 on midsummer's evening; Helvellyn, nine kilometres away, keeps it until 22:00. Putting the two together showed something neither half did: summits lose light far more easily than they gain it. The best gain is seven minutes at dawn; the worst loss is 143 minutes at dusk. It still will not tell you when to start down.

The two decisions in the fell first-aid guide you will least want to be reading off a phone screen - the DR ABC primary survey, and whether to walk someone out or call Mountain Rescue - are now on four cards you print on a single sheet of A4, cut into quarters and fold. No double-sided printing and no special paper: the backs are printed upside down on purpose so each card lands the right way up, and Lee tested the geometry on real paper first. The other cards carry the kit list, the hypothermia signs, and a plan you fill in by hand before you go. Every word is lifted from the guide unchanged. It is still not a substitute for a first-aid course, and the cards say so.

Our place pages told you what a valley was and which fells were near it, then went quiet on the three things people actually ask: where do I park, can I sleep there, and how long is the walk. Fifty-five of the fifty-six now carry the car parks near them with whether each charges and who runs it, thirty-two carry the land manager's own published position on overnight parking quoted word for word, and forty-nine list the walks that start on the doorstep. Where no published rule exists we say nothing, because that is not the same as permission. We print no price, no claim about how early a car park fills, and no time to leave by.

Sadgill in Longsleddale was 236 m from the parking it describes, the Kentmere Institute 179 m, and the Gale Road pin for Latrigg 150 m. Those three and fourteen more are now on the mapped car park. Seven are corroborated by one of our own recorded walk starts landing within 25 m of the new position; the rest rest on an unambiguous OpenStreetMap match with nothing competing. Three came out of a pile we had set aside as unanswerable, where several car parks sat close together and no name matched - asking which one our own walks started inside settled it. What this does not promise is the entrance: 74 of our 157 spots sit on a mapped area more than 25 m across.

A new feature page answering a question nobody else answers in walker-shaped form: can you charge the car while you are on the fell? We checked 51 candidate chargers against operators' own published information and 18 survived. The honest headline is the gap - 47 of the 157 places we publish parking for have nothing recorded within 5 km, including Rosthwaite and Stonethwaite, which are how you reach Great Gable and the Scafells. Six walks in the whole National Park start beside a verified public charger, and they are really three car parks. Charge is given in kWh, never as a percentage of a battery.

Maggie's Bridge at Loweswater, Walna Scar Road above Coniston, Cow Bridge at Hartsop and Scout Scar outside Kendal were all shown here as free, and all four charge. Our own data had them right; the name and the note printed on the page did not, and the name is what you read. Both are corrected, and the build now refuses to ship an entry whose name contradicts its own fee, so this particular mistake cannot come back. The same sweep re-checked all 157 parking locations against the surveyed map, our recorded walks and the tracks on our own routes, and found two pinned in the wrong valley: Seathwaite in the Duddon was pinned at Coniston, seven kilometres away, and Dockray was pinned in Penrith, fifteen. Both now point where their names say. The pin at the top of Honister Pass has moved onto the slate mine's own car park and now says which one it is — there are two up there, about 170 metres apart, and the other belongs to the National Trust.

Reading about contours is not the same as reading contours. The field guide now has three self-check drills built from the elevation model this site already runs on: fifteen patches of real ground to call spur, saddle or re-entrant; eight two-kilometre panels with a summit marked, for six-figure grid references; and six legs between named summits with a compass bezel you turn to the bearing — drag it, or use the buttons. Nothing is scored, nothing is timed and no streak is kept. A wrong answer explains the shape, or names the tenths you misjudged, or tells you how far off the summit you would finish. Every answer is computed rather than remembered, and each grid reference matches the one on that fell's own page. Section six has stopped asserting a date-stamped fact and started calculating it: magnetic north currently sits about three quarters of a degree east of grid north over the Lakes, and the page now says so with the figure, the source and the year the advice expires.

The Clag Index page told you which fells are hardest to stay found on. It now shows you. Twelve summits where the trap is confirmed are drawn from the elevation model twice over: the whole top as a map shows it, and then only the ground you could see in cloud — everything beyond about fifty metres removed. High Raise's circle is nearly empty, and that emptiness is the whole problem. Helvellyn's contains the edge. Two of the twelve are boulder mazes that a 30-metre model simply cannot draw, and those two say so on their face rather than looking deceptively calm.

