{"problems":[{"id":"problem_be1211f5-9004-4004-805a-ed92cfe63e29","title":"Automotive Buyer's Agent on Open Dealer Protocols (UCP, MCP, A2A)","category":"Product","ownerName":"Nick Baguley","ownerEmail":"nicholas@badlabels.com","company":"Bad Labels","description":"Build a consumer buyer's agent connected to DMC-12 at Mark Miller Subaru: discover fit, shortlist real inventory, request asking-price quotes with itemized out-the-door estimates, optionally soft-hold a VIN, and preserve shopper preferences across a multi-day purchase journey, using live MCP, A2A, and UCP surfaces instead of ad-hoc integrations. An optional Per Se outcome and approval layer can be added on top for declared intent, human approval gates, and proof receipts.","affectedAudience":"Vehicle buyers facing multi-hour, multi-day purchase cycles, and dealership groups seeking agent-native commerce across AI platforms.","currentCost":"Disconnected inventory search, opaque pricing, repeated data entry, scheduling friction, and post-sale service coordination gaps that erode loyalty.","desiredOutcome":"A demo-ready buyer's agent that reduces friction in acquisition by automating discover, quote, hold, and handoff boundaries on a verified protocol endpoint, with a credible path to servicing and network expansion.","pledgeAmount":500,"status":"submitted","voteCount":4,"pledgedAmount":500,"createdAt":"2026-06-02T04:12:36.693Z","updatedAt":"2026-06-02T22:38:24.837Z"},{"id":"problem_0c73133c-2d3a-4bae-a4e4-0259d79aa359","title":"Shared Team Memory for AI Agents","category":"Product","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"A shared memory workspace where people and agents propose, accept, reject, revise, cite, and retire memories, so every agent on a project works from the same approved facts. AI can extract candidate memories from conversations, documents, code reviews, and tasks while human-approved governance decides what becomes durable team knowledge.","affectedAudience":"Startups, agencies, enterprise teams, software and consulting teams, and anyone running multiple agents across projects.","currentCost":"One agent knows the latest decision, another uses stale facts, and a third preserves a bad assumption. Context gets re-entered constantly and trust in multi-agent systems stays low.","desiredOutcome":"A demo using Markdown, JSON, Git, or Slack exports with provenance, versioning, confidence, role-based visibility, and a human approval step for durable memories.","pledgeAmount":500,"status":"submitted","voteCount":3,"pledgedAmount":500,"createdAt":"2026-06-02T22:23:41.562Z","updatedAt":"2026-06-02T22:38:14.996Z"},{"id":"problem_47e64ace-fa33-481d-ba94-284c97d04c64","title":"Conversational AI Operations Layer for Dance Studio Management","category":"Process","ownerName":"Maria Ivanova","ownerEmail":"maria@dfdancestudio.com","company":"DF Dance Studio","description":"Dance studio front-desk staff lose hours each week navigating a complex, bug-prone CRM (Mindbody) to add classes, check members in, update pricing, and pull reports. Staff need extensive training before working independently. Operational questions like 'How many people attended Wednesday's Latin class?' require manual report navigation. When staff turn over, training repeats from scratch. Bugs occasionally cause check-in errors or scheduling conflicts. An AI layer that staff can talk to in plain language — asking questions and taking guided actions, with human confirmation before any write — would cut overhead, reduce training costs, and make studio operations faster and more reliable.","affectedAudience":"Dance studio owners and front-desk staff managing class schedules, memberships, check-ins, and pricing through Mindbody or similar studio software.","currentCost":"Significant weekly staff hours lost to CRM navigation, bug workarounds, and recurring training when staff turn over.","desiredOutcome":"Staff can add classes, check members in, update pricing, and pull attendance reports through a conversational AI interface, with human confirmation before any write action.","pledgeAmount":500,"status":"submitted","voteCount":2,"pledgedAmount":500,"createdAt":"2026-05-26T13:51:27.764Z","updatedAt":"2026-05-26T17:49:10.774Z"},{"id":"problem_f98a8228-7e74-4822-ac3b-7e24d18e7f16","title":"AI Capability Navigator for Builders and Leaders","category":"People","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"A navigator that asks what someone is trying to accomplish, then recommends the frameworks, capabilities, tools, and learning path most relevant to that goal. AI can maintain a capability map, compare frameworks, explain tradeoffs by user type, and generate a practical plan for what to learn or build next.","affectedAudience":"Small-business owners, executives, students, builders, educators, product managers, and nontechnical professionals trying to keep up with AI.","currentCost":"People hear about MCP, A2A, agents, memory, RAG, evals, local models, and many frameworks, but cannot tell which matter for their goals. That leads to wasted time, wrong tool choices, and shallow