MungerWunger Posted February 5 Posted February 5 I think owning hard assets are in again ... (waste management, airplane engines, railways etc.)
Junior R Posted February 5 Posted February 5 36 minutes ago, MungerWunger said: Did Anthropic announce something again? MCO, SPGI, FDS, TRI, MSCI, FICO ... Yeah they did around noon est
UK Posted February 5 Posted February 5 (edited) 35 minutes ago, MungerWunger said: I think owning hard assets are in again ... (waste management, airplane engines, railways etc.) +1 'Who crawled through a river of shit and came out smelling like a rose?' Edited February 5 by UK
brobro777 Posted February 5 Posted February 5 39 minutes ago, MungerWunger said: I think owning hard assets are in again ... (waste management, airplane engines, railways etc.) And tobacco/nicotine outperforming in 2026 just like last year!
dipod Posted February 5 Posted February 5 More CPNG leaps. Pricing on this common stock is not making much sense at all.
Rainier Posted February 5 Posted February 5 (edited) 1 hour ago, MungerWunger said: Did Anthropic announce something again? MCO, SPGI, FDS, TRI, MSCI, FICO ... https://www.bloomberg.com/news/articles/2026-02-05/anthropic-updates-ai-model-to-field-more-complex-financial-research https://www.anthropic.com/news/claude-opus-4-6 The press release was only from 2 hours ago Edited February 5 by Rainier
multidecade Posted February 5 Posted February 5 35 minutes ago, Rainier said: https://www.bloomberg.com/news/articles/2026-02-05/anthropic-updates-ai-model-to-field-more-complex-financial-research https://www.anthropic.com/news/claude-opus-4-6 The press release was only from 2 hours ago So, is this now going to replace Capital IQ? Can anyone with access to Claude try it out? It's obviously not going to replace indices, but still very interesting.
gfp Posted February 5 Posted February 5 As a marker of a general reprieve in the software rout, Microsoft has hit 397 which is basically where I've been expecting it to go this entire time. Now it can bounce for long enough for people to forget
formthirteen Posted February 5 Posted February 5 2 hours ago, rogermunibond said: Complex financial research I asked Mr. Claude how he plans on replacing us humans: Spoiler How Anthropic and OpenAI Will Disrupt Financial Research Role 1: Senior Equity Research Analyst Layer 1 — Understanding: The Current Workflow and Its Pain Points Financial research today is a pipeline of five activities, each with compounding inefficiencies: Data collection. Analysts pull SEC filings, earnings transcripts, market data feeds, credit reports, and alternative data into spreadsheets and internal systems. Pain point: this is manual, repetitive, and error-prone. A single coverage initiation can require 40+ hours of data gathering before any analysis begins. Financial modeling. Analysts build three-statement models, DCFs, LBOs, and comps in Excel. Pain point: models are fragile, hard to audit, and most of the time is spent on formatting and mechanical formula work rather than judgment calls about assumptions. Qualitative synthesis. Reading management commentary, industry reports, competitor filings, and news to form a thesis. Pain point: the volume of information far exceeds any individual's reading capacity. An analyst covering 15 names cannot realistically process every relevant document. Report writing. Producing the actual research note—often 20-50 pages for an initiation. Pain point: writing is the bottleneck. Senior analysts spend evenings editing junior drafts rather than thinking about markets. Client interaction. Calls, emails, bespoke analysis for buy-side clients. Pain point: most requests are commoditized ("send me your model," "what's your EPS estimate for Q3") and don't require senior judgment. Layer 2 — Analysis: Where AI Capabilities Map to Pain Points Anthropic (Claude) and OpenAI (ChatGPT/GPT series) now offer capabilities that directly target each stage: Pain Point Capability Product Data collection Long-context processing (Claude's 1M token window), web browsing, PDF/filing ingestion Claude Financial Analysis (with PitchBook, Morningstar, Daloopa connectors); ChatGPT Deep Research Financial modeling Code execution, structured reasoning, spreadsheet manipulation ChatGPT agent (benchmarked on IB analyst modeling tasks); Claude Code with agent teams Qualitative synthesis Multi-step agentic research, source comparison, citation OpenAI Deep Research (synthesizes hundreds of sources in minutes); Claude's deep research mode Report writing Long-form generation, formatting, tone control Both platforms generate structured, cited reports Client servicing Agentic workflows, MCP integrations, tool use Claude Cowork plugins; ChatGPT connectors and scheduled tasks The key development as of early 2026: both companies are moving beyond "chat assistants" into agentic workflow products that execute multi-step tasks autonomously. Anthropic's Claude Cowork plugins and OpenAI's ChatGPT agent represent the shift from "tool you query" to "agent that works." Assessment: Every stage of the research pipeline is now addressable by frontier AI. The question is no longer if disruption occurs but how fast organizational inertia, regulation, and trust gaps slow adoption. Role 2: AI Strategy Consultant Layer 3 — Synthesis: Business Model Shifts and Market Structure Changes Taking the analyst's assessment as