en

New ideaFintech

Open-source "trader skills" pack

Free PT/ES skill pack that lets Claude or Codex scan B3 and BMV charts, read CVM filings and local news, and replay setups — the day-trader-skill pattern for LatAm.

2 weeks ago · Scored 26 Oct 2026 · ID HBV-261026-03

Working title: Open-source "trader skills" pack for Brazilian (B3) and Mexican (BMV) retail traders, in Portuguese and Spanish (working name "Pregão Skills")

Summary

BriefingVerifiedHypothesis

New HBV GitHub repo idea (POP 5.6). Benchmark: Jakeschincariol/day-trader-skill (accelerating comparable) — Total: 1.8k (github.com page text, read 26 Oct 2026 07:06 BRT; page shows no exact count); API 1,816 (OSS Insight/gh meta 06:57 BRT) and 1,824 (second gh read) — live growth.…

Read the rest

Created 2026-10-20 (gh API cross-check only; not visible on the repo page), MIT (browser). 24h (last full UTC day, 25 Oct): +519 stars. 7d (19–25 Oct): +1,674 = 239/day over 7 days (repo is 6 days old, so ≈279/day over its real life). Since creation: ~1.8k in 6 days ≈ 303/day. Acceleration: 7v7 n/a (repo younger than 7 days; prior 7 days = 0, not computable); 3v3 (23–25 Oct 1,136 vs 20–22 Oct 538) = +111% (accelerating); peak day 25 Oct (+519, the last full day) → not past peak. Daily series: 20 Oct 37 · 21 Oct 216 · 22 Oct 285 · 23 Oct 347 · 24 Oct 270 · 25 Oct 519 · 26 Oct partial 223 (OSS Insight analyze-recent-stars, fetched 26 Oct 06:57 BRT). Not on github.com/trending daily, weekly, python weekly or typescript weekly at 07:06 BRT (weekly #5 cutoff ≈2,790 stars/week). All star numbers Verified (browser total; API/OSS Insight series). Sibling Jakeschincariol/trader-skill: 650 stars (browser, 07:06 BRT), created 14 Oct 2026 (API), MIT; 24h +84; 7d +586 (7v7 +1,174% vs 46); 3v3 −24% (252 vs 331); peak 21 Oct (+169) → past peak (OSS Insight, 07:08 BRT). Existence check: gh search 26 Oct ~07:10 BRT: "B3 ibovespa skill OR claude OR agent" → 0 repos; "trading skill claude created:>2026-09-26" → no B3/LatAm repo (Verified). gap Verified (gh search "B3 ibovespa skill OR claude OR agent" → 0 repos; "trading skill claude created:>2026-09-26" → no B3/LatAm repo, 26 Oct ~07:10 BRT). Below the bar because the LatAm demand is small and noisy (PT query 376 posts/7d, rising +19% 3v3, Verified), monetization (broker affiliate/sponsorship) is Hypothesis only, the sibling comparable trader-skill is decelerating (−24% 3v3, past peak 21 Oct) and CVM rules on recommendations add compliance load. POS 62 ≤70. Stars: Comparable repos (no own repo yet): Total: 1.8k (github.com page text, read 26 Oct 2026 07:06 BRT; page shows no exact count); API 1,816 (OSS Insight/gh meta 06:57 BRT) and 1,824 (second gh read) — live growth. Created 2026-10-20 (gh API cross-check only; not visible on the repo page), MIT (browser). 24h (last full UTC day, 25 Oct): +519 stars. 7d (19–25 Oct): +1,674 = 239/day over 7 days (repo is 6 days old, so ≈279/day over its real life). Since creation: ~1.8k in 6 days ≈ 303/day. Acceleration: 7v7 n/a (repo younger than 7 days; prior 7 days = 0, not computable); 3v3 (23–25 Oct 1,136 vs 20–22 Oct 538) = +111% (accelerating); peak day 25 Oct (+519, the last full day) → not past peak. Daily series: 20 Oct 37 · 21 Oct 216 · 22 Oct 285 · 23 Oct 347 · 24 Oct 270 · 25 Oct 519 · 26 Oct partial 223 (OSS Insight analyze-recent-stars, fetched 26 Oct 06:57 BRT). Not on github.com/trending daily, weekly, python weekly or typescript weekly at 07:06 BRT (weekly #5 cutoff ≈2,790 stars/week). All star numbers Verified (browser total; API/OSS Insight series). X: 376 posts/7d (Rising +19% 3v3 (23–25 Oct 166 vs 20–22 Oct 139; daily full UTC days 51, 44, 44, 48, 57, 61); peak 25 Oct 61; 25 Oct is 0% below peak; window 2026-10-19 10:30 UTC → 2026-10-26 09:50 UTC (X API recent counts, granularity day, pulled 26 Oct 2026 ~06:55–07:15 BRT; Sun 19 and Mon 26 are partial UTC days; full days Mon 20 → Sun 25; 3v3 = (23+24+25) vs (20+21+22))). Revenue hypothesis: ~$1K–$8K/month at month 6 (HYPOTHESIS; affiliate terms not checked).