adoption.","desiredOutcome":"A curated capability graph with rubric-based recommendations and links to official docs, scoped to one audience first, without claiming any one tool is universally best.","pledgeAmount":500,"status":"submitted","voteCount":0,"pledgedAmount":500,"createdAt":"2026-06-02T22:23:43.786Z","updatedAt":"2026-06-02T22:23:43.786Z"},{"id":"problem_b6a690b2-603f-4af0-a433-d6258c0a841f","title":"Permit Packet Pre-Check Agent","category":"Process","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"An agent that reviews a permit packet before submission, flags missing forms or conflicting information, and produces a clean revision checklist so applications pass first review. AI can extract requirements, compare uploaded documents against checklists, summarize deficiencies, and prepare a human-readable correction packet without approving permits.","affectedAudience":"Remodelers, builders, architects, engineers, permit applicants, property owners, and city or county reviewers who catch preventable errors.","currentCost":"Permit delays often start before formal review because applications are incomplete or inconsistent. Applicants find out weeks later, and reviewers spend time on errors that good pre-checks would catch.","desiredOutcome":"A document-AI pre-check demo for one permit type that improves packet quality, cites the source checklist for each flag, and keeps a human in the loop for all approvals.","pledgeAmount":0,"status":"submitted","voteCount":1,"pledgedAmount":0,"createdAt":"2026-06-02T22:23:39.309Z","updatedAt":"2026-06-02T22:37:55.452Z"},{"id":"problem_b7ac2063-6b44-4f4b-8210-a07f3b040e37","title":"Caregiver Coordination Copilot","category":"People","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"A copilot that turns scattered caregiving work into one shared plan: task lists, appointment summaries, medication and document reminders, family update drafts, and a short what-needs-attention-next briefing. AI can summarize scattered notes, pull tasks out of messages and documents, suggest next steps, and prepare privacy-safe updates for family or an employer conversation, without giving medical advice.","affectedAudience":"Working adults caring for aging parents, spouses, disabled relatives, or medically complex family members who become unpaid project managers on top of jobs and their own families.","currentCost":"Caregiving lives across texts, calendars, portals, PDFs, and memory. Handoffs between siblings and providers are weak, important details get lost, and the work drives burnout and missed appointments.","desiredOutcome":"A demo-ready caregiver copilot on calendar, email, document, checklist, and notification workflows with human review before anything is sent to family, employers, or providers.","pledgeAmount":0,"status":"submitted","voteCount":1,"pledgedAmount":0,"createdAt":"2026-06-02T22:23:37.803Z","updatedAt":"2026-06-02T22:37:55.638Z"},{"id":"problem_397c8c25-842f-4a72-9ace-98f61fe9672b","title":"AI Can't Branch a Conversation When Thinking Branches","category":"Process","ownerName":"Allen Ulbricht","ownerEmail":"allen@snowcapconsulting.com","company":"Snow Cap Consulting","description":"AI agents routinely surface multiple questions or angles at once, but conversation tools force a single linear thread. There is no way to explore one branch deeply without contaminating the others — and no shared data layer to keep parallel threads coherent.\n\nWho feels it: Knowledge workers, architects, researchers, and executives using AI for complex problem-solving — anyone whose AI regularly surfaces multiple questions simultaneously. Friction is mild for casual use and severe for sustained analytical work.\n\nToday's state: An AI working through a complex problem typically surfaces three to five sub-questions in one response. The user answers all of them in a numbered reply. If they want to go deeper on question three — challenge an assumption, explore a dependency, ask a follow-up — there is no structural mechanism. Pursuing it in-thread drags context from the other threads along. Opening a new session loses the parent. The conversation enforces linear structure on branching intellectual work. Users manage parallel thread state manually in their heads; the AI has no awareness that multiple concerns are in flight simultaneously.\n\nThis exposes a second problem beneath the UX one: data access across threads. When a branch forks, it needs more than isolated context — it needs decisions, constraints, and facts being established in parallel threads and in the parent. Without a shared data layer underneath all branches, forking just creates silos. A thread working on implementation cannot see that the architecture thread just changed a core constraint. When branches merge, reconciliation becomes a guess. This is also where threading and context relevance converge: the moment you fork, the question of what context is relevant to this specific branch becomes unavoidable — and current AI sessions have no mechanism to answer it.