DHAHAB · ذهب · Stack
The real tools behind the trading, the code, and the cut — no fluff, only what I use daily.
Most "my stack" pages are a logo wall. This one is the honest version: for every tool, what it replaced, why it stayed, what it costs, and what adopting it cost me. Money is tight, so the default is the free tier until it blocks something real — and the ledger further down shows exactly where that rule breaks.
11 line items · 6 free · 5 paid · one software subscription, four consumables
Some links may become affiliate links, meaning I may earn a commission at no extra cost to you. I only list what I genuinely use, and the free-versus-paid note stays honest either way. Details: Terms · Privacy
01 · At a glance
Four numbers, each one checkable against the ledger in section 05.
A tool earns a place here by replacing something specific, and it keeps that place only if I can say plainly what adopting it cost me. So every card carries a tradeoff line. Where a choice is genuinely opinionated — a young package manager, a runtime that is not quite Node, a local model measurably weaker than the hosted one — the cost of being wrong is written on the card rather than left for you to discover.
02 · Trading
Charting, execution, and the screener I built when nothing off the shelf did what I needed.
Every chart I post starts here — it's the only platform where my anchored-VWAP workflows feel native instead of bolted on. The free plan already covers the workflow; the paid tiers mostly buy more indicators per chart and better alerting, neither of which has blocked me yet.
tradingview.com →The broker API my execution engine is built against — clean REST endpoints and a real free paper environment for testing execution logic without risking a dollar. The part that decided it: it's an official API. Reverse-engineered brokerage APIs get accounts restricted and usually have no paper mode at all.
alpaca.markets →My own CLI: scans a watchlist for AVWAP-reclaim setups and backtests them with no lookahead — I built it, then used it to break my own edge. The walk-forward result is on the build page, including the part where the headline number did not survive.
Read the build →Deeper on this section: Alpaca vs Interactive Brokers vs Robinhood · Anchored VWAP · Walk-forward backtesting · Paper to live · the execution engine · free risk calculators
Risk disclaimer: nothing on this page is financial advice. The trading tools listed here are what I use for research and paper trading; markets carry real risk of loss and past results — mine included — do not predict future ones. My own AVWAP-reclaim result is not a validated edge: +0.117R over 4,933 trades, 95% CI [+0.057R, +0.174R], against a risk-matched random entry that already captured +0.086R, leaving the signal's own contribution at +0.030R with a confidence interval crossing zero.
03 · AI & Code
Agents where they earn their keep, local models where they don't, and tooling that never makes me wait.
My daily driver for agent workflows — it runs my nightly automation, drafts my outreach pipelines, and ships most of the code on this site. It is also the only line on this page with no free tier, which is precisely why the local model exists: anything that is volume rather than judgement gets routed away from it.
claude.com/claude-code →A local 30B mixture-of-experts model on my own machine handles the bulk drafting work overnight — nothing per token, the data never leaves my laptop, and it keeps running when the network doesn't. It is not as sharp as a frontier model. That's the trade, and I make it on purpose for work that is long rather than hard.
ollama.com →Replaced pip entirely for me — environments and installs resolve in seconds instead of minutes, which matters when every project is Python. One binary now covers what pip, venv and pyenv used to do separately. Nothing since has made me want pip back.
astral.sh/uv →My default JS runtime and package manager — one binary covers install, run, test, and bundle, and it's fast enough that I stopped thinking about it. No separate package manager, test runner and bundler to keep in version-sync with each other.
bun.sh →How the local half of this works, in full: running a local LLM with Ollama · Ollama vs LM Studio vs llama.cpp · running a MoE model on a MacBook · getting structured output · prompt caching
04 · The Cut
The short list behind the 260→200 lb cut. Categories only — I'll name brands when I've vetted the deals, not before.
The one supplement I never skip — 5g daily kept my lifts from falling off a cliff while cutting hard. Plain monohydrate, not a branded blend; the fancier forms cost more per gram to do the same job.
Shop creatine →Hitting protein targets in a deficit is the whole game — a shake covers the gap on days when food alone can't. It is a convenience, not a requirement: if you can hit the number with food, skip it and save the money.
