AI-Assisted Research / Systematic Trading
Brendan builds trading research systems with AI and shows the plumbing: data, backtests, alerts, and trade audits, with the honest caveat that asking an AI for a profitable strategy does not produce one. Here is the stack behind those workflows, what each piece needs to run (code, an API key, or a server), and the tools that can do the same jobs.
Hybrid research and automation education. Demos span stocks, options, crypto, and futures. Not a single stock day-trading playbook. Caveat: asking Claude for a profitable strategy does not establish one (https://www.youtube.com/watch?v=CPkrCoIbBIA&t=42s).
Subscribers
81K subscribers · 2026
Public statement, not verified
Reports two years building trading systems with Claude, math and economics study at UCLA, and more than three years in investment banking at Raymond James.
Source videoEntity
person
Channel joined 2024
Subscribers
81K
Channel views
3.2M
Videos
61
On YouTube since
2024
Avg views per video
about 52,500
Stack range
$0 to $49/mo
Did you know?
3.2 million views from 61 videos in about two years.
Avg views per video is total channel views divided by public video count, the channel average. Stack range runs from the Free Start bundle to the priciest bundle. All figures 2026.
Market focus
Holding period
Preferred conditions
Train on 12 months, test blind on the next 3 months, roll forward, include costs, and lock settings before the blind test.
This evaluates a rule, not an entry signal. The windows are an example, not a fit for every strategy.
Upload a broker CSV with your own rules and real market bars; code calculates P&L, holding time, MAE, and MFE first, then AI reviews rule breaks and proposes changes.
No entry signal comes out of this. Missing bars cannot be filled from a model's memory.
An indicator webhook sends candle data and a chart image to an agent that posts an analysis to Telegram for human review.
A message pipeline, not autonomous execution. Stop and exit rules are not publicly detailed as a repeatable stock setup.
AI tags news, earnings, and filings against explicit choice and rating questions to set research priority or route a record for review.
Classification is not verified catalyst truth, a price prediction, or a valuation. Set your own confidence threshold and human escalation.
STEP 01
1. Before any research
Write fixed rules, define the instrument, and decide what the AI will and will not do.
As taught by AI Pathways · Source
STEP 02
2. Historical prep
Get bars, choose training and blind windows, add trading costs.
As taught by AI Pathways · Source
STEP 03
3. Event-driven analysis
An alert or ticker request pulls candles, news, and a chart image for AI interpretation; the output is a research message.
As taught by AI Pathways · Source
STEP 04
4. Post-trade review
Import trades and rules, compute metrics in code, send a compact result to Claude.
As taught by AI Pathways · Source
STEP 05
5. Ongoing positions
Monitor holdings, events, and new information against the current thesis.
As taught by AI Pathways · Source
STEP 06
6. Recurring agent tasks
Hermes with task memory and scheduled checks, connected to Telegram.
As taught by AI Pathways · Source
STEP 07
Premarket, open, and midday
No fixed daily schedule found in his public videos.
As taught by AI Pathways
STEP 08
Weekly
No fixed weekly routine found in his public videos. The flow is data in, fixed calculation or classification, traceable output, then human review.
As taught by AI Pathways
The Full Stack
Their pick on the left. Same job, different tool on the right. Prices sit side by side so you can do the math.
Scanner recipe
Walk-forward test, then the other demo parameters he states, each labeled with its instrument.
The 12 and 3 month windows are an example, not a prescribed fit for every strategy. Simulation results are not forward profit expectations. Stock scanner settings (price, volume, relative volume, float, gaps, market cap, sector) were not found in his public videos and are not inferred from the Bitcoin model.▶ 2:11
Load this recipe into
$99/mo (rate offered to the first 400 members)
His paid community for building AI trading systems.
Affiliate link
Book a call, no public price
Affiliate link
Their pick
Basic $0 · Essential $155.40/yr ($12.95/mo equivalent); Essential month-to-month not published on the page we checked
Exchange data is separate (U.S. stock bundle $9.95/mo, non-professional).
Webhook alerts need two-factor authentication on the TradingView account.
Why they use it. An indicator alert fires a webhook that starts the AI analysis, with the chart and candle data attached.
Affiliate link
Their pick
Basic $0 / Pro $7/mo
Basic is 50 requests a day with a watermark; Pro is 500 a day, unwatermarked. A chart image is not a price feed or an exchange license.
