My AI-Powered Macro Analysis Workflow — How I Use Claude to Read the Fed, CPI, and Crypto Sentiment

Why Macro Matters for Crypto

There is a persistent myth that crypto trades independently from traditional markets. It does not. Since 2020, Bitcoin’s correlation with macro factors has been too consistent to ignore:

• Fed rate decisions have preceded 7 of the 10 largest BTC single-day moves since 2022
• CPI releases regularly cause 3-5% BTC swings within hours
• DXY (US Dollar Index) and BTC have maintained a roughly -0.6 to -0.8 inverse correlation over rolling 90-day periods
• Liquidity conditions (M2 money supply, reverse repo facility) have been leading indicators for crypto market cap direction

You do not need to become a macroeconomist. But if you are trading crypto without checking what the Fed is doing, you are flying blind in one of the most macro-sensitive asset classes that exists.

My approach is practical: I use Claude to help me process macro data faster and more systematically than I could alone. Claude does not give me an edge in predicting the Fed’s next move — nobody has that edge consistently. What it gives me is structure, speed, and a second pair of eyes that never gets tired of reading dense Federal Reserve language.

The Weekly Macro Workflow Overview

Every Sunday, I spend about 40 minutes building a weekly macro thesis. Here is the structure:

1. Gather macro data (10 min) — FOMC calendar, recent CPI/PPI, DXY chart, yields
2. Feed the Fed to Claude (10 min) — Summarize recent Fed communications
3. Crypto sentiment check (5 min) — Fear & Greed, social sentiment
4. Synthesize with Claude (10 min) — Build the weekly thesis
5. Log and compare (5 min) — Record the thesis, check last week’s accuracy

Let me walk through each step with the exact prompts I use.

Step 1: Gathering Macro Data

Before opening Claude, I spend 10 minutes collecting data from these free sources:

I copy these numbers into a plain text note. This takes about 10 minutes once you know where to look. The CME FedWatch Tool is especially valuable — it shows you what the bond market is pricing in for the next rate decision, which is a much better predictor than pundit opinions.

Step 2: Feeding the Fed to Claude

This is where Claude shines. Federal Reserve communications are written in deliberately vague, hedged language that is designed to convey information without being too explicit. Parsing this language is tedious for humans but is exactly the type of structured text analysis where LLMs excel.

Prompt: Fed Minutes Summary

After each FOMC meeting, the Fed releases minutes about three weeks later. I copy the key sections (the “Participants’ Views” section is the most important) and paste them into Claude:

You are a macro analyst specializing in Federal Reserve communications.
I am pasting a section from the most recent FOMC meeting minutes.

[Paste the "Participants' Views on Current Conditions and the Economic
Outlook" section here — usually 3-5 pages of text]

ANALYZE THIS TEXT:

1. HAWKISH vs DOVISH SCORE: On a scale of 1 (extremely dovish) to 10
   (extremely hawkish), rate the overall tone. Explain your rating by
   citing specific phrases.

2. KEY PHRASES: List the 5 most important phrases or sentences that
   signal the committee's direction. For each, explain what it means
   in plain English.

3. CHANGES FROM PREVIOUS: If I tell you the previous meeting's tone
   was [X/10], what shifted? Are they moving toward more hawkish or
   more dovish language?

4. INFLATION LANGUAGE: How is the committee describing inflation?
   Are they using words like "elevated," "persistent," "moderating,"
   "progress"? The specific adjectives matter.

5. EMPLOYMENT LANGUAGE: How do they describe the labor market?
   "Tight," "rebalancing," "softening"? This signals how much room
   they feel they have to cut or hold.

6. DISSENT SIGNALS: Are there any phrases suggesting disagreement
   among participants? Words like "some participants," "a few
   participants," "several participants" indicate different faction
   sizes.

7. CRYPTO IMPLICATION: Based on this monetary policy direction, what
   is the likely impact on risk assets in general and crypto
   specifically? (Hawkish = headwind for crypto, Dovish = tailwind)

Be precise. Quote the actual text when possible. I want to see the
specific words they chose, not your paraphrase.