Type a fell name at the top of the Clag Index and you get a short briefing on that one top — how hard it is to stay found on, which of the three traps it has, and the note from actually walking it where one exists, word for word as it appears on the fell's own page. The address updates as you go, so a warning can be pasted straight into a group chat. The full table is still there underneath.

Reading contour shapes is a skill you cannot get from a diagram of an imaginary hill. The contour section now offers fifteen patches of actual fellside — spur, saddle or re-entrant? — drawn from the same elevation data as the rest of the site, and it tells you afterwards exactly which hillside you were looking at. Nothing is scored, timed or counted: getting these right is not the same as being able to navigate, and the page says so every time.

The parking list grew from 119 to 157. Every position was checked on the map before it went in, and the ones that are roadside or a verge rather than a car park are named and labelled as such, because turning up expecting a car park and finding a single-track lane is worse than not knowing. Also fixed: a fell's page could name the nearest car park we held even when it was miles away — Brant Fell above Bowness was pointing at Hawkshead, across the lake.

The boat companies sell the lake as a cruise and the walking sites start every Ullswater route from a car park — so the most useful thing about the steamer goes unsaid. It solves a parking problem: Howtown sits under a cluster of quiet Far Eastern fells (Hallin Fell, Steel Knotts, Bonscale Pike, Arthur's Pike) at the end of a narrow lane with nowhere to leave a car, and the pier gets you there without one. It also turns out-and-backs into point-to-points — sail Glenridding to Howtown and walk back over Place Fell and you get a high traverse instead of the same ground twice. We've added the boat as transport to the Ullswater, Glenridding and Patterdale pages (Pooley Bridge already said it), and to four routes it genuinely helps. One piece of advice throughout: take the boat first and walk back, so a slow day can't become a missed last sailing. We deliberately publish no departure times or fares — a stale one would be worse than none — only a link to the operator's timetable and the date we checked.

All 155 route pages were titled “… Hiking Route”, but the fell pages have always said “Walking Guide” — so the site was using two words for the same thing. British walkers overwhelmingly say walk (Catbells: 4,700 searches a month for “walk” against 10 for “hike”), and that holds even on Scafell Pike, so it is now walking everywhere in page titles. The bad-weather guide is renamed to match the name the rest of the site already links to it by. No web addresses changed, so every existing link and bookmark still works.

A collection only showed the little description under each walk once nearly all of them had one, so a page with a couple of gaps showed none at all and read as bare. The rule was counting the fells on those pages as missing descriptions when they had perfectly good ones of their own, which meant some collections could never have qualified however much was written. Fixed — 22 collection pages now show them, up from one.

Each fell page was loading Lee's entire Strava history — 454 walks — before it would finish, to fill in a “my hikes over this fell” panel near the very bottom that usually matches one of them. It now loads only if you scroll that far, which takes about 70% of the JavaScript off the page. The fell pages also stop laying out the parts of the page you cannot see yet, and the stylesheets arrive as one file instead of four.

Is it a Wainwright, where do I park, and how hard is it — the three questions that sit under nearly every search we see. They are now answered on each fell page from the data the site already holds: the list memberships, the nearest car park within 2.5 km measured straight-line, and the route's difficulty axes. Where a fell already has a hand-written answer to one of the three, the hand-written one stands and the generated one stays out of the way. A fell with no route on file gets no difficulty answer rather than a guess.

A glossary of 256 terms, from the ones you can guess (col, tarn, scree) to the ones nobody explains (clag, Birkett, bagging, a rake, a rigg). Each is written as a walker would say it rather than as a dictionary would, and where a word appears on our own pages it links back to the fells and routes that use it. The second half turned out to be the interesting one. Of the 443 name-elements behind the 214 Wainwrights, most are not English. Fell, tarn, force, gill, howe and thwaite are Old Norse; crag, pen and blen are Brittonic, older still, left by people who were here before either. The map carries three languages at once, and the names quietly record who named what. That is drawn rather than described. All 214 are plotted on a relief map, marked by the oldest language in the name, so the Norse ground and the Brittonic pockets separate on sight. Where a name has a contested origin we say so instead of choosing the tidier story.