DescriptionHypothesis

Free PT/ES skill pack that lets Claude or Codex scan B3 and BMV charts, read CVM filings and local news, and replay setups — the day-trader-skill pattern for LatAm. Who: Brazilian and Mexican retail traders and investors who use ChatGPT/Claude and want tools that understand B3/BMV tickers, CVM filings and local news in their language.…

Read the rest

Pain: The fast-rising trader skill packs are US-only (US tickers, SEC filings, English news); nothing open-source covers B3/BMV data, CVM/B3 filings or PT/ES news. Monetization: Open-source core + hosted Pro tier (alerts, replays) + broker affiliate/CPA programs (XP, BTG, Rico, Clear) and labelled sponsorship slots (Hypothesis). Build: ~2–4 agent-weeks (skills, free data adapters for Yahoo.SA / B3 / CVM, PT/ES prompts, replay); stack: Claude/Codex skill files (MIT); Python data adapters (Yahoo.SA, B3/CVM public filings); optional Next.js Pro tier on Vercel; Stripe/Pix. GTM: GitHub launch + PT/ES README, SEO/GEO pages per ticker setup, faceless X/TikTok clips; seed in PT/ES trading communities.

Target customer

Brazilian and Mexican retail traders and investors who use ChatGPT/Claude and want tools that understand B3/BMV tickers, CVM filings and local news in their language.

Pain

The fast-rising trader skill packs are US-only (US tickers, SEC filings, English news); nothing open-source covers B3/BMV data, CVM/B3 filings or PT/ES news.

Why this score

Why it scored lower: Small, noisy LatAm demand signal and Hypothesis-only monetization; CVM rules on recommendations.

MoatHypothesis

Moderate (local data + language): First PT/ES pack with B3/BMV/CVM adapters; GitHub stars become distribution. Local data adapters, community contributions and SEO in Portuguese and Spanish. Status: Hypothesis.

Demand signals

X posts (7 days)

376

Signal trend

Rising +19% 3v3 (23–25 Oct 166 vs 20–22 Oct 139; daily full UTC days 51, 44, 44, 48, 57, 61); peak 25 Oct 61; 25 Oct is 0% below peak

Web demandVerifiedHypothesis

No B3/BMV skill repo exists (gh, Verified). The PT query is noisy ("opções" also matches non-trading posts). Brazilian retail-investor base size not fetched this run (Hypothesis that it is large). Other X counts (same window, Verified X API): 434 posts (Falling -55% 3v3 (23–25 Oct 108 vs 20–22 Oct 239; daily full UTC d…

Read the rest

ays 73, 81, 85, 61, 28, 19); peak 22 Oct 85; 25 Oct is 78% below peak — ≥50% below peak: POP 5.3 window-closed signal): ("day trading" OR daytrading) (AI OR Claude OR ChatGPT) -is:retweet

Market and competitors

TAM

Not sized this run — Scope: Global. Method: No third-party report fetched. (Basis: n/a)

SAMHypothesis

Not sized — Scope: Target markets above. Method: HBV assumption. (Basis: Hypothesis)

SOMHypothesis

~$12K–$96K ARR (Hypothesis) — Scope: Month-6 run-rate. Method: From revenue hypothesis ×12. (Basis: Hypothesis)

Competitors

Status Invest, Investidor10, TradingView (B3 data), broker research apps (XP, BTG, Rico, Clear), Telegram signal groups; US skill packs (no B3 coverage).

