Crypto Accumulation Research

Long-horizon plan · verified bounties · scam avoidance

Triage of the demand-signal list against the earning goal

Source: ~180 "unmet need" rows scraped from Reddit, Hacker News and Indie Hackers (Aug 23-25, 2026), supplied in chat on 2 Sep 2026 as a table with columns date, platform, post title, unmet need, niche concept, monetization, source URL. Rows are referenced below by their post titles.

The goal is crypto earned through skill-based work. The list is validated demand for products, which is a different thing: a product earns fiat over months after distribution and support, not stablecoins in weeks. So the useful question is narrower than "which idea is best": which rows can be shipped inside the open bounty windows, get paid in crypto, and keep earning afterwards.

What actually connects to the pipeline

1. Agentic-payment bounties want exactly one shape: an API agents pay per call

Three ETHOnline tracks (Sep 4-16) and the KeeperHub main track reward a service that an autonomous agent can discover, pay for in USDC, and consume:

Track Prize What it asks for
Hedera "AI & Agentic Payments" $2,000 x 3 A working x402 service
Bazantic "Agentify a new API" $500 / $300 / $200 Add a new API service to Bazantic's x402/MPP gateway and a reusable recipe
Arc "Best Agentic Economy Application" $1,667 Agents that pay/settle in USDC, ideally via Circle Agent Stack / Nanopayments
KeeperHub main track $2,000 / $1,200 / $800 KeeperHub as execution layer inside a live project, value moving through it

A per-call API from the list, priced in USDC over x402, is one build that qualifies for several of these and is a real product afterwards. That is the only place the list and the goal overlap directly.

Rows that fit "an agent pays per call" (developer-facing, deterministic or verifiable output, stateless request/response, no consumer distribution):

Row Unit of value Verifiability Effort Verdict
Convert PDF bank statements into CSV reliably (Ask HN) per page Arithmetic: running balance must reconcile row to row, so wrong extractions are detectable Medium-high (text PDFs; OCR out of scope) Build. Highest willingness to pay on the list; bookkeeping agents are a real buyer; reconciliation makes the output trustworthy without trusting a model
Strict-spec document/photo pre-flight resizer per image Deterministic (dimensions, bytes, DPI) Low Second endpoint if time allows
HAR file PII/auth-token sanitizer per file Deterministic Low Cheap add-on, low value
Breaking API surface gatekeeper (semver) per check Deterministic per language Medium Later; language-specific
PR-diff-aware CI test selector per PR Hard to verify High No
Cost-optimised LLM router per request Quality unverifiable, needs upstream keys Medium No; OpenRouter exists
Multi-OS untrusted code sandbox API per second Infra-heavy Very high No

Decision: the ETHOnline entry is a paid API whose headline endpoint turns text-based bank statement PDFs into reconciled CSV, priced per page in USDC over x402. Design in ../x402-api/DESIGN.md. Per ETHGlobal's rules nothing project-specific is written before Sep 4; the design is the only artifact until then.

2. Rows that validate the CreditPassport thesis (use in the deck, do not build)

Several rows ask for portable, verified history without a trusted middleman, which is what CreditPassport does with Attestcoin proofs:

These go on the "market pull" slide as evidence the problem is felt, with their URLs. None is built now.

3. Rows that are not a crypto-earning path in any open window

Roughly 150 of the rows are consumer or SMB apps (ADHD tools, screen-time lockers, photo journals, meetup apps, household trackers, macOS utilities). They need app-store distribution, marketing and support, they earn fiat, and the first dollar is months away. Some are good businesses; none is an answer to "gain crypto assets," and building any of them now would displace work that pays in stablecoins in September. They are parked, not rejected.

Two rows are close to what the parallel agent runs already built here: "Covered call backtester with historical IV surfaces" and "Options income backtester" resemble the Ground Truth backtester at the repo root, but with a paid-data dependency (options surfaces) that makes them a fiat SaaS, not a bounty.

What this changes in the pipeline