Why they use it. Turns the chart into an image the model can read.
Affiliate link
Their pick
Basic $0 (8 API credits a minute, 800 a day) / Grow from $79/mo
Basic is for internal, non-display use. Extended-hours and broader coverage need a plan check.
Why they use it. Pulled the candle data for his first AI agent prototype.
Affiliate link
Their pick
Not in our catalog
Data rights and coverage still need checking before you rely on it.
Why they use it. Example historical bars for the walk-forward demo.
Same job, different tool
$0 extra if you already have charts
Replaces. The chart-image API step
What you give up. Manual upload; not a scheduled image API or a numerical price feed
What the AI actually does. Claude reviews the chart image you upload in a normal chat.
Affiliate link
Same job, different tool
$0 for eligible Schwab clients
Replaces. Chart alerts and native scans
What you give up. Alerts stay inside Schwab's platform, not a universal connection to another tool or broker; custom studies need thinkScript work
Affiliate link
Keeping the TradingView webhook route makes sense if you want an indicator to start the analysis on its own. The manual route covers chart review, but nothing fires until you upload the chart.
Their pick
$0.042 per million input tokens, output tokens free (metered, no flat subscription)
Text only. Data extraction, other models, and integrations are separate. Classification is not verified catalyst truth or a price prediction.
Why they use it. Labels incoming news, earnings, guidance, and filing text against explicit questions so the important records get looked at first.
Affiliate link
Their pick
Developer $0 / Business from $449/mo
The free Developer plan is for development and testing only, with a 24-hour article delay and 100 requests a day. It is not a free live catalyst feed.
Why they use it. Fetched recent headlines for his 2025 agent prototype.
Affiliate link
Same job, different tool
$0
Replaces. Confirming the catalyst from the primary source
What you give up. No squawk or broad fast-wire discovery; EDGAR filings and company releases arrive as separate events
Same job, different tool
$0
Replaces. Labeling incoming text by explicit rules
What you give up. Cannot match a model's judgment on ambiguous text; rules and scripts need upkeep
Affiliate link
Setup effort: Low (manual checklist), Technical (scripted). The scripted version needs code.
Same job, different tool
$20/mo or $200/yr
Replaces. The n8n research and digest prototype
What you give up. No indicator webhook, no custom API pipeline, no guaranteed run or data volume
What the AI actually does. Summarizes dated sources into a structured report with citations.
Affiliate link
Setup effort: Low (manual), Moderate (scheduled). Scheduling: https://support.claude.com/en/articles/13854387-schedule-recurring-tasks-in-claude-cowork
Same job, different tool
Free
What you give up. Updates about four times a day, so it is research, not a live wire or an API
What the AI actually does. Connects news and data to explain catalysts.
Affiliate link
Their pick
Claude Pro $20/mo or $200/yr (Claude Code included, subject to plan limits) + Python $0
Display layer shown: Streamlit and Plotly (price not in our catalog).
Why they use it. Claude Code writes the engine; fixed code does the math, and the AI works around the calculation, not inside it.
Affiliate link
Their pick
$20/mo; no annual Plus discount verified
API-key use is billed separately; a subscription is not unlimited API credits.
Why they use it. Shown as a second coding assistant for model and API integrations.
Affiliate link
Same job, different tool
Standard $299 / Professional $379 one-time (24 months of upgrades)
Replaces. A ready-made backtest engine
What you give up. Data downloaders and data are extra; your rules still need coding in its own language
Affiliate link
Same job, different tool
$389 one-time; optional $159 for another year of updates
Replaces. A ready-made backtest engine
What you give up. Historical data is a separate purchase; rules still need scripting
Affiliate link
Keep the original engine where a working, checked version already exists. Replacing an engine to trim a modest subscription can create more coding and verification work than it removes. Data, costs, and testing are needed either way.
Their pick
Community $0 (you host it; hosting extra) · Starter about $269/yr (billed €240/yr in euros, about $22/mo equivalent, 2,500 workflow executions a month); Starter month-to-month not verified
Feed and model calls cost extra; n8n does not include unlimited third-party model tokens.
Why they use it. Links the alert, data pull, AI step, and message into one workflow.
Affiliate link
Their pick
Hermes software: open source · Hostinger KVM 1: $6.49/mo promotional equivalent, $11.99/mo renewal on a two-year renewal term (initial charge for the selected term not established)
Model tokens, extra services, and maintenance are separate. Not counted as a $6.49 month-to-month plan.
