September 2026 evidence-first software opportunities

This release adds 50 software opportunities designed for solo AI-assisted developers and small teams: 28 focused web SaaS products, 9 iOS apps, 9 developer reliability tools, and 4 platform-native apps. Every idea includes a complete implementation brief that can be copied into a coding agent.

What changed this month

• 104 candidates were ruled on exactly once by an editorial business review: 44 kept, 35 sent back for rework, 25 dropped. The published slate is the 44 keeps plus 6 rewritten concepts.
• Every idea passed the software-only builder-fit gate: no physical venue, custom hardware, proprietary-data dependency, marketplace-liquidity-first model, or required enterprise/government procurement.
• All 50 were checked against the existing library and against each other. Zero collisions.
• 1,497 web and App Store result rows were hand-labeled as direct competitor, adjacent, informational, or unrelated. No classifier or heuristic produced any label.
• Every idea carries an authored business model with concrete price points and a stated reason the buyer compares the price to recovered cost rather than to seat count.

ProvenScore overview

Every idea in this drop has a ProvenScore v3 on the same locked contract as the rest of the library: demand, opportunity gap, and builder feasibility at 45/30/25 weights, with low confidence shrinking unsupported extremes toward the midpoint rather than inflating them.

• Range: 50–64
• Median: 57; mean: 56.06
• Tiers: 0 Strong, 7 Promising, 43 Exploratory, 0 Weak
• Confidence: 0 High, 7 Medium, 43 Low
• Validation state: 42 Validated, 8 Needs validation

This drop skews exploratory, and that is the honest read rather than a disappointment. The evidence behind most of these niches is thin: keyword metrics returned no attributable row for many of the phrases, and a missing row means demand is unknown, not zero. The score reflects that uncertainty instead of papering over it.

"Validated" means the idea passed the current builder-fit and source-evidence contract. It does not mean the business is guaranteed to succeed. The score is a comparable research signal; confidence explains how much evidence supports that signal.

The 8 marked "needs validation"

8 ideas did not have every primary evidence source their type requires, so their score is shrunk hard toward the midpoint and they are labelled accordingly. Treat them as leads worth your own validation pass:

• Support Policy Regression Monitor
• Vendor Access Compliance Tracker
• Weather-Triggered Field Operations
• Flaky Test Governance
• Deployment Environment Drift Monitor
• Shopify Return Reason Coder
• Slack On-Call Handoff Digest
• HubSpot Stale Deal Hygiene

Highest-scoring opportunities

The 12 highest-scoring ideas in this release, with the tier and evidence confidence behind each number:

• Prescription Refill Photo Remainder — ProvenScore 64 · Promising tier · Medium confidence
  Focused iOS Apps. Photo the remaining pills, set a refill threshold, and remind before the pharmacy's lead time.

• ClickHouse TTL Dry-Run — ProvenScore 60 · Promising tier · Low confidence
  Developer Reliability Tools. Dry-run TTL expressions against a sampled partition and list columns and rows that would expire.

• Kid Medication Time Proof — ProvenScore 60 · Promising tier · Medium confidence
  Focused iOS Apps. Record each dose with time, photo of remaining medication, and a caregiver acknowledgement.

• Law Exhibit Number Registry — ProvenScore 60 · Promising tier · Low confidence
  Focused Workflow SaaS. Assign stable exhibit numbers, lock the registry at cutoff, and export a Bates-aware index.

• OpenTelemetry Cardinality Cap — ProvenScore 60 · Promising tier · Low confidence
  Developer Reliability Tools. Parse metric views and instrumentation for unbounded attributes and fail when a new high-cardinality key is introduced.

• S3 Lifecycle Dry-Run — ProvenScore 60 · Promising tier · Low confidence
  Developer Reliability Tools. Dry-run lifecycle documents against a sampled inventory and list objects that would expire in the next window.

• npm Provenance Attestation Checker — ProvenScore 60 · Promising tier · Low confidence
  Developer Reliability Tools. Check lockfile packages against required provenance attestations and fail when a new install lacks them.

• Flag Combination Collision Linter — ProvenScore 59 · Exploratory tier · Low confidence
  Developer Reliability Tools. Parse flag configs, compute colliding audience pairings, and fail CI when an untested combination is introduced.

• Hangboard Session Planner — ProvenScore 59 · Exploratory tier · Medium confidence
  Focused iOS Apps. Plan timed hangs with edge, added weight, and rest, then chart failure across those three dimensions.

• Plant Watering by Microclimate — ProvenScore 59 · Exploratory tier · Medium confidence
  Focused iOS Apps. Log light and room, then schedule watering from the last soil check instead of a generic species timer.

• Sourdough Starter Timeline — ProvenScore 59 · Exploratory tier · Medium confidence
  Focused iOS Apps. Log feed ratio, room temp, rise height, and aroma tags on a timeline that predicts the next bake window.

• Agency Change-Order Desk — ProvenScore 58 · Exploratory tier · Low confidence
  Focused Workflow SaaS. Capture every out-of-scope request, map it to the signed deliverable, and require a change-order before work continues.

All 50 ideas from this drop are in the product, each with its full ProvenScore breakdown, evidence notes, and complete implementation brief.

Evidence, limitations, and cost

September evidence collection spent $0.291080 through DataForSEO with zero paid retries, against a hard ceiling of $0.50. Two organic SERP tasks returned provider errors, were charged, and were deliberately not retried; those queries remain unknown rather than being backfilled with a guess.

Keyword Overview charges per returned item, not per requested phrase. 296 phrases were requested and 19 returned an attributable row. The remaining phrases have unknown volume. They are not zero-volume phrases, and nothing in this drop treats them as such.

App Store ranks, ratings, reviews, and search results are directional market signals. They are not exact market size and not guaranteed revenue.