Why Claude for this task

I have tried this with ChatGPT, Gemini, and Claude. Claude consistently produces the best output for Fed analysis for two reasons: it is better at quoting specific phrases from the source text rather than paraphrasing, and it is more calibrated about uncertainty. When the Fed language is genuinely ambiguous, Claude is more likely to say “this could be interpreted either way” rather than forcing a directional conclusion.

The “dissent signals” section (point 6) is particularly valuable. The Fed uses specific quantifiers — “a few” typically means 2-3 members, “some” means 3-5, “several” means more than half, “many” means most but not all. Claude reliably picks up on these distinctions and explains what they mean for consensus.

Prompt: CPI Data Interpretation

When a new CPI number drops, I paste the headline data into Claude within the first hour:

New CPI data just released. Here are the numbers:

HEADLINE CPI:
- Month-over-month: [X]%
- Year-over-year: [X]%
- Consensus expectation was: MoM [X]%, YoY [X]%

CORE CPI (excluding food and energy):
- Month-over-month: [X]%
- Year-over-year: [X]%
- Consensus expectation was: MoM [X]%, YoY [X]%

KEY COMPONENTS (if available):
- Shelter: [X]% MoM
- Used cars: [X]% MoM
- Services ex-shelter: [X]% MoM

CONTEXT:
- Fed funds rate currently at: [X]%
- Next FOMC meeting: [DATE]
- Market was pricing [X]% probability of a [cut/hold/hike] before this release

ANALYZE:
1. BEAT/MISS/INLINE: Relative to consensus, is this hotter, cooler,
   or in line? By how much?
2. TREND: Looking at the YoY number relative to the previous 3 months,
   is inflation accelerating, decelerating, or flat?
3. CORE vs HEADLINE: Is there a divergence? What does it tell us about
   underlying inflation trends?
4. SHELTER COMPONENT: Shelter has been the stickiest component. Is it
   showing any signs of rolling over?
5. FED REACTION: Given the Fed's stated focus on [whatever they
   emphasized in the last meeting], how does this print affect the
   probability of action at the next meeting?
6. MARKET REACTION PREDICTION: Based on historical patterns, what
   would you expect the initial market reaction to be? (Note: the
   initial reaction is often wrong and reverses.)
7. CRYPTO IMPACT: How should crypto traders interpret this number?
   Hotter CPI = Fed stays tight = headwind. Cooler CPI = Fed can ease
   = tailwind. But how significant is this particular print?

Quantify everything. Use actual numbers, not just "higher" or "lower."

Reading the output

The most actionable part of Claude’s CPI analysis is the comparison between the actual print and consensus expectations. Markets do not react to absolute inflation levels — they react to surprises. A CPI of 3.2% YoY that was expected at 3.4% is bullish for risk assets, even though 3.2% is still elevated by historical standards. Claude is good at framing the data in terms of “relative to expectations” when you prompt it to do so.

Step 3: Crypto Sentiment Check

After the macro data analysis, I check crypto-specific sentiment. This takes about 5 minutes.

Fear & Greed Analysis

Current Crypto Fear & Greed Index reading: [SCORE] ([LABEL])
7-day history: [list daily scores for the past week]
30-day average: [AVG]

BTC price: $[PRICE]
BTC 7-day change: [X]%

ANALYZE:
1. Is the current reading an extreme (below 20 or above 80)?
   Historically, extremes have been contrarian indicators.
2. What is the trend over the past 7 days — moving toward more fear
   or more greed?
3. Is there a divergence between sentiment and price? (Example: price
   rising but Fear & Greed falling = potential hidden weakness)
4. Based on historical patterns, what has typically happened to BTC
   in the 14 days after readings at this level?
5. How much weight should I give this indicator? (It is a composite
   of volatility, momentum, social media, surveys, dominance, and
   trends — some components are noisy)

Be specific about limitations. The Fear & Greed Index is a blunt
instrument, not a precise tool.