Why they use it. Runs a persistent research agent with task memory and scheduled checks on a local machine or a VPS.
Affiliate link
Same job, different tool
$20/mo or $200/yr
Replaces. Part of the persistent agent job
What you give up. Not an arbitrary Hermes runtime, a continuous custom process, or a broker-API execution service
What the AI actually does. Runs a recurring research check through supported connectors and returns a report.
Affiliate link
Same job, different tool
$0 (subject to free-plan limits)
What you give up. Nothing runs on a schedule; every step starts with you
What the AI actually does. You paste dated sources and get a structured summary back.
Affiliate link
The demonstrated pipeline sends research messages for human review. It is not proof of an autonomous broker execution system. Keeping n8n and Hermes makes sense if you want custom triggers and always-on agents; the no-code route covers research and digests only.
Their pick
A NinjaTrader trade export appears as example input in the audit video. That is not proof of his execution platform.
AIThe AI alternative
Let rules or an AI agent place the trades through a broker API. Automation executes your plan faster; it doesn't make a bad plan good.
Alpaca
Alpha Score 77/100
API-first brokerage built for developers to build and deploy algorithmic trading strategies in equities and options.
QuantConnect
Alpha Score 65/100
Cloud-based algorithmic trading platform with a backtesting engine, live trading across multiple brokerages, and a community strategy library.
StockHero
Alpha Score 64/100
Cloud-based trading bot platform with prebuilt bot templates, paper trading, and multi-exchange support for automated crypto and stock strategies.
Keep the broker that gives you the data and permissions your instruments need. A broker API in a transcript does not establish reliable unattended live trading.
Or the tools the other educators use for execution
Prices are subject to change fast, like AI technology.
Check before you switch
Their pick
Python $0 + Claude Pro $20/mo or $200/yr
Missing bars cannot be filled from a model's memory. Screenshots add context but do not replace numerical bars.
Why they use it. Imports a broker CSV, rebuilds each trade against real minute bars, computes P&L, holding time, MAE, and MFE in code, then hands Claude a compact result to coach on rule breaks.
Affiliate link
Same job, different tool
Full journal replacement after checking one imported day; partial substitute for custom audit software
Pro $19.99/mo / Platinum $29.99/mo
Pro $179/yr, Platinum $269.88/yr billed annually
Replaces. Imports, metrics, and chart review
What you give up. A custom minute-bar trade path and arbitrary strategy rules may differ; check instrument and import coverage
What the AI actually does. AI Coach and AI review on paid plans; paid REST and MCP access lets your own AI read the journal.
Same job, different tool
$49.95/mo (Free $0 for 30 trades a month; Silver $29.95/mo)
Replaces. MFE, MAE, exit, and fee analysis with broker sync
What you give up. No generative AI feature established; no annual price verified
Affiliate link
Same job, different tool
$0
Replaces. Manual P&L and rule-check log
What you give up. Everything is manual, including the contract multiplier for futures and options
Affiliate link
Prices are subject to change fast, like AI technology.
Dedicated local computer or hosted VPS for a persistent agent
A hosting choice from his agent video, not a workstation inventory. Desk, monitor, and chair models were not found in his public videos.
FullStack Alpha suggestion; no beginner setup found in his videos. Start with the computer you already own (no purchase needed).▶ 1:26
Apple MacBook Air 13.6" M5, 16GB, 512GB
$1,000 to $2,000
Affiliate link
Acer SB242Y H1bi (x2)
Under $100
Affiliate link
VIVO STAND-V002 Dual Mount
Under $100
Affiliate link
Elgato Stream Deck Mini
Under $100
Affiliate link
Prices change daily on Amazon. Check each item's current price. Compare more in the Gear Lab
Ready-made stacks
The AI Full Stack
One-time. Swing Starter desk $200 to $300 + computer $1,000 to $2,000. Check current prices on Amazon. Gear Lab
Claude ProAI research, scheduled research task, review questions$20/mo
TradesViz PlatinumAI imports, AI review, MAE/MFE-style trade review$29.99/mo
TradingView BasicCharts; no webhook pipelineFree
TenjiAI catalyst researchFree
SEC EDGARCompany IR + SEC EDGAR, primary-source catalystsFree$49.99/mo month to month
$469.88/yr
No code, no API keys, no server. $469.88/yr ($39.16/mo equivalent). Setup effort: Low to Moderate. Stays manual: chart review, catalyst checks, and any strategy change.