\n\nMarket signal: Branched conversation is among the most consistently requested features in AI tool communities. The numbered-list workaround is so universal it is a recognized pattern. No major consumer AI product has shipped native branching with a shared data layer.\n\nWhat AI shifts: A branch-aware system with a shared data layer lets users fork a sub-thread from any response. Each branch maintains focused context but can query shared state for decisions made elsewhere. When a branch resolves, findings merge back with full traceability.\n\nBuild readiness: Graph-structured conversation models are well-understood. Core challenges are UX (branch navigation, merge visualization) and context routing (what each branch inherits vs. queries on demand). The shared data layer is the harder architectural piece — retrieval, not just storage. Key failure mode to design against: branch proliferation.\n\nDemand signals: Requests for follow up on just this part appear regularly across Claude, ChatGPT, and coding assistant communities. Workarounds — separate sessions per sub-topic, manual thread summaries — are widespread.\n\nUnresolved question: When a deep branch merges back into the main thread, what is the right summarization boundary to avoid overwhelming the parent context?","affectedAudience":"Knowledge workers, architects, researchers, and executives using AI for complex problem-solving where the AI surfaces multiple questions simultaneously","desiredOutcome":"A branch-aware AI conversation system with a shared data layer, allowing users to fork sub-threads, explore them independently, query decisions from parallel threads, and merge findings back with full traceability.","pledgeAmount":0,"status":"submitted","voteCount":1,"pledgedAmount":0,"createdAt":"2026-05-28T07:02:21.413Z","updatedAt":"2026-05-28T12:12:45.217Z"},{"id":"problem_c2793d50-6717-4e7d-8843-048200b30d0a","title":"AI Sessions Have No Control Over Their Own Context","category":"Process","ownerName":"Allen Ulbricht","ownerEmail":"allen@snowcapconsulting.com","company":"Snow Cap Consulting","description":"Knowledge workers doing sustained AI-assisted work lose continuity and control as sessions grow. Context is managed by recency, not relevance — and users have no tools to change that.\n\nWho feels it: Systems architects, strategy consultants, senior engineers, and executives doing multi-hour AI-assisted planning or design work. Anyone whose session involves large reference documents and builds on decisions made earlier in the same conversation.\n\nToday's state: AI session context is managed by linear progression and lossy compaction, not relevance. As turns accumulate, content is retained or evicted based on recency — not on what is actually useful to the current question. Older material gets compacted into degraded summaries regardless of whether it contains a critical decision or key constraint. At turn 50 of a planning session, a pivotal decision from turn 5 may exist only as a lossy fragment, while recent but peripheral exchanges are preserved in full. No per-turn intelligence asks what does this question actually need from everything that has happened — the system applies uniform compression to whatever did not fit. Compounding this: users have no manual controls to intervene. No way to flag content as durable, purge irrelevant material, or compact a single loaded document without affecting everything around it. Context reflects conversational recency, not task relevance.\n\nMarket signal: Every major AI provider is racing to expand raw context window size — itself a signal that the problem is real. But larger windows delay relevance failure; they do not solve it. Extended AI sessions are becoming standard for high-value knowledge work, making context control a ceiling on session value.\n\nWhat AI shifts: A session-aware system could score relevance per turn and load only what the current question needs — moving evicted content to a retrievable store rather than discarding it. Users could manually flag durable content, purge stale material, or compact specific items on demand. Sessions become continuous rather than bounded by a compression policy.\n\nBuild readiness: Retrieval-augmented relevance scoring is mature technology. User control UI is a design challenge, not a research problem. Main open question is latency — per-turn scoring cannot add perceptible delay. Privacy matters: planning sessions often contain sensitive material; any persistent store needs local-first or encrypted options.\n\nDemand signals: Context window management is consistently cited as a top friction point in AI power-user communities. Workarounds — manual summarization, session restarts, pasted context blocks — are widespread and well-documented.