How I Used 4 AIs to Figure Out Venezuelan Oil in Minutes ↓

The US just took control of the world’s largest oil reserves.

At least that’s the headline. 303 billion barrels. More than Saudi Arabia. Regime change, oil flows, America wins.

But here’s the problem…

You saw the news. You bookmarked some tweets. You’re “monitoring the situation.”

You know what that means. It’s the investor equivalent of “I’ll start the diet Monday.”

Meanwhile your feed is full of newly minted oil experts. Some say oil spikes. Some say it crashes. Some say this takes a decade. Without understanding what’s actually true, you can’t answer the question that matters: who wins and who loses?

I’m going to show you a 4-step AI workflow I use to get the complete picture on any breaking event, and extract trade ideas, in minutes.

I went from “monitoring” to a fully mapped view with ranked winners and losers. Here’s how.

Step 1: Crowdsource the Questions

I wanted a prompt that finds the best tweets on this topic. The people who actually know what they’re talking about, not the people who discovered what “heavy crude” means yesterday.

I use Grok for this. It has native Twitter integration that other models don’t have. You can ask it to search recent posts, filter by account quality, and pull verbatim text with URLs. Claude and ChatGPT can’t do this natively.

I seeded the prompt with a few tweets I thought were solid, then let it find more.

What came back was a corpus of 25 posts organized by angle, with author credentials and reasoning for why each view matters.

Here’s what people are debating:

• Supply reality vs. headlines. 77% of Venezuela’s oil is extra-heavy crude, heavier than water. It requires specialized refineries and competes directly with Canadian heavy. Venezuela can’t even export without importing diluent first.
• Timeline for recovery. Opinions split between months and decades. Infrastructure is rotting, upgraders are offline, and 18,000 engineers were fired in 2003 and never came back. This is not Iraq 2003.
• Price direction. Short-term geopolitical premium, medium-term bearish. Some analysts see oil capped near $50 if barrels return at scale into an oversupplied market.
• Winners and losers. Gulf Coast refiners were built for this crude. Chevron never left. Oilfield services get paid for activity. Canadian heavy producers face direct competition.
• Legal optionality. ConocoPhillips holds an $8.7 billion unpaid arbitration award. Regime change might finally create a path to collect.

Now instead of bookmarked tweets you’ll never read, you have a corpus you can actually use. That’s the next step.

Step 2: Build Understanding

Tweets tell you what themes experts are focused on. That’s a much better starting point than going in cold.

If you just ask an AI “tell me about Venezuelan oil,” you get Wikipedia-level stuff. But if you start with expert takes, you can extract themes and research each one comprehensively.

For this I use ChatGPT Deep Research. It has the most analytical lens of any model I’ve tested. The tradeoff: reports are dense and hard to read. We’ll solve that later.

I use a meta-prompt. Instead of writing a research prompt myself, I give the AI the tweet corpus and ask it to generate the prompt for me. It extracts themes, identifies what needs explaining, then builds a comprehensive research agenda. You’re outsourcing the prompt engineering to the AI itself.

What came back was a 35 page report. The kind of thing that would take a junior analyst a week, except it took minutes and didn’t ask for a deadline extension.

Here’s what the analysis reveals:

The World’s Largest Oil Reserves Are Not What You Think

• 77% is extra-heavy crude that cannot flow through pipelines unless blended with imported diluent.
• If diluent supply gets cut for six weeks, production starts declining.
• Venezuela cannot export its primary product without imports.

The Collapse Began with a Self-Inflicted Brain Drain

• In 2003, Chavez fired 18,000 skilled PDVSA employees, nearly 40% of the workforce.
• They emigrated to Canada, Colombia, and the Middle East.
• Institutional knowledge that took decades to build evaporated in months.

This Isn’t a Quick Fix

• The Iraq analogy is flawed; Iraq had conventional crude, maintained fields, and an intact workforce.
• Wood Mackenzie estimates $15-20 billion and 5-10 years just to add 500,000 barrels per day.
• The country with the largest oil reserves in the world currently imports gasoline.

This Is Geopolitics, Not Just Oil

• Venezuela, Iran, and Russia supply 41% of China’s crude imports.
• China has extended $60 billion in loans and receives 80-85% of Venezuelan exports.
• The US operation is about flipping a key supplier from China’s orbit.

The report covers a lot more: value chain mapping, refinery analysis, arbitration specifics, price scenarios, and historical precedents.

Now you have the research. How does this translate into concrete ideas? That’s the next step.

Step 3: Extract Trade Ideas

The deep research surfaced names, but ChatGPT reports are painful to read. The insights are buried in walls of text.

Claude is better at writing. I feed it the research and ask for trade ideas with thesis, risks, and key sensitivities.

Here’s a sample:

Gulf Coast Refiners (VLO, PBF): Long

• Coking refineries literally built for Venezuelan crude.
• Before sanctions, Valero was the largest US importer of Venezuelan barrels.
• When supply returns, cheaper feedstock and wider margins.

Chevron (CVX): Long

• Only US major still operating there.
• PDVSA owes them $3 billion, which they’ve been collecting in oil instead of cash.
• First-mover advantage in a multi-decade rebuild.

Oilfield Services (SLB, HAL): Long

• Paid for activity, not commodity prices.
• Rehabilitation means drilling, workovers, equipment deployments.

ConocoPhillips (COP): Long

• Not operating there, but holds $8.7 billion unpaid arbitration award.
• Regime change creates a path to recovery.

Canadian Heavy Producers (SU, CVE, CNQ): Underweight

• Venezuelan crude competes directly with Western Canadian Select.
• When those barrels return, WCS discounts widen.

More trade ideas in the full results linked above.

Now you have ideas. How do you make sense of everything? That’s the final step.

Step 4: Create Your Second Brain

Remember how I said ChatGPT reports are dense and hard to read? NotebookLM solves that.

Instead of manually slogging through 30 pages, I drop the report into NotebookLM and let it do the work. I can generate a mind map to see everything structured visually. I can chat with the report and ask specific questions. No more scrolling through walls of text trying to find that one paragraph about upgrader capacity.

I add three things: the tweet corpus from Step 1, the research from Step 2, and the trade ideas from Step 3.

First thing I do is create a mind map. I need to see everything structured visually. That’s how my brain works.

Now I can ask follow-up questions. “What’s the timeline for upgrader rehabilitation?” “Which refiners have the most coking capacity?” The answers pull from everything I’ve built.

As I dig deeper, I add more sources: 10-Ks, earnings transcripts, news as it comes out. The notebook grows with me.

My next steps: evaluate catalysts, schedule AI to surface relevant news, and start diligence on names I’m excited about. All from the NotebookLM.

The Bottom Line

This process works for any trending topic. Silver. Copper. BOJ policy. Tariffs.

You know the pattern. Topic starts trending. You bookmark tweets because some random guy replied “worth reading.” You tell yourself you’ll dig in later. Later never comes. You never form a view.

This workflow changes that.

Four tools. Minutes of work. From “monitoring the situation” to full mental models and specific trade ideas. Not days. Minutes.

Venezuela was the example. The process works for anything.