Business model

MonetizationHypothesis

Open-source core + hosted Pro tier (alerts, replays) + broker affiliate/CPA programs (XP, BTG, Rico, Clear) and labelled sponsorship slots (Hypothesis).

Revenue hypothesis (monthly)Hypothesis

~$1K–$8K/month at month 6 (HYPOTHESIS; affiliate terms not checked).

Jobs to be done

When I prepare my B3 trading day, I want to have my AI read B3 charts, CVM filings and local news in Portuguese, so I can decide faster without paying for signal groups.

HXC

Rafael, 29, mini-índice day trader in Recife (illustrative) · Trades WIN futures at lunch; uses ChatGPT with screenshots. Buying signal: Asks in Telegram groups for "robô" setups; follows US AI-trading repos.

Build and go-to-market

Tech stack

Claude/Codex skill files (MIT); Python data adapters (Yahoo.SA, B3/CVM public filings); optional Next.js Pro tier on Vercel; Stripe/Pix

GTM channels

GitHub launch + PT/ES README, SEO/GEO pages per ticker setup, faceless X/TikTok clips; seed in PT/ES trading communities

Risks

Main risk

Small, noisy LatAm demand signal and Hypothesis-only monetization; CVM rules on recommendations.

All risksVerifiedHypothesis

Value: Attention — PT query 376/7d, noisy (Verified) | Usability: Supported — Skill UX copied from an accelerating repo | Feasibility: Supported — Small build, free data | Business Viability: High — Revenue Hypothesis; CVM rules

Next validation step

Re-measure the PT/ES query with a cleaner phrasing on 2 Nov 2026; check broker affiliate terms; watch whether day-trader-skill keeps accelerating (7v7 on 28 Oct).

Sources

  1. gh search (26 Oct ~07:10 BRT)
  2. github.com
  3. github.com (browser 07:06 BRT)
  4. OSS Insight (06:57 and 07:08 BRT)
  5. X API counts (window above)
  6. sheet row 104 (HBV-261024-04) for dedupe

ideagithub-newfintechtradingopen-sourceai-agentsbrazilmexicolatamglobal

Want the next one the day it is scored?

The score, the sources, the risks and the next step, by email.

Free. Sent with beehiiv. One click to unsubscribe.

  • New ideaFintech· 2 weeks ago

    Hosted "day-trader skills" desk

    A web app that runs the open-source day-trader skills for you — scan charts, track news, time entries and replay outcomes before you risk a dollar — no terminal needed.

    Why it scored lower: Financial-advice/regulatory exposure and market-data licensing cost; the MIT code lets anyone host the same thing.

    Open ProductBots.ai →

  • New ideaFintech· 1 week ago

    Agent-native checkout

    Drop-in agent checkout and prepaid credits wallet for solo SaaS: MCP tool + embed so coding agents can buy credits or unlock features; merchant dashboard with card and LatAm PIX.

    Why it scored lower: Incumbent rails ship the same wedge; regulated payments support load.

    Open ProductBots.ai →

  • Existing businessFintech· 3 weeks ago

    AgentaOS

    AgentaOS (Aristokrates OU, Estonia) becomes the official seller for its users' digital products and services, handling card checkout, subscriptions, VAT/sales tax in 135+ jurisdictions, invoices and chargebacks, and paying out twice a month in EUR or USD to the seller's own bank in ~85 countries, Brazil included.

    Why it scored lower: Regulatory and partner dependency (licenses, acquiring, Pix requires licensed partner); thin per-sale margins; crowded fee war (Creem 3.9%, Dodo 4%, Paddle 5%); large sellers churn to cheaper rails; Paddle already sup…

    Open ProductBots.ai →

Research, not investment advice. Numbers keep the label they were published with.