One-time. Swing Starter desk $200 to $300 + computer $1,000 to $2,000. Check current prices on Amazon. Gear Lab
Claude ProAI · Claude Code builds the engine and audit code · NEEDS CODE$20/mo
PythonWalk-forward engine and trade audit · NEEDS CODEFreeTradingView EssentialAlerts and webhooks · CUSTOM WORKFLOW CONFIG$14.95/mo
n8n StarterWorkflow, about $269/yr (billed €240/yr); month-to-month not verified · CUSTOM WORKFLOW CONFIGPrice varies
chart-img ProChart images for the model · NEEDS API KEY$7/mo
Twelve Data BasicBar data, 800 credits a day; Grow from $79/mo if you outgrow it · NEEDS API KEYFree
TypeSafe JevAI · metered at $0.042 per million input tokens, output free · NEEDS API KEYPrice varies
Hostinger KVM 1Always-on host: $6.49/mo promotional equivalent, $11.99/mo renewal, or the mini PC slot · NEEDS VPS OR ALWAYS-ON COMPUTERPrice varies$41.95/mo month to month
$439.40/yr
Code, API keys, and an always-on host. Known subtotal: $41.95/mo + n8n Starter (annual) + model tokens + hosting. Known subtotal: $439.40/yr + n8n Starter about $269/yr (billed in euros) + Jev tokens + hosting. A complete live-agent total is not quoted: usage, data rights, hosting, and model tokens vary by build. Setup effort: Technical.
$0/mo
SEC EDGAR$0
LibreOffice Calc$0Claude Free$0Stays manual. Record cleanup, catalyst collection, and review; no unattended agent. Setup effort: Low (all manual)
Subject to free-plan limits and your broker's data entitlement.
$20/mo
Or $200/yr ($16.67/mo equivalent) $200/yr ($16.67/mo equivalent)
SEC EDGAR$0LibreOffice Calc$0Claude Pro$20Stays manual. Checking data, catalysts, and any proposed strategy change. Setup effort: Low (manual research), Technical (if you build code)
The coding branch needs scripts and proper historical data. No API bill for ordinary included plan use within limits.
$49.99/mo
Annual comparison $469.88/yr ($39.16/mo equivalent) $469.88/yr ($39.16/mo equivalent)
TradesViz Platinum$29.99Claude Pro$20Stays manual. Checking data gaps, costs, adjustment rules, screenshots, and proposed strategy changes. Setup effort: Low to Moderate
Replaces repeated import, chart, and review work, not the full live-agent pipeline.
Totals cover software and selected data only. Commissions, exchange fees, hardware, hosting, model tokens, and taxes are not included. Prices checked 2026. TradesViz annual amounts are the advertised offers on the research date. The advanced webhook and agent branch is not a fixed $20 or $49.99 stack. See the builder column of the AI Full Stack for its known component prices.
AI Pathways' broker is not named in the public videos we reviewed, so we do not guess. Keep the broker you already have if it covers the market and the data your routine needs.
A dedicated scanner is not part of AI Pathways' documented process.
AI Pathways charts on TradingView alerts and webhooks (used in 2026; Basic $0 · Essential $155.40/yr ($12.95/mo equivalent); Essential month-to-month not published on the page we checked); chart-img.com (demonstrated in 2026; Basic $0 / Pro $7/mo); Twelve Data (historical in 2025; Basic $0 (8 API credits a minute, 800 a day) / Grow from $79/mo); Yahoo example history (demonstrated in 2026; Not in our catalog).
An indicator alert fires a webhook that starts the AI analysis, with the chart and candle data attached. Same job, different tool: Manual chart screenshot into Claude at $0 extra if you already have charts.
The software stacks on this page run $0 to $49/mo. The cheapest, Lowest-cost Research and Review, is Your existing broker's records and charts + company IR + SEC EDGAR + LibreOffice Calc + optional Claude Free for commentary.
Our pick, AI-Assisted Research, is Lowest-cost + Claude Pro; Claude Code and Python only if you choose to build an engine. Commissions, exchange fees, and hardware are separate. Prices checked 2026.