input, three structural shifts become clear: Shift 1: The collapse of the labor-arbitrage model. Sell-side research has historically monetized junior analyst labor—data gathering, model maintenance, formatting—and packaged it as "research." AI agents now perform this labor at near-zero marginal cost. Anthropic's CEO has stated that AI could displace half of all entry-level white-collar jobs in the next 1-5 years. The immediate casualty is the economics of large research departments that bill for headcount rather than insight. Shift 2: Data becomes the moat, not analysis. When any firm can deploy Claude or ChatGPT to analyze public filings in minutes, the differentiator shifts to proprietary data access. This explains Anthropic's partnerships with PitchBook, Morningstar, and Daloopa—the value is in the connector, not the model. Firms with unique datasets (transaction data, satellite imagery, expert networks) gain; firms that repackage public information lose. The recent $285 billion selloff in software/data stocks reflects this repricing: investors are questioning whether Thomson Reuters' Westlaw or FactSet's terminals command the same premium when AI agents can synthesize raw sources directly. Shift 3: Democratization of institutional-grade research. A $200/month ChatGPT Pro subscription or a Claude Enterprise seat gives a solo fund manager capabilities that previously required a team of five analysts. This compresses the advantage of scale. Independent research and smaller buy-side shops gain disproportionately. New market structures become possible: AI-powered research-as-a-service platforms, dynamic "living" research notes that update continuously, and real-time competitive intelligence dashboards generated by agent teams. Segment-by-segment disruption ranking: Segment Disruption Level Rationale Sell-side equity research Very High Labor model directly attacked; MiFID II already pressured revenues Independent/boutique research High (positive) Massive capability uplift at low cost Buy-side fundamental High Internal research teams shrink; AI augments remaining PMs Quantitative/systematic Medium Already automated; AI adds at margins (NLP alpha, alternative data) Retail/self-directed High (positive) Institutional-quality analysis becomes accessible Layer 4 — Evaluation: Barriers and Acceleration Points Barriers that will slow disruption: Regulation. Financial regulators (SEC, FCA, ESMA) have not yet established frameworks for AI-generated research. Liability for errors in AI-produced investment recommendations remains unclear. Compliance teams will move cautiously. Trust. Portfolio managers making $50M allocation decisions will not blindly trust an AI report. Hallucination risk (Claude has been documented hallucinating credentials and data points) means human-in-the-loop review remains mandatory for high-stakes calls. Proprietary data moats. Bloomberg Terminal, Refinitiv Eikon, and similar platforms bundle data access with workflow. Switching costs are high. AI agents still need data inputs—they don't replace the data itself. Organizational inertia. Banks and asset managers are notoriously slow adopters. Many still run COBOL systems. Cultural resistance from senior professionals who built careers on the current model will slow internal rollout. Accelerators that will speed disruption: Cost pressure. Sell-side research budgets have been declining since MiFID II unbundled research from execution (2018). AI offers a way to maintain coverage breadth while cutting headcount. Competitive dynamics. Once one major firm deploys AI research at scale, others must follow or risk losing clients. Bridgewater's public adoption of Claude signals that tier-1 buy-side firms are already in production. Model capability jumps. Claude Opus 4.6 outperforms GPT-5.2 on finance-specific benchmarks (GDPval-AA) by ~144 ELO. ChatGPT agent benchmarks show it outperforms deep research on investment banking modeling tasks. Capability is improving quarter over quarter. Agent ecosystems. Both Anthropic (MCP protocol, Cowork plugins) and OpenAI (connectors, scheduled tasks) are building the integration layer that embeds AI into existing workflows rather than requiring workflow redesign. Strategic outlook: Disruption will be uneven. Junior research roles compress within 2 years. Mid-level roles transform within 3-4 years. Senior roles (with client relationships and judgment) persist but with radically different workflows. The overall research headcount at a typical sell-side bank drops 40-60% by 2029, while output per remaining analyst increases 3-5x. Role 3: Chief Investment Officer Layer 5 — Conclusion: A Three-Year Restructuring Roadmap Drawing on both prior assessments, here is how an investment firm should restructure its research function. Year 1 (2026): Foundation — Deploy and Learn Actions: License Claude Enterprise (Financial Analysis tier) and ChatGPT Enterprise. Run both in parallel to evaluate which excels at which tasks. Current evidence suggests Claude is stronger on long-context document analysis; OpenAI's agent is stronger on structured modeling and web-based research. Deploy AI agents on the lowest-risk, highest-volume tasks first: earnings transcript summarization, comp table generation, data pull automation, and first-draft model updates. Retain all current headcount but reassign 30% of junior analyst time from data work