The commonest disappointment at a Lakeland waterfall is arriving to find a damp streak on a rock. All 15 now carry a verdict for the next two days, worked from the rainfall forecast for the fell area each one drains against how much rain that particular fall needs before it performs. They are not equal: a big catchment keeps a fall going through a dry week, while a small steep one needs rain in the last day or two or there is nothing to see. Each verdict shows the number behind it, so you can disagree with it. It is a forecast of rain rather than a measurement of water, and a fall below a reservoir answers to the gate rather than the sky - both of which the page says on its face.

Put in the dates you are in the Lakes and the events page now shows what is on while you are there. It has looked like that was possible for a while. It was not: the code doing the filtering sat on a different page entirely, where the first thing it did was look for a control that was not there and give up. The section rendered and did nothing, on every visit, from the day it was added. It works now, and the dates go into the address bar, so a filtered list is a link you can send to whoever you are going with.

Card pictures on the lakes and waterfalls pages are about 290 pixels wide at most. We were sending files up to 1400 wide to fill them - roughly five times the detail the space could show, seventeen times over on a single screen. Each card now arrives in a size chosen for the screen asking for it, taking a typical one from about 90 kilobytes to 17. Nothing looks different, which is the point. It matters on a phone on mobile data at the start of a walk, which is when these pages actually get opened.

The View Report's table of what can be seen from each summit was built on pale striped rows. In dark mode the stripes stayed pale while the text turned light, leaving 321 cells of near-white on near-white - a contrast of about 1.15 to 1, where 4.5 is the readable floor. The note about seeing the Isle of Man was worse at 1.07, and the good-conditions verdict on the things-to-do page had the same fault. All three were backgrounds given a fixed colour long before dark mode existed, so they never changed with it. They were found by measuring every element on nineteen pages in both themes rather than by looking - which is the only way this kind of fault surfaces, since it is invisible unless you use the theme it breaks.

The lakes page now opens its comparison section with a drawing rather than a table: all 17 lakes in one continuous cross-section, deepest first, on a shared waterline. Wast Water's 79 m against Elter Water's 7 m is more than eleven to one, and it was invisible in a column of figures. The uncertainty is drawn too - only 6 of the 17 depths come from a survey we can cite, and the other 11 fade towards their floor behind a broken line, meaning we are less sure of the number, not that the lake is shallower. Depth and length cannot share a scale, so the vertical is exaggerated about 294 times and the chart says so. No figure appears in it that the table beneath does not already publish. The page around it also picks up the site's current design, which it had never been switched on for: headings are one size for one role instead of three, the content sits on a measured page rather than running edge to edge, and the tap targets are finger-sized.

On the waterfalls page the Download the data button was rust-brown text on its own dark green background - 1.02 to 1, which is to say invisible. Not a waterfalls fault: the redesign recolours links inside a content card, and this button style was left off the list of exceptions, so it broke wherever one appeared. Fixed everywhere, with the other button styles audited for the same collision. Under Managing bodies and sources the same page printed a row of bare dots - eight empty links and eight keyboard stops announcing nothing, on the section that backs its claim to be independently sourced. It now lists the two real sources, and the build refuses to publish a source it cannot name. Separately, two feature pages were unreadable in dark mode: the central claim on the drinking-water piece sat at 1.25 to 1, and parts of the mountain rescue report near 2 to 1. Both pages now pass in either theme.

Every Wainwright page told you when the sun first clears that summit's own skyline. It now tells you when it leaves. The two are not mirror images - a fell can have an open eastern horizon and a wall to the west - but both come from one calculation against one 360-degree profile, so they cannot disagree. Castle Crag loses the sun at 19:30 on a midsummer evening; Helvellyn, ten miles away, holds it to 22:00. 48 of the 214 keep it past the valley's sunset on every date we model. Only those 214 have a profile; the rest show nothing rather than a guess. The admission: on 24 fells the sentence above the sunrise table contradicted the figures beneath it - Loft Crag claimed first light within half an hour of sunrise directly above a row reading 50 minutes. It described a fell's typical morning, and a fell whose sunrise bearing swings 87 degrees across the year has no typical morning. Descriptions are now pinned to each fell's hardest day and name the date. We have also stopped averaging horizon height across a 90-degree arc, which cannot see the one ridge in the way: Green Gable averages 1.8 degrees and still loses 92 minutes to Great Gable. Pages now name the bearing and what stands there. Model outputs, each with its plus-or-minus band, and not a descent plan.