Yes. AI Pathways uses n8n (used in 2026; Community $0 (you host it; hosting extra) · Starter about $269/yr (billed €240/yr in euros, about $22/mo equivalent, 2,500 workflow executions a month); Starter month-to-month not verified); Telegram (demonstrated in 2026; Not in our catalog); Hermes agent on a Hostinger VPS (demonstrated in 2026; Hermes software: open source · Hostinger KVM 1: $6.49/mo promotional equivalent, $11.99/mo renewal on a two-year renewal term (initial charge for the selected term not established)). Links the alert, data pull, AI step, and message into one workflow.
18:23 · 2026
AI Pathways shows how to use an AI model to tag news, earnings, and filing text against explicit questions so the important records get read first. Treat the labels as triage, not verified catalyst truth or a price call, and set your own confidence threshold.
Watch on YouTube20:34 · 2026
Brendan audits trades by loading a broker CSV with your rules and real minute bars, letting code compute P&L, holding time, MAE, and MFE before Claude reviews rule breaks. It produces coaching, not entry signals, and missing bars can't come from the model's memory.
Watch on YouTube19:06 · 2026
Brendan uses Claude to monitor open holdings, events, and new information against each position's current thesis. The $102k in the title is his own result from a one-month test, not a benchmark to expect.
Watch on YouTube20:49 · 2026
Brendan sets up Hermes as a persistent research agent with task memory and scheduled checks, running on a local machine or a VPS and reporting to Telegram. Budget for model tokens, extra services, and upkeep, since hosting alone does not cover them.
Watch on YouTube17:49 · 2026
Before any research, Brendan writes fixed rules, defines the instrument, and sets what the AI will and won't do, then has Claude Code write an engine where fixed code handles the math. The key caveat: asking Claude for a profitable strategy does not establish one.
Watch on YouTube16:31 · 2026
Brendan's walk-forward method trains on 12 months, tests blind on the next 3, rolls forward, includes trading costs, and locks settings before the blind test. The windows are an example, and simulation results are not forward profit expectations.
Watch on YouTube22:11 · 2026
Brendan walks through using Claude Code to build trading strategies with a quant-style process. It sits inside his testing recipe, so treat any strategy it produces as a hypothesis until it survives your own testing.
Watch on YouTube24:32 · 2026
Brendan wires a TradingView indicator alert through n8n, sending candle data and a chart image to an AI agent that posts its analysis to Telegram for human review. It is a message pipeline, not autonomous execution, and webhook alerts need two-factor authentication turned on.
Watch on YouTube18:28 · 2025
Brendan's 2025 n8n prototype pulls candle data and recent headlines into an AI agent day trader. Check data plans first: a free developer news plan carries a 24-hour article delay and is not a live catalyst feed.
Watch on YouTubeDollar figures in video titles are the creator's own.
Watch the foundation lesson
Start with "10 Ways To Actually Use Jev For Trading (Here's What Works)" before the advanced material. Open on YouTube
Write the setups down
Copy the 4 setup cards on this page into your own words: conditions, entry, exit, risk.
Build the Lowest-cost Research and Review stack
Your existing broker's records and charts + company IR + SEC EDGAR + LibreOffice Calc + optional Claude Free for commentary. Learn the process before you pay for speed. Load it in the Stack Builder
Log 20 trades or paper trades
Journal every one, including the bad ones. Your journal, not a video, decides what works for you.
Upgrade to AI-Assisted Research
Lowest-cost + Claude Pro; Claude Code and Python only if you choose to build an engine. Add paid tools one layer at a time, starting with the job your journal says costs you most. Load it in the Stack Builder
A FullStack Alpha study order built from AI Pathways's public lessons. Process beats prediction: master one layer before you pay for the next.
Incoming transmission | Free weekly newsletter
FullStack Alpha has tested 200 AI stock tools. Be honest: how many of yours do the same job, and which one did you forget you signed up for?
One email a week fixes that: which tools are worth the money, what they cost right now, and how the traders on this site wire them into a routine. Five minutes to read. Zero noise.
Inside every issue
5 MIN READ
The Verdict
One AI trading tool tested and scored. What it does well, what it costs, and the catch the sales page leaves out.
Price Watch
Price changes and live promos on the tools traders actually use, monthly and annual side by side, so you stop overpaying.
Workflow Teardown
One educator's routine taken apart: the scan, the chart, the tools, and the free swap for every paid one.
Scan of the Week
Exact scanner settings you can copy, pulled from the educator's own lesson with the timestamp.
Process beats prediction. This is the email that keeps your process honest.
