SB StartupBasket

About

A research pipeline,
not a creative-writing exercise.

Most AI idea generators remix what already exists or hallucinate plausible-sounding concepts. Both produce slop. StartupBasket runs every idea through a 6-stage pipeline grounded in real-world signals, then scores it honestly against a 4-layer rubric. The goal: surface ideas a small team could realistically ship and scale to $1M–$5M ARR.

The persona behind every idea

An experienced tech entrepreneur — fifteen-plus years shipping software, two prior exits — currently angel-investing while hunting for the next thing to build. Direct, opinionated, slightly impatient, allergic to fluff. Every idea is written in that voice, filtered through that taste.

Rejects

  • — Unicorns, billion-dollar TAMs, "change the world" pitches
  • — Ideas needing $10M+ before first revenue
  • — Regulated markets unless regulation is the moat
  • — Crypto / NFT / web3 revivals
  • — "Uber for X" without a genuine insight

Hunts for

  • — Boring markets where incumbents have bad UX
  • — Workflows where AI collapses 2 hours into 2 minutes
  • — Tools professionals will pay for this week
  • — Localized plays — India, SEA, LatAm, MENA — and global SMB SaaS where the wedge is real
  • — Micro-SaaS too small for VCs, perfect for a bootstrapper

The 6-stage pipeline

Signals → Triangulation → Shape → Filter → Research → Score → Publish. Most candidates die between stage 2 and stage 4. That's the point — filtering before generation is what keeps the bar honest.

  1. 01

    Signal harvesting

    5–10 raw signals per run from complaint mines, demand trackers, regulatory shifts, technology unlocks, and geographic arbitrage. Each signal gets a source URL and a date.

  2. 02

    Triangulation

    An idea candidate requires three independent signals: one demand, one feasibility, one economic. Single-signal ideas are noise. They get discarded.

  3. 03

    Shape

    Answer six concrete questions — who exactly, what pain how often, what do they pay today, what the product literally does, why now, why not already. If any answer is fuzzy, the idea isn't ready.

  4. 04

    Audience filter

    Kill filter against the builder profile: no $50K+ capex, no 6-month enterprise sales, no network-effect chicken-and-egg, no regulatory approval gates. 60–70% of candidates die here. That's correct.

  5. 05

    Deep research

    Market map, customer voice (5–10 verbatim quotes), unit economics, distribution playbook with conversion math, competitive moat. Zero tolerance for fabricated stats.

  6. 06

    Score & publish

    4-layer rubric: gating → 7-axis scoring → qualitative modifiers → structured verdict with kill criteria. Provenance block trails every idea back to its source signals.

The 3-signal rule

A single complaint on Reddit is noise. Three independent signals pointing at the same pain are an opportunity. Every published idea carries a Provenance block naming the three signals — one demand, one feasibility, one economic — with source URLs and dates. If a claim can't be sourced, it doesn't make it in.

Provenance:
Signal 1 (demand): 42% of STR operators spend 51–100 hrs/yr on lodging-tax compliance — avalara.com/... — 2026-02
Signal 2 (feasibility): California SB 346 data-sharing law live Jan 2026 — bbklaw.com/... — 2026-01
Signal 3 (economic): Avalara MyLodgeTax priced $27/mo/property + $299 setup — strspecialist.com/... — 2026-02
Category: Regulatory arbitrage

The 4-layer rubric

Survivors of the pipeline get scored. No single number is enough — each idea passes through four layers.

  1. Layer 1

    Gating criteria — pass/fail

    Legal in target markets · ethical floor · market evidence · buildable by 1–5 people · launchable under $50K. Fail any one, the idea is rejected before scoring begins.

  2. Layer 2

    7-axis weighted scoring — 100 points

    Each axis has explicit scoring bands. A 78 isn't a vibe — it's 17+13+10+12+12+8+6 with notes on each one.

  3. Layer 3

    Qualitative modifiers

    Founder-fit tags (technical-heavy, sales-heavy, domain-expertise-required, etc.), 3–5 assumptions with specific validation methods, and risk flags the score doesn't capture.

  4. Layer 4

    Structured verdict + kill criteria

    Score, verdict, confidence, best-fit builder, time to revenue, capital to launch, top three assumptions to validate first, and — most importantly — measurable kill criteria. When to quit matters more than when to start.

The seven axes

AxisWeightWhat it measures
Problem intensity 20 How painful is this problem? How often is it felt? Would users pay to solve it today?
Demand evidence 15 Hard signals — existing spend, competitor traction, search volume, forum complaints. Not hunches.
Build feasibility 15 A small team can ship v1 in 90 days with available tools and standard infra.
Distribution clarity 15 A concrete, cheap path to first 100 customers. Named channels, named lists, realistic conversion math.
Revenue mechanics 15 Clear pricing, willingness to pay, unit economics that work below $5M ARR.
Time to first revenue 10 Weeks to first paying customer. For bootstrappers, this matters more than TAM.
Defensibility 10 Soft moats — workflow lock-in, data advantage, community, regulatory knowledge.

Verdict thresholds

80+ = STRONG GO · 70–79 = GO · 55–69 = VALIDATE · below 55 = PASS. PASSes stay in the catalog as negative examples. An honest 68 is more useful than an inflated 82 — if every idea scored 80+, the rubric would mean nothing.

Portfolio balance, on purpose

Left alone, an idea generator drifts — five India SaaS copilots in a row, ten DevTools plays, every idea a B2B workflow. Before every run, the pipeline audits the last 15–20 published ideas and deliberately targets an under-represented category or geography. Category targets:

Tech-unlock

~25%

Became possible in the last 6–12 months — new model capabilities, cheaper inference, new APIs, new hardware.

Geographic arbitrage

~20%

Proven in one market, absent in another. US SMB tools missing in India; Indian patterns useful in LatAm/SEA.

Underserved niches

~20%

Markets too small for a VC fund but perfect for a focused bootstrapper.

Regulatory arbitrage

~15%

New law creates new compliance needs — DPDP, EU AI Act, SB 346, GST shifts.

Workflow automation

~15%

Manual work in $1B+ industries that AI can collapse from hours to minutes.

Platform shifts

~5%

New platforms create new app categories — WhatsApp-as-OS, UPI ubiquity, ESP32 generation.

What this is not

Start exploring

Pick an idea. Steal it. Ship it.