Backlog

The roadmap says what belongs together and why. This page says what order.

Everything starts in the pool at the bottom and works its way up. The bands are not sprints and they are not dates — they are a rough answer to "what would we pick up next", so that nobody has to re-derive it from twelve phases every time something finishes.

:::note The order is loose, and expected to move

Nothing on this page is a commitment to sequence. It is one defensible ordering, written down so there is something to argue with — moving an item up because it turned out to matter more, or down because it turned out to be harder, is the page working rather than the page being wrong.

What is not loose: the dependency arrows in the roadmap. An item cannot jump above something it structurally needs, however much we want it. :::

How to read it

Marker	Means
Decide	A question, not a build. Costs an afternoon, unblocks weeks — see decisions we owe
Build	Actual work
🔑	On the critical path. Something expensive is waiting behind it

Status lives on the roadmap, not here. That page's tables are the record of what has shipped; this page is only the running order. Two places tracking "done" is two places to get it wrong.

Now — the core

The bit you cannot ship without. Deliberately short: if this band grows past about ten items it has stopped being "now" and started being "everything".

#	Item	Kind	Why it is here
1	Identity — supersede ADR-0007 🔑	Decide	Six phases wait on it and it is the hardest thing to retrofit
2	Accounts and identity, with device-id migration	Build
3	Image storage and delivery — bucket, sizes, CDN	Decide	Nothing visual works until this is settled
4	Image storage and delivery	Build
5	Crag view — sectors in walking order, parking, approach	Build	The screen the whole app is an excuse for
6	Access restrictions, prominent and dated	Build	Safety, cheap, and belongs with the crag view rather than after it
7	Sector view — topo viewer, pinch, pan, route lines	Build	`TopoOverlay` already draws lines
8	Tap a line to select a route; tap a route to light its line	Build	The core interaction. Everything else is navigation to reach it
9	Route detail — description, grade, gear, first ascent	Build
10	Download this crag, manually, with size and progress	Build	Manual first; automatic is a later problem

This band is done when you can log in, pick a crag, look at a sector, tap a line, read the route, download the lot, and do all of it again in airplane mode.

Next — a guidebook you would actually take

#	Item	Kind	Why it is here
11	Multi-pitch and alpine data model 🔑	Decide	Free to decide now, a migration to decide later. Build is in Then
12	Route tag vocabulary	Decide	Free text fragments within a week and cannot be un-fragmented
13	Route tags and pills	Build	Safety tags as their own class
14	Real data for at least two crags	Build	A licensing question, not a technical one — start it early, it has lead time
15	Filter by grade band, type, stars and tag	Build	`gradeBand` is already stored for exactly this
16	Offline search across everything downloaded	Build	SQLite FTS5. No network path at all
17	Approximate grade conversion beside the published one	Build	ADR-0010
18	Log an ascent, offline, through the outbox	Build	Schema and sync already support it
19	Logbook screen — history, filters, totals	Build
20	Tab structure	Decide	Cheap, and it gets more expensive with every screen built against the old one
21	The five tabs — Nearby, Map, Plan, Logbook, Feed	Build	Feed tab can land empty; the slot is the point

Then — the things a book cannot do

#	Item	Kind	Why it is here
22	Offline map tiles	Decide	Mostly a licensing question wearing a technical hat
23	Approach — parking pin, walk-in, offline tiles	Build	"Where do I park" comes before "which route"
24	Pitches as first-class records	Build	Against the model decided at #11
25	Descent and retreat as their own records	Build	More days go wrong on the way down
26	Schematic pitch topos	Build	A 300 m route does not fit a photograph
27	Dated condition reports that visibly age	Build	Alpine facts decay; the record has to say when it was last true
28	Storage tiers and eviction	Decide	Cross-cutting to sync
29	Three-tier auto-download with an LRU budget	Build	The crag you looked at on Thursday is there on Saturday
30	Partner tagging and mutual ticks	Build	Needs a consent model — an unaccepted tag must not create a tick
31	Wishlist, ticklist and trip planning	Build	Also what makes pre-emptive download worth having

After that — the community, and paying for it

Everything in this band waits on identity, and most of it waits on a licence.

#	Item	Kind	Why it is here
32	Contribution licence 🔑	Decide	Cannot be applied retrospectively. Must land before contribution one
33	Add a crag from the crag, fully offline	Build	The keystone — everything downstream consumes what it produces
34	Draw a route line onto a photo, in-app	Build	The hard, valuable one
35	Corrections as field-level proposals, with reasons	Build
36	Route and crag change history, visible in the app	Build	The best argument that we are more trustworthy than the book
37	One-tap problem reports and immediate safety banners	Build	No review latency on this path
38	Trust and moderation model	Decide	The other one where the failure mode is somebody getting hurt
39	Moderation queue and reviewer tooling	Build	Probably a small web app
40	Maintainer roles, attribution, crag adoption	Build
41	Revenue model	Decide	Determines what the community is incentivised to document
42	Contributor splits and what counts as use	Decide	Whatever is chosen becomes what people optimise for
43	Attribution ledger, payouts, contributor dashboard	Build	KYC and tax are a real cost, not a footnote

Later — reach

#	Item	Kind	Why it is here
44	Follow, feed, sessions, reactions	Build	ADR-0009
45	Crag conditions posts, expiring after 48 hours	Build	The most useful post on any climbing feed
46	Sharing, redistribution and quiet crags	Decide	A share is redistribution under somebody else's licence terms
47	Share cards, Instagram Stories, sticker pack	Build	The cheapest growth the product has
48	Live camera, stage A — which topo am I looking at	Build	Useful on its own, and needs nothing new in the model
49	Live camera, stages B and C — homography, anchoring	Build	Where it becomes magic
50	3D asset pipeline	Decide	Constrains the #33 capture flow, which is much earlier
51	Sector geometry, bolts, gear placements, beta	Build
52	External references and UKC	Decide	Start with a conversation, not a scraper
53	UKC import, route matching, full export	Build	The cost of entry for anyone with fifteen years of ticks

The pool

Everything else, and where everything starts.

The wishlist is the pool: ideas nobody has committed to, captured so the architecture does not rule them out by accident. It is deliberately unordered, because ranking things nobody has committed to produces an argument about the ranking.

An item leaves the pool by being promoted into a band on this page, which is also the moment it gets a phase on the roadmap and, if it carries a technical commitment, an ADR.

Currently in the pool: bolt condition history, goal-driven training, the conditions answer, access alerts, offline crag sharing between phones, a web contributor tool, crag ethics and etiquette, first-ascent stories, partner finding, publisher licensing, Apple platform extras, and everywhere that is not the UK.

Keeping this page honest

Order changes freely. Moving an item is not a decision and does not need an ADR.
Status does not live here. Shipping something means updating the roadmap table, per CLAUDE.md.
Adding something new means adding it to the pool, not to Now. Things earn their way up.
If "Now" has more than about ten items, it is not Now. Push something down rather than widening the band.

How to read it​

Now — the core​

Next — a guidebook you would actually take​

Then — the things a book cannot do​

After that — the community, and paying for it​

Later — reach​

The pool​

Keeping this page honest​