Hi there,
Three stories landed this week that together sketch what AI-powered finance actually looks like in practice: DeepSeek V4 drops the biggest open-weight model ever at $0.14/M tokens, Robinhood bets $75M on OpenAI while giving retail traders a plain-English buy/sell agent, and Bank of America hands 18,000 advisors back four hours per client meeting.
🔥 Featured Post
DeepSeek V4 Opens the Frontier, Robinhood Bets on OpenAI, and BofA Gives 18,000 Advisors Their Hours Back
- DeepSeek V4 Pro: 1.6T parameters, 49B active, 1M token context as default — runs on Huawei Ascend 950 chips, priced at $0.14/M input tokens
- Robinhood invested $75M in OpenAI and launched Cortex — a retail AI agent that buys, sells, and researches in plain English
- BofA's Meeting Journey deploys to 18,000 Merrill advisors, saving up to 4 hours per meeting across millions of annual client interactions
- OpenAI crossed $25B annualized revenue, raised $122B at $852B valuation — but is still losing $14B per year
- The pattern: frontier AI is now simultaneously open-sourced, retailized, and embedded in wealth management
📚 In Case You Missed It
From 3 Days to 3 Minutes: AI's Underwriting Revolution, the Fed's Stability Warning, and the $8B Model Risk Boom — AI is collapsing insurance underwriting from 3 days to 3 minutes, the Fed published a framework warning of 'model monocultures' as a new systemic risk, and 49% of consumers are already using AI for savings decisions — finance's AI transformation is now measured in minutes, not years.
Wall Street's AI Arms Race: Agentic Finance, Foundation Models for Fraud, and 5,000 Layoffs — All at Once — Wall Street's AI arms race hit full speed in Q1 2026: BlackRock launched Asimov for equity research, JPMorgan scaled its LLM Suite to 200,000 employees, Feedzai dropped RiskFM — the first tabular foundation model for financial crime — and OpenAI quietly acquired personal-finance startup Hiro.
GPT-5.5, Google's 8th-Gen TPU, and Why AI Is Finally Learning to Say 'I'm Not Sure' — GPT-5.5 nearly doubles FrontierMath Tier 4 scores vs. Opus 4.7, Google's TPU 8 superpods hit 9,600 chips and 2 PB memory, and MIT's RLCR slashes hallucination calibration error by 90% — three stories shaping how fast AI moves and how much you can trust it.
More posts dropping every day. Stay curious.
— Bhanu @ superml.dev