lanework/research/01-market-overview.md

Market Overview — Synthesis of the Broad Pass

This is the synthesis of the 2026-08 broad pass over the kanban and adjacent todo/task-management markets, drawing on the four segment documents (02–05) and the agent-intersection deep slice (06). It maps the landscape, places Lanework in it, and distills what the evidence says about how apps in this space succeed and fail. Facts are cited loosely here; the segment docs carry the sources.

Researched: 2026-08-01

The map: four rings around Lanework

The market divides cleanly into four rings, from farthest to closest to Lanework's position.

Ring 1 — mainstream team SaaS (doc 02). Trello, Asana, Monday, Jira, ClickUp, Notion, Linear, Airtable. Per-seat subscriptions, cloud-hosted, multi-billion-dollar revenues. Lanework does not compete here and shouldn't try; this ring matters as the source of the frustrations (bloat, seat-step billing, AI credit metering, cloud lock-in) that push users toward everything in the inner rings. Two structural facts stand out: kanban is usually a view over some other primitive (database, canvas, list), rarely the product itself; and offline/local-first is a near-universal gap treated as degraded fallback, never as architecture.

Ring 2 — personal todo apps (doc 04). Todoist, Things 3, TickTick, OmniFocus, plus the daily-planner wave (Sunsama, Akiflow, Motion). The durable businesses here are bootstrapped (Todoist, TickTick); the VC-funded ones keep pivoting out of the category (Motion to "AI Employees," Amie to meeting notes). Apple Reminders is the market's free ceiling — differentiation against it has to be structural, not featural. Kanban views exist (Todoist Boards, TickTick, Any.do, Lunatask) but always bolted onto a flat task list; nobody in this ring treats the board as the native data model.

Ring 3 — native/indie/file-based tools (doc 03). Lanework's actual neighborhood. The striking finding: at least six independent developers converged on "folder = column, file = card" within roughly 2024–2026 (Signboard, Tasks.md, Kanban MD, Kanbanero, banban, FileBoard). Tasks.md's 2.2k GitHub stars is the clearest demand signal for exactly this pattern. But every occupant is compromised for Lanework's purposes: Signboard is Electron; Tasks.md needs a self-hosted Docker container; Kanbanero is JSON with no card-body richness; FileBoard is brand-new and thin. The largest existing audience for markdown-backed kanban sits inside PKM tools (Obsidian's Kanban plugin, Logseq's plugin-only kanban) — real demand, always a secondary feature.

Ring 4 — the agent intersection (doc 06). The newest and fastest-moving ring. Markdown files are already the default substrate for feeding work to AI coding agents; Beads (~26k stars in months) proved pent-up demand for "an issue tracker an agent can drive"; OpenAI's Symphony turned a Linear board into a Codex control plane and reported a 500% increase in landed PRs — the strongest evidence anywhere that a board with real states unlocks agent throughput. Every mainstream tracker converged on hosted MCP + assignable agents within twelve months, so agent API access is now table stakes. A dense cluster of "kanban for agents" apps exists (Vibe Kanban, Conductor, Kanban Code), but all keep app-owned state.

Where Lanework sits: the whitespace, precisely

The broad pass supports a precise statement of the gap: nobody combines a genuinely good native human kanban UI with files-on-disk as the true source of truth, equally legible to any agent, script, or editor.

Every agent-first tool has a weak or absent human GUI (Backlog.md, Beads, GNAP). Every polished GUI tool owns its state (Kanban Code's ~/.kanban-code/*.json, Vibe Kanban's internal store, Kanbanero's JSON, Sprout's CoreData). Every mainstream tracker mediates agents through a server API. Lanework's "boards are folders and Markdown files; the app is the best human surface over that substrate" is validated from three independent directions — the file-over-app movement (Obsidian's scale, Steph Ango's essay), the convergent folder-as-board hobby projects, and the agent-tooling wave — yet unoccupied as a polished native product.

Two qualifications keep this honest. First, the window is narrowing: agent-friendliness went from novelty to expected-by-default in months, and an existing player (Signboard already ships an MCP server) could plausibly close the gap faster than Lanework builds breadth. Second, the demand signal for local-first is broad but not kanban-specific — the 9,300-post Reddit analysis found local-first demand at 7% of app wishes but no kanban-format requests within it; the kanban-specific demand evidence is developer-shaped (GitHub stars, HN threads), which says something about where early adopters will come from.

What makes apps in this space succeed

Synthesized across all four rings, ordered by how consistently the evidence supports each.