Social Sentiment from Twitter/X

For social sentiment, I do not use a paid tool. Instead, I spend 5 minutes scrolling Crypto Twitter, noting the dominant narratives, and then asking Claude to help me interpret the prevailing mood:

I just spent a few minutes reading Crypto Twitter. Here is a summary
of what I observed:

DOMINANT NARRATIVES:
- [Narrative 1, e.g., "Many accounts are talking about BTC breaking
  through resistance at $45K"]
- [Narrative 2, e.g., "Several prominent accounts are warning about
  a potential correction based on overextended funding rates"]
- [Narrative 3, e.g., "Altseason discourse is increasing, with focus
  on AI tokens and RWA sector"]

TONE:
- Overall mood: [Optimistic / Cautious / Fearful / Mixed]
- Engagement level: [High — lots of debate / Low — apathetic / Normal]

ANALYZE:
1. Based on these narratives, what is the crowd positioned for?
   (Expecting up, expecting down, or uncertain)
2. Historically, when the crowd is positioned for [X], what has
   tended to happen? (Crypto Twitter has a contrarian track record)
3. Are any of these narratives based on actual data, or are they
   mostly vibes and chart patterns shared without context?
4. What is the crowd NOT talking about that might be important?
   (The most useful signals are often in what people ignore)
5. If I wanted to fade the crowd (take the opposite position), what
   would that trade look like? Do not recommend it — just describe it
   for analytical completeness.

Be blunt. Most social media analysis is noise. Tell me if you think
any of these narratives have substance.

Why I include social sentiment

Social sentiment is the noisiest input in this workflow, and I almost excluded it entirely. I kept it for one reason: extremes in social sentiment are genuinely useful contrarian signals. When every account on Crypto Twitter is convinced BTC is going to $100K next week, that widespread agreement often marks a local top. When everyone is calling for $20K, bottoms tend to form.

The signal is not in the specific narratives — it is in the uniformity of opinion. When I tell Claude “the crowd is unanimously bullish,” it correctly identifies that as a contrarian warning sign. When I describe mixed or uncertain sentiment, Claude correctly notes that this is neutral and does not provide a clear signal.

Step 4: The Weekly Synthesis

This is the final and most important prompt. I take all the outputs from the previous steps and ask Claude to build a single coherent thesis:

I am building my weekly macro thesis for crypto. Here are my inputs
from today's analysis:

FED STANCE:
[Paste summary from Step 2 — hawkish/dovish score, key phrases,
direction of change]

LATEST CPI/ECONOMIC DATA:
[Paste summary — beat/miss, trend, Fed reaction likelihood]

DOLLAR AND YIELDS:
- DXY: [level], [1-week change]
- US10Y: [yield], [trend]
- Direction: Dollar strengthening/weakening

CRYPTO SENTIMENT:
- Fear & Greed: [score]
- Social sentiment: [summary from Step 3]
- BTC price: $[PRICE], [7-day change]

FED FUNDS FUTURES (CME FEDWATCH):
- Next meeting probability: [X]% hold, [X]% cut
- Market pricing [X] cuts by year-end

BUILD MY WEEKLY THESIS:
1. MACRO ENVIRONMENT (2-3 sentences): Summarize the macro backdrop
   for risk assets this week. Is the environment supportive, hostile,
   or neutral?

2. KEY RISK: What is the single biggest macro risk for crypto this
   week? (Upcoming data release, Fed speech, geopolitical event, etc.)

3. KEY CATALYST: What is the single biggest potential positive catalyst?

4. POSITIONING: Given this macro environment, should I be:
   - Aggressive (increase exposure)
   - Neutral (maintain current positions)
   - Defensive (reduce exposure, tighten stops)
   One word plus 2-3 sentences of reasoning.

5. WATCHLIST: What specific data releases, events, or levels should
   I monitor this week? Give me a day-by-day calendar if possible.

6. THESIS INVALIDATION: What would need to happen for this thesis
   to be completely wrong? Be specific.

7. CONFIDENCE: Rate your confidence in this thesis from 1-10.
   I expect most weeks to be 4-6. Only rate above 7 if the signals
   are unusually aligned.

This is a framework for thinking, not a trading signal. Optimize for
intellectual honesty over conviction.

What I do with the output

The synthesis prompt produces my “weekly macro note.” I copy the output into a Google Doc organized by date. Each week gets one entry. Over time, this builds into a journal that I can reference when similar conditions arise.

The most valuable section is “Thesis Invalidation” (point 6). It forces me to think about what would prove me wrong before the week starts, so I am not caught rationalizing when new data contradicts my view.