\n\nUnresolved question: Can per-turn relevance scoring run within the main response pipeline without user-perceptible latency, or does it require a dedicated fast-model pass?","affectedAudience":"Systems architects, strategy consultants, senior engineers, and executives doing multi-hour AI-assisted planning or design work","desiredOutcome":"A session-aware AI system that dynamically loads relevant context per turn, moves evicted content to a retrievable store, and gives users manual controls to manage what the AI is working with.","pledgeAmount":0,"status":"submitted","voteCount":1,"pledgedAmount":0,"createdAt":"2026-05-28T07:01:40.151Z","updatedAt":"2026-05-28T12:13:52.144Z"},{"id":"problem_9780e277-3fc2-422e-b1a5-91f43e836e74","title":"Accounts Receivable and Cash-Flow Forecast Agent","category":"Profit","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"An agent that predicts which invoices are likely to be late, prioritizes follow-up, drafts customer reminders for approval, and turns invoice history into a short-term cash forecast. AI can combine invoice aging, customer behavior, email context, and payment history to recommend the next best collections action.","affectedAudience":"Small and medium businesses, freelancers, agencies, studios, contractors, bookkeepers, and fractional CFOs.","currentCost":"Many owners do not know which invoices will be late until cash is already tight. Collections live across accounting systems, email, spreadsheets, and memory.","desiredOutcome":"A demo using synthetic accounting data or QuickBooks-style exports with human-approved customer outreach and financial figures clearly labeled as estimates.","pledgeAmount":0,"status":"submitted","voteCount":0,"pledgedAmount":0,"createdAt":"2026-06-02T22:23:45.845Z","updatedAt":"2026-06-02T22:23:45.845Z"},{"id":"problem_33ee9837-1dc5-499d-9e21-f9fc6ef1b779","title":"Sponsor Problem Studio","category":"Process","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"A problem-shaping agent that interviews a sponsor, separates public from private context, identifies the affected audience, turns the problem into a public card, and suggests several solution pathways. AI can turn messy sponsor conversations into structured, buildable briefs while preserving human approval and privacy boundaries.","affectedAudience":"Event organizers, sponsors, chambers of commerce, local businesses, nonprofits, universities, and community partners.","currentCost":"Sponsors often have real problems but struggle to express them as buildable challenges. Too vague and builders cannot act. Too narrow and only one team can solve it.","desiredOutcome":"A demo built on the existing problem shaper and schema that converts sponsor input into public cards with public-private separation and human approval before publish.","pledgeAmount":0,"status":"submitted","voteCount":0,"pledgedAmount":0,"createdAt":"2026-06-02T22:23:44.801Z","updatedAt":"2026-06-02T22:23:44.801Z"},{"id":"problem_7ba41646-6bed-4dfa-8c5d-1846130579fc","title":"Vacant Lot Opportunity Agent","category":"Purpose","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"An agent that evaluates vacant or underused parcels and recommends realistic activation paths, including likely use cases, zoning constraints, required approvals, incentive options, and projected public benefit. AI can combine parcel records, zoning text, nearby business patterns, public meeting notes, and tax assumptions into a ranked set of possible projects for human review.","affectedAudience":"Cities, redevelopment agencies, property owners, chambers of commerce, brokers, developers, small businesses, and neighborhoods.","currentCost":"Opportunity is scattered across zoning rules, parcel data, ownership records, utility access, neighborhood needs, traffic patterns, incentives, and political constraints. Lots sit idle while corridors lose vitality.","desiredOutcome":"A civic-data demo using public parcel, zoning, and business data that recommends activation paths while a person reviews all legal, zoning, and financial assumptions.","pledgeAmount":0,"status":"submitted","voteCount":0,"pledgedAmount":0,"createdAt":"2026-06-02T22:23:42.541Z","updatedAt":"2026-06-02T22:23:42.541Z"},{"id":"problem_5376dd2f-0181-442e-b74d-8c29da285e5d","title":"Text-Only Appliance Repair Agent","category":"Product","ownerName":"Nick Baguley","ownerEmail":"baguleyllc@gmail.com","company":"Bad Labels","description":"A repair agent that works entirely over text message. It walks a consumer through basic diagnosis, collects model and photo details, checks warranty or manuals, finds likely parts or issues, and books a human technician when needed. AI turns a messy request into a structured repair workflow and keeps context so the consumer never starts over.","affectedAudience":"Everyday consumers, renters, homeowners, property managers, repair companies, warranty providers, and local technicians.","currentCost":"Most consumers will not install agent software or configure APIs. To get a washing machine fixed today, people search, self-diagnose, call around, compare, schedule, check warranty, and repeat the same details to everyone.","desiredOutcome":"An SMS-style demo that guides diagnosis, checks warranty or manuals, and hands off to a human technician with approval before any booking or spending.","pledgeAmount":0,"status":"submitted","voteCount":0,"pledgedAmount":0,"createdAt":"2026-06-02T22:23:40.588Z","updatedAt":"2026-06-02T22:23:40.588Z"}]}