Most AI newsletters cover announcements. This one covers what those announcements mean for your money. Bloomberg Terminal charges $27,000 per year for AI-assisted financial analysis. JPMorgan is paying $185,000 base salary for the skills in today's exercise. You're reading this for free. Here are the five stories that actually moved markets this week — plus one skill you can build in the next ten minutes.
Today's Top Stories
Google and Amazon led the round. The company's valuation doubled in under 12 months. Claude usage at enterprise customers grew 4x quarter-over-quarter, with financial services being the fastest-adopting vertical.
Enterprise AI spend is accelerating, not plateauing. If your firm isn't already experimenting with Claude for document analysis or deal flow, your competitors who are have a widening edge. The fundraise signals Anthropic can sustain the long model-quality arms race — bet on Claude Code capabilities continuing to expand.
Bloomberg's AI division confirmed they are fine-tuning Claude for financial data interpretation, earnings call analysis, and real-time market commentary. The models will be integrated into Bloomberg Terminal.
If Bloomberg Terminal is integrating Claude-based models into the world's dominant financial data platform, AI-augmented analysis is no longer optional for institutional desks. The workflows we cover every week are the playbook that professionals will pay Bloomberg $27k/year to access. You're reading this for $15/month.
The SEC's Division of Investment Management published new guidance clarifying that AI-assisted research still requires human review and attestation. Firms must maintain audit trails for AI-generated content used in client communications.
This is guardrail-setting, not a ban. It actually accelerates adoption by removing regulatory ambiguity. The message: use AI in your research pipeline, document your process, keep a human in the loop on the final output. Firms that build the audit-trail infrastructure now will be ready to scale when competitors are still asking their compliance teams for permission.
Multiple quant funds reported using Claude API with the 200k context window to analyze full earnings transcripts + 10-Q filings simultaneously, extracting guidance revisions and capex signals before traditional analyst notes publish.
The speed advantage is real and measurable. Analyst notes that used to land 2–4 hours post-earnings can now be replicated in 90 seconds per company. The skill-up exercise below walks you through exactly how to build this pipeline yourself, no API budget required — just Claude Pro.
JPM's technology division posted 200+ roles with explicit requirements for prompt engineering, LLM integration, and AI workflow design. Base salaries for mid-level roles start at $185k.
"AI skills required" is now a compensation multiplier. The skills practiced in today's exercise are exactly what JPM is paying $185k to hire. You can learn them in 10 minutes per day. We help you get there.
By the Numbers
AI adoption is not evenly distributed. Financial services leads every other sector by a widening margin — and the gap is accelerating.
This is the same workflow quant funds are using. You don't need API access or a Bloomberg Terminal. You need Claude Pro, a company name, and 10 minutes. By the end, you'll have a reusable analyst that processes any earnings call in 90 seconds.
Step 1 — Set the context (2 min)
Copy this prompt and paste it into a fresh Claude conversation. This sets the analyst's behavior for the entire session.
You are an expert equity analyst specializing in reading between the lines of management communication. Your role is to analyze earnings call transcripts and identify signals that differ from the headline numbers. For every transcript I give you, extract: 1. Forward guidance changes vs. prior quarter (explicit and implied) 2. Management tone shifts — are they more or less confident than last quarter? 3. Capex and R&D signals — are they investing more or cutting back? 4. Language patterns that historically precede guidance cuts or beats 5. Divergence from consensus — what is management saying that analysts haven't priced in? Score each signal on a 1-5 scale (5 = high conviction divergence, 1 = noise). Only flag signals scoring 3+. Respond with a structured brief I can copy into my research notes. Confirm you understand the framework before I share the first transcript.
Step 2 — Feed the transcript (3 min)
Find any earnings call transcript on Seeking Alpha (free), The Motley Fool, or directly from SEC EDGAR. For today's exercise, try Apple's most recent Q report. Paste the full transcript text into Claude after the setup prompt.
Here is the earnings call transcript for [COMPANY, QUARTER YEAR]. [PASTE FULL TRANSCRIPT HERE] Run the full analysis framework. Flag all signals 3+. Include the specific quote from management that supports each signal. Format as a structured brief I can share with a colleague.
Step 3 — Extract the edge (5 min)
Once Claude returns the brief, run this follow-up to identify the highest-conviction divergence between what management said and what the market has priced in.
Based on your analysis, what is the single highest-conviction signal that the market has likely NOT priced in yet? Format your answer as: - Signal: [one sentence] - Evidence: [the exact management quote that supports it] - Why the market may have missed it: [1-2 sentences] - Time horizon for this to resolve: [weeks/months/quarters] - Risk to this thesis: [1 sentence]
Save this 3-prompt workflow. Tomorrow you can run it on any earnings transcript in under 5 minutes. By end of earnings season, you'll have a consistent analytical framework applied across 20+ companies — something that takes a full analyst desk weeks to produce manually.
Quick Hits
- Perplexity Finance: Now indexing SEC filings in real-time. Useful for quick 10-K searches before Claude analysis.
- OpenAI Operator: Early access users report ~70% success rate on simple brokerage account tasks. Not ready for live trading but watching closely.
- Karpathy's CLAUDE.md: Hit 66k GitHub stars this week. If you use Claude Code and haven't read it, do that before the next issue.
Sample issue • Pro subscribers get this every morning at 6:00 AM ET. One new Skill-Up exercise per day.
Subscribe to get the next one →That's the issue. Tomorrow: how to build a macro regime detector using FRED data and Claude — the workflow that pre-emptively flags the kind of rate-shift environment we saw in 2022.