Real Walkthrough: Week of January 27, 2026

Let me show how this played out during a specific week.

Context

The FOMC had met on January 22 and held rates at 4.25-4.50%. The statement was slightly more dovish than expected — they removed the phrase “further progress” on inflation and replaced it with language about “continued progress,” which Claude rated as a shift from 6/10 hawkish to 5/10. CPI from January 14 had come in at 2.8% YoY, below the 2.9% consensus. DXY had been declining for two weeks, from 104.2 to 103.1.

My synthesis input

I pasted the Fed analysis, CPI summary, DXY data, and sentiment data (Fear & Greed at 52, neutral social sentiment) into the synthesis prompt.

Claude’s weekly thesis (paraphrased)

Macro environment: Mildly supportive. The dovish Fed tilt and cooling CPI create a favorable backdrop for risk assets, but the magnitude of the shift does not warrant aggressive positioning. The declining DXY is a tailwind.

Key risk: February 7 jobs report. A hot employment number could reverse the dovish narrative instantly.

Key catalyst: If DXY continues below 103, this could trigger a broader risk-on move as dollar weakness historically correlates with crypto strength.

Positioning: Neutral, leaning constructive. Maintain positions, consider small adds on dips, but do not chase. Wait for the jobs report before increasing exposure significantly.

Confidence: 5/10. Signals are directionally aligned (dovish Fed + cooling CPI + weak dollar = bullish for crypto) but the magnitudes are small. This is a “lean,” not a “conviction call.”

What actually happened

BTC moved from $43,200 to $46,800 that week, roughly an 8.3% gain. The dollar continued weakening. The February 7 jobs report came in roughly in line with expectations, which removed the key risk Claude identified without creating a new one. The macro tailwind Claude described played out.

Would I have traded differently without the analysis? Probably not dramatically — I was already leaning bullish. But the structured framework helped me define exactly what would change my mind (a hot jobs print), which kept me from second-guessing my position during the normal intra-week volatility.

Historical Macro Events and BTC Impact

Here is a reference table I built with Claude’s help, covering major macro events and their crypto impact:

The pattern is consistent: the direction and magnitude of the surprise relative to expectations matter more than the absolute data. A rate cut is not inherently bullish — it is bullish if the cut is larger than expected, and can actually be bearish if it is smaller than expected.

Claude helped me compile and format this table by analyzing each event when I described it. I cross-checked the price data manually against CoinGecko historical data.

Building a Macro Calendar

I maintain a Google Calendar with recurring macro events. Here is what I track:

Monthly (dates vary):

• CPI release (usually second week)
• PPI release (usually day after CPI)
• Non-farm payrolls (first Friday)
• FOMC meeting (8 per year, roughly every 6 weeks)

Weekly:

• Initial jobless claims (every Thursday)
• Fed speaker schedule (varies, check Fed website)

Quarterly:

• GDP advance estimate
• Fed Summary of Economic Projections (dot plot, 4x per year)

I set alerts for one day before each event. On the morning of, I run the relevant Claude prompt (CPI prompt for CPI day, Fed prompt for FOMC day). On non-event days, the weekly synthesis note carries me through.

The calendar prevents surprises. The single worst thing for a trader is getting caught off-guard by a macro event you did not know was coming. Even if you do not trade around these events, knowing they exist lets you manage risk — for example, reducing leverage before a CPI release.

When to Override Your Macro Thesis

The hardest lesson I have learned: macro changes slowly, but price changes fast. Your weekly thesis might be “dovish Fed = bullish for crypto” and that can be correct on a multi-week timeframe while being completely wrong on a Tuesday afternoon when an unexpected headline drops.

Scenarios where I override or discard the weekly thesis:

Black swan events. Unexpected events (bank failures, geopolitical escalation, protocol exploits) override all macro analysis. When something genuinely unexpected happens, I reduce exposure first and analyze second. The thesis becomes irrelevant until the new information is processed.

Data that directly contradicts the thesis. If my thesis is “dovish pivot incoming” and a CPI print comes in 50 basis points above expectations, I do not wait until Sunday to update my view. I re-run the synthesis prompt with the new data immediately.