1. A sharp core promise kept undiluted beats feature breadth. Things 3 thrives with no AI and no board view because the core promise (fast, beautiful, pay-once, reliable) never diluted; Linear won the Jira-refugee segment on focus and speed; every all-in-one player draws bloat complaints even while growing. Nobody has cracked simple-and-complete; picking a side and holding it is the proven pattern.
2. One-time or low-commitment pricing is a trust signal, not just a price point. Subscription fatigue is measured and rising (active cancellation 31%→47%, 2024→2026). The $0–25 one-time range dominates native Mac kanban; users revolt not at price but at value-removal (Todoist's regional-pricing pull, Akiflow's billing surprises). Agenda's "pay once, keep what you unlocked" hybrid and Curio's MAS-subscription-or-perpetual-key dual track are the interesting patterns worth studying for Lanework's own model.
3. Bootstrapped survives; funded pivots or dies. Every notable 2024–2026 failure (Height, Clockwise, Rise, Mayday) was VC-funded; the durable individual-focused businesses (Todoist, TickTick, Obsidian, TaskPaper's 15+ years) are all bootstrapped. Obsidian is the existence proof for the ceiling: ~$25M ARR, 18 people, zero VC, local-first — real money without the enterprise axis.
4. Native polish and speed are stated purchase reasons in the indie niche. Repeated Electron-sluggishness complaints, repeated "feels like a real Mac app" praise; keyboard-first workflows lead the marketing of nearly every respected tool in ring 3. Lanework's native-Swift, keyboard-complete architecture is squarely what this buyer says they want.
5. AI that removes friction sticks; AI that replaces judgment doesn't (yet). Natural-language capture demonstrably stuck (Todoist Ramble); summarization is used but not a switcher; full-delegation auto-scheduling is commercially fragile (Clockwise dead, Motion fleeing to B2B), and Height died five months after going all-in on autonomous PM. For Lanework the corollary is convenient: because the substrate is files, agents bring their own AI — Lanework never has to run, meter, or upsell compute, sidestepping the AI-credits resentment that plagues ring 1.
6. Abandonment is fatal in the indie niche; users tolerate minimalism but not staleness. TidyCards (beloved, then stalled, then complained about) and Wunderlist (acquired, starved, killed) are the standing cautionary tales. Visible, steady development is part of the product.

Threats and cautions

• Obsidian Bases shipping a first-party kanban view could commoditize markdown-board demand inside the largest existing file-based audience.
• Signboard (Electron, same philosophy, MCP server, shipping since fall 2025) is the closest live competitor; "native, faster, more Mac-idiomatic" is the differentiation story against it and it must stay true.
• Pure-file storage has known scaling costs. Taskwarrior's retreat from flat files to SQLite, Beads' entire thesis (markdown degrades under concurrent agent writers), and the Backlog.md thread's context-budget caveats all mark real technical territory Lanework must navigate — the write-provenance ledger and atomic-write architecture are already aimed at this, but multi-agent concurrent writes remain the hard open problem in the whole space.
• Category absorption from above: Notion/Tana/Capacities absorbing tasks into cloud PKM graphs squeezes the casual end, and Apple Reminders squeezes the bottom. Both squeeze toward the same conclusion: Lanework's buyer is the person for whom file ownership and agent legibility are structural requirements, not nice-to-haves.

The direct competitive set (for the narrow pass)

Ordered by how much attention each deserves in pass two: Signboard, Tasks.md, Kanban Code, Backlog.md, Beads, Kanbanero, Obsidian Kanban plugin (+ Bases trajectory), Vibe Kanban, FileBoard, Sprout, Conductor. Things 3 and Obsidian stay on the list not as competitors but as the business-model and community-playbook references respectively.

The Signboard deep dive is done: 07-schema-comparison-signboard.md compares the two on-disk formats and concludes the differentiation is durable and structural — Signboard's Obsidian-shaped format (mutable filenames as identity, prefix ordering, denormalized location keys) cannot adopt Lanework's posture (immutable UUIDs, rank ordering, versioned contract) without breaking every existing vault, and vice versa. It also sets the template for the remaining pass-two profiles.

Open questions for the narrow pass

• Positioning and messaging study of the direct set above: exact feature matrices, pricing, and each one's actual traction beyond star counts.
• Concrete launch case studies (HN/Product Hunt) for file-based task apps with real numbers — this pass found none; searching HN/PH directly rather than via aggregators is the fix.
• Whether the agent-tooling audience (git/CLI discovery funnel) and the Mac-indie audience (MAS/Setapp/r/macapps funnel) overlap at all, since Lanework needs both and they may never see each other's channels.
• Reddit/App Store review mining for kanban-specific (not PKM-general) local-first demand.
• Pricing-model design: test the Agenda hybrid and Curio dual-track patterns against a straight Things-style one-time purchase for Lanework specifically.
• Superlist's post-merger direction, as the one well-funded personal-task player still actively expanding.