Shop whey →Cheap insurance for joints that take a beating from lifting and years of football and wrestling. This is the one I would drop first if the budget got tighter — I keep it because it is a few dollars a month, not because I can point at a measurement.
Shop fish oil →Taken before bed for sleep quality — the glycinate form because it doesn't wreck my stomach like the cheap oxide stuff. That difference is the whole reason to pay a little more here instead of buying the biggest bottle.
Shop magnesium →The part that actually moved the number: the calorie math behind the cut · weight tracking, signal vs noise · the free offline weight tracker I built
Not medical advice. This is what I take, not a recommendation for you. Supplements are the smallest lever in a cut — sleep, calories and protein do the work. Talk to a doctor before adding anything, especially if you take medication.
05 · The money
Every tool above, with what it replaced and whether it costs anything. No "pro tier recommended" hedging.
| Tool | What it replaced | Free tier | What paying actually buys |
|---|---|---|---|
| TradingView | Broker-native charting | Yes | More indicators per chart, better alerting. Has not blocked me yet. |
| Alpaca | Unofficial broker APIs | Yes — API and paper trading are free | Higher market-data tiers. Not needed for daily-bar research. |
| Swing Screener | Paid stock screeners | Free — I wrote it | Nothing. It is mine. |
| Claude Code | Writing boilerplate by hand | No | This is the subscription. There is no free version, and it is the one place I break my own rule. |
| Ollama + qwen3 | Metered API calls for bulk drafting | Free and open source | Nothing. You pay in hardware you already own and electricity. |
| uv | pip, venv, pyenv | Free and open source | Nothing. |
| Bun | Node and npm on my own projects | Free and open source | Nothing. |
| Supplements | Magnesium oxide, in one case | No | The only genuinely recurring cost on this page, and the first thing cut when money is tight. |
The rule: stay on the free tier until it blocks something real. Most "pro" upgrades buy convenience, not capability, and convenience is the cheapest thing to live without when you are a student. The one exception is the coding agent — and even there, everything that is volume rather than judgement gets routed to the local model so it never touches a metered API. Longer version: free vs paid tools for students.
06 · FAQ
More of the reasoning is written up in the guides, and the things I built with this stack are on the apps page.
Trading research runs on TradingView for charts and a screener I wrote myself, with the execution engine built against Alpaca's API. Code runs on Claude Code, a local model served by Ollama, uv for Python, and Bun for JavaScript. Everything else on this page is a supplement, not software.
Speed, and one less decision. uv resolves and installs an environment in seconds where pip took minutes, and a single binary covers what pip, venv and pyenv used to do separately. Every Python project I own now starts with uv, and nothing has made me want pip back.
One binary covers install, run, test and bundle, so there is no separate package manager, test runner and bundler to keep in sync. It is fast enough that install time stopped being something I think about, which is the whole point of a runtime choice.
Because the bulk of the work is not hard, it is just long. Drafting and rewriting passes run overnight on a mixture-of-experts model served by Ollama on my own machine: nothing per token, and the data never leaves the laptop. Frontier reasoning still goes to a hosted model. The local one handles volume, not judgement.
Most of it is free. uv, Bun, Ollama and the open-weight models cost nothing, Alpaca's API and paper environment are free, TradingView has a free plan, and the screener is mine. The line that genuinely costs money is the coding agent, which has no free tier. Money is tight, so the rule is simple: stay on the free tier until it blocks something real.
Alpaca. The paper environment is real and free, the REST endpoints are clean, and it is an official API rather than a reverse-engineered one. That last part matters: unofficial brokerage APIs get accounts restricted and usually have no paper mode at all.
Some of them may become affiliate links, which means I could earn a commission at no extra cost to you. Nothing is on this list because of a payout. Every item is something I already use, and the free-versus-paid note next to each one stays honest either way.
Want this stack pointed at your problem?
Same tools, same rule — free tier until it blocks something real, and the tradeoff stated up front. Recruiting instead? The full background is on the about page.