Price action that defies the thesis for more than 5 days. If macro is bullish but BTC drops 10% over a week, the thesis might be wrong — or the market might be responding to something I am not seeing. After 5 days of contradictory price action, I do a full reassessment rather than assuming the market will “catch up” to my view.

Internal disagreement. If Claude’s synthesis says “neutral” but my gut says “something feels wrong,” I pay attention to the discomfort. I do not override the analysis based on gut alone, but I will tighten stops and reduce position sizes. Unexplained unease has saved me more than once.

The principle I follow: macro provides the backdrop, not the script. It tells me whether the wind is at my back or in my face, but it does not tell me exactly when or how far the boat will move. I use macro to set my default posture (aggressive, neutral, defensive) and let shorter-term signals handle the actual entry and exit timing.

The “Macro Changes Slowly, Price Changes Fast” Principle

This deserves its own section because it is the single most important thing I have learned about macro trading.

The Fed does not change its stance weekly. CPI trends take months to reverse. Dollar trends play out over quarters. When you build a macro thesis on Sunday, it should be roughly correct for the entire week and potentially for several weeks.

But BTC can move 10% in a day on a single headline. Price responds to new information instantly, while macro conditions evolve gradually.

The practical implication: your macro thesis should inform your position sizing and directional lean, not your specific entries and exits. If macro is bullish, be willing to buy dips and hold through volatility. If macro is bearish, tighten stops and reduce size. But do not expect your macro thesis to tell you “buy here, sell there” — that level of precision is not what macro analysis provides.

I see traders on Twitter posting things like “CPI was low, time to long BTC with 10x leverage.” That is using macro incorrectly. The CPI print supports a bullish lean, but it does not tell you the right entry price, the right leverage, or the right time horizon. Those decisions come from technical analysis, risk management, and personal circumstances — not macro.

Combining Macro with Other Research

Macro analysis is one pillar of my research routine. It works best when combined with:

On-chain data — Macro tells you the direction of the wind, on-chain tells you what participants are doing with their coins. A bullish macro backdrop combined with strong exchange outflows is a much stronger signal than either alone. See my on-chain data analysis workflow for the complete setup.

Technical analysis — I use my daily AI research routine for chart reading. Macro provides the context, charts provide the timing.

Bot parameters — When my macro thesis shifts from neutral to defensive, I adjust my bot settings: wider grids, smaller position sizes, tighter DCA intervals. The macro read does not trigger specific trades — it adjusts the parameters that my automated systems use.

Honest Assessment: Does This Actually Help?

After three months of running this workflow, here is my honest take:

What it does well: It forces me to check macro data systematically instead of randomly. It gives me a written record I can review. It identifies upcoming risk events so I am never blindsided by a scheduled data release. Claude’s Fed minutes analysis is genuinely useful — it catches nuances in language that I would miss when reading quickly.

What it does not do: It does not predict price movements with useful precision. My weekly thesis has been directionally correct about 60% of the time, which is marginally better than chance but not a reliable trading signal on its own. The 40% of weeks where my thesis was wrong included some of the biggest moves, which is exactly when you most want to be right.

Where I have been wrong: I have been consistently too slow to recognize narrative shifts. My macro thesis tends to lag actual market conditions by about a week because I rely on lagging data (CPI is backward-looking by definition, Fed minutes are three weeks old by release). This is the fundamental limitation of macro analysis: you are making forward-looking decisions based on backward-looking data.

The actual edge, if any: The edge is not in the accuracy of the thesis. It is in the risk management framework. On weeks when my thesis is bearish, I run smaller positions and tighter stops. On weeks when it is bullish, I am more willing to hold through drawdowns. This asymmetric sizing has had a measurable positive impact on my portfolio, even though the directional calls are only slightly better than random.

Next Steps

• Add on-chain data to your macro workflow — See how I use ChatGPT for on-chain analysis to combine macro signals with whale tracking and exchange flows
• Build a complete daily routine — My daily AI trading research routine is the 15-minute daily complement to this weekly macro workflow
• Learn the foundational prompts — The ChatGPT trading analysis playbook covers the basic prompt templates that underpin this workflow
• Understand the full AI trading landscape — If you are newer to AI-assisted trading, start with the AI Trading 101 guide