The Forum
“Who has this problem?” as a structured quest — the problem, must-have criteria, and how the poster knows it's real. Builders claim the gaps no existing product fills.
Quests written by a person — posted through the Forum or seeded by the founder with a cited source. Nothing here was drafted by a model.
Showing 12 quests
In a 3–20-staff Indonesian service business — villa management, tour operations, contracting, catering, agencies — the staff pay for things and the receipts arrive as photos in a WhatsApp chat. At month end the owner has to hand something an external bookkeeper will accept, and the gap between those two states is manual work somebody pays for: fastwork.id's bookkeeping category is a live marketplace of freelancers selling exactly that transcription from Rp60.000, which is about as direct a demand signal as an unserved workflow produces. The tools around it miss on both sides. BukuWarung owns the micro segment but has become a payments company whose bookkeeping is simple cash-in/cash-out, and its users mostly do not have this problem — under PPh Final 0.5% a micro business owes simple revenue pencatatan, not expense documentation. Kledo, Mekari Jurnal and Accurate Online are full web-first accounting suites aimed at larger SMEs, and Mekari already ships AI OCR for receipts, so OCR alone settles nothing. Every serious incumbent is Indonesian-first or Indonesian-only, so an owner who works in English with staff who work in Indonesian is served by nobody. The open question: is the thing worth paying for the capture (getting the receipt out of WhatsApp), or the hand-off (a record the bookkeeper accepts without re-keying)?
3 must-haves · 1 bonus · 0 votes
There is a specific person who writes down what they shipped every day — a WIP todo, a commit, a changelog line — and whose X or LinkedIn account has been silent for months. They already do the input behaviour as a habit; they have failed at the output behaviour repeatedly. WIP.co is the densest visible concentration of them: its entire core mechanic is a daily shipped-todo streak, and its front page on 2026-08-08 showed same-day todos from real makers. The ~264K-member Build in Public X community is the same population stating the aspiration out loud. Six-plus tools now pipe git logs straight to social — BuildInPublic.so (alive, growing, $9/$19/$49), oidapost (10+ platforms, scheduled auto-posting), Forg, and newer entrants gittomarket.io, shiptopost.com, git-to-x.com — and all of them optimise for volume. The unanswered question is what the quiet shipper is actually blocked on: is it the writing, or is it the fear of publishing something that sounds like a machine wrote it? If it is the second, then a tool that refuses to auto-post and forces you to approve every line is the feature. If it is the first, that same gate is just friction someone will route around. Nobody has tested which.
3 must-haves · 1 bonus · 0 votes
People manage notification overload one app at a time: muting, unsubscribing, then reinstalling and resubscribing months later because something important got buried. On a 78-comment Ask HN thread, the original poster missed a close friend's birthday reminder in the noise; commenters' workarounds were all per-app (OS-level limits, pull-not-push email, disabling everything except a manual allowlist) — nobody named a tool that curates across apps instead of within one.
2 must-haves · 1 bonus · 0 votes
Users of AI coding tools hit usage limits with no visibility into what consumed the budget — model choice, reasoning effort, or context bloat. The most-reacted issue on the topic has 305 reactions; a related request for programmatic quota access has 114.
2 must-haves · 0 votes
Long-running and multi-session AI coding agents start each session knowing nothing about what came before. Teams lose the reasoning behind past decisions, and the agent silently re-derives or contradicts them. Surfaced independently on Hacker News and in three separate GitHub issue threads, including one that consolidates five other community requests.
2 must-haves · 0 votes
Sites and apps ship fuzzy-only search with no way to force an exact match. Commenters independently reported the same failure across several major products, one estimating most of their daily searches would be better served by exact matching.
2 must-haves · 0 votes
Developers who need local or offline models for privacy or client-confidentiality reasons are shut out of tools that hard-require cloud AI and a login. The request has been open since 2024 with 1,424 reactions.
2 must-haves · 0 votes
Agentic coding replaces sustained focus with a wait-and-review cycle; developers report they no longer work deeply and cope with ad hoc workarounds rather than any tool built for running several agent sessions attentively.
2 must-haves · 0 votes
Repeatedly asked in public by founders who have already built something: the build is done and no repeatable way to reach the first users exists. Four independent posts in a single 240-post research window described this same gap.
2 must-haves · 1 vote
Agent SDKs expose raw token usage but leave every team to maintain its own pricing tables and to ignore non-LLM tool, database and retrieval calls — so the true cost of an agent run is reconstructed by hand, differently, in every project.
2 must-haves · 0 votes
Agents avoid rewriting existing code (it costs context) and append near-identical functions instead, so codebases accumulate duplicate and dead implementations. Multiple developers in the same thread posted their own examples; no tool was named as detecting it.
2 must-haves · 0 votes
Every discussion of local models for daily coding is a pile of unreproducible anecdotes — quantization, VRAM, context size and agent harness are almost never stated together, so nobody can replicate a reported result.
2 must-haves · 0 votes