The graveyard

Ideas our own engine told us not to build

These went through the full pipeline — clustered from real signals, written up, reviewed by a human, published. Then the verdict engine read the counter-evidence and returned kill. They are still here, with their sources and their reasoning intact.

We publish them because a validation tool that only ever says yes is a marketing tool. If you cannot see what we reject, our approvals are worth nothing.

7 killed ideas, newest first. Each one is readable in full — the research does not stop being useful because the answer was no.

Bot-traffic triage dashboard for indie site operators using Cloudflare

A lightweight log-analysis layer that classifies bot vs human traffic and flags scrapers Cloudflare's default rules miss, for solo site owners.

Kill it· low

What killed it

  • An incumbent can ship thisCloudflare already provides bot scoring and detection for customers sitting in the request path, and the stated cause of death explicitly acknowledges this incumbent advantage. The product would need defensibility from features Cloudflare won't build, not from dashboards on signals Cloudflare already surfaces.
  • The evidence is too thin to judgeThe signal cluster shows operators complaining about bot traffic problems, but none express willingness to pay for a standalone solution or mention trying paid tools that failed. The HN complaints represent awareness, not purchasing intent.
  • A free alternative already covers itThe core pain is resource consumption and blocking bad actors, not classification or dashboards. A dashboard that identifies scrapers without actually blocking them or reducing costs leaves the fundamental problem unsolved, while Cloudflare can both detect and mitigate in one step.

Unified web-parity mobile client for SimplePractice/Jane clinics

A mobile booking app for small clinics that mirrors the clinic's web portal exactly — same login, same appointment history, same payment info — for practices running SimplePractice or Jane.

Kill it· low

What killed it

  • An incumbent can ship thisSimplePractice and Jane both have active products and direct API access to the clinic data, payment rails, and patient records needed to ship the requested features. A third-party cannot offer 'same login' or 'credit cards on file' without those companies granting integration access—which they have no incentive to do if mobile parity becomes table stakes.
  • A regulatory barrierThe stated wedge requires clinic software vendors to expose patient login credentials, payment methods, and appointment history via API to a competitor. No evidence shows SimplePractice or Jane offering this level of access to third parties, and HIPAA compliance would require the clinic to own the patient relationship—eliminating the 'unified login' that differentiates this from yet another booking layer.
  • The evidence is too thin to judgeAll three signals come from app store reviews in a 9-month window (Oct 2025–Jun 2026), and all reference existing clinic software apps that are actively maintained. If mobile parity were a structural blocker, we'd see sustained complaint volume across years; instead this reads like a feature gap the incumbents are already working to close.

Third-party calibration app to suppress false theft-lock triggers during running (Android)

A companion app for Android runners that detects a 'run session' (via connected earbuds/watch or manual toggle) and temporarily suppresses Google's theft-detection lock so it stops mistaking running for a phone snatch.

Kill it· low

What killed it

  • The evidence is too thin to judgeAll seven signals are from a single 48-hour window (Aug 6-8, 2026) on one platform (HackerNews), likely representing one viral thread being quoted rather than independent demand. No follow-up complaints exist in the data to suggest this persists beyond the initial discussion.
  • A regulatory barrierAndroid's security model increasingly restricts background access to motion sensors and lockscreen control, making a third-party app unable to reliably detect 'run mode' or suppress theft alerts without requiring intrusive permissions that would block Play Store distribution.
  • An incumbent can ship thisGoogle already owns the theft-detection feature being complained about and can ship a native workout exception or sensitivity toggle in a single OS update, eliminating the problem before a third-party solution gains distribution. The stated cause of death explicitly predicts this.

Escalation-to-human triage add-on for SMB SaaS AI support widgets

A drop-in escalation layer that SMB software vendors plug into their AI chatbot so frustrated users can reach a real human, sold to the vendor not the end user.

Kill it· low

What killed it

  • The evidence is too thin to judgeAll complaint signals come from end users, not vendors. There is zero evidence that SMB SaaS companies are searching for, spending on, or expressing intent to purchase escalation tooling. The market evidence shows pain exists but not that the stated buyer — the software vendor — recognizes it as a problem worth outsourcing.
  • The unit economics do not closeWave — the primary example in the evidence — is an active product whose users are complaining about *lack* of human escalation, yet Wave has not purchased this solution. If the need is acute enough to generate multiple complaint signals, but the vendor experiencing the complaints has not bought or built an obvious fix, this suggests vendors do not prioritize solving this problem at a price point a third party could charge.
  • An incumbent can ship thisThe only spend signal in the evidence is for Telegram-specific support triage with automatic handoff, indicating that buyers want platform-native solutions integrated into their existing stack, not a generic drop-in layer. This reinforces the stated cause of death: vendors will build this themselves as a routing rule rather than pay for a third-party add-on.

Skill-sync CLI for Claude Code / Codex agent configs across repos

A CLI and background sync daemon that keeps Claude Code/Codex agent skills, prompts, and settings identical across a developer's repos, flagging drift before it causes side effects.

Kill it· medium

What killed it

  • An incumbent can ship thisAnthropic and OpenAI own the config schemas and have direct retention incentive to ship native workspace-level sync. The stated cause of death identifies this as the most likely outcome within a release cycle, and no evidence shows demand for agent-agnostic sync that would differentiate from the native solution.
  • A free alternative already covers itGit submodules or symlinks already solve config sync across repos for free, and developers coordinating multi-agent work are explicitly managing git worktrees and tracking context per project. A paid CLI that only syncs configs adds no value over existing version control primitives.
  • The evidence is too thin to judgeOnly one signal directly describes the core problem of config drift across repos. The rest describe multi-agent coordination, testing infrastructure, or environment setup — adjacent problems but not validation that developers will pay specifically for config sync tooling.

AI usage governance dashboard for engineering teams using Claude Code/Cursor/Copilot

A visibility and adoption dashboard that lets engineering managers see how their team actually uses AI coding tools and where shadow-tool sprawl or data leakage is happening, for SMB engineering orgs.

Kill it· low

What killed it

  • An incumbent can ship thisMicrosoft already owns the natural distribution channel (M365 admin console) and the primary AI coding tool in this market (Copilot), making a third-party dashboard structurally disadvantaged. Organizations are already defaulting to built-in Microsoft options over more capable standalone tools, and the cluster shows no signals of buyers specifically seeking cross-vendor dashboards.
  • The unit economics do not closeThe target buyers are explicitly cost-sensitive about incremental AI seat licenses, yet this product asks them to pay for a separate seat-based dashboard on top of the AI tools themselves. The evidence shows organizations questioning whether to buy Copilot for employees who won't use it fully, not willingness to add another layer of tooling cost.

Offline-first time clock and job schedule for rural field service crews

A mobile time-tracking and scheduling app for field service technicians that fully functions without cell signal, syncing entries when connectivity returns.

Kill it· low

What killed it

  • An incumbent can ship thisOffline sync is a standard mobile engineering pattern, not a defensible technology moat. The incumbent already has the harder problems solved—payroll integration, dispatch workflow, customer data—and adding offline mode is a feature sprint, not a platform rebuild.
  • An incumbent can ship thisThe complaints clearly identify Housecall Pro as the incumbent, and these are paying customers trapped by switching costs who explicitly say 'switching to another platform is not a simple process once you're invested'. They are complaining, not churning, which means the pain threshold for departure is high.
  • The evidence is too thin to judgeThe signal set is extremely thin for validating a standalone business: 9 complaints over 14 months, all from app stores, with no evidence of search volume, willingness to pay for a solution, or businesses actively seeking alternatives. One mention of operational losses is immediately followed by acknowledgment that switching is too hard.

How an idea ends up here

A kill needs a decisive risk — severity 5 on the published scale, or two separate risks at severity 4 — and enough evidence to see it. An idea with no data does not get killed; it gets insufficient evidence, because a kill is a claim too and we hold it to the same bar as a recommendation.

Risks that count: an incumbent can ship it as a feature, a free alternative already covers it, the unit economics do not close, a regulatory barrier, or products in the space have already died. Thin evidence and shrinking demand are recorded but never kill on their own — one is missing data, the other is a timing problem.

Revenue figures cited in the evidence carry one of 6 labels for how they were obtained, from payment-verified down to third-party estimate. Only the top tier counts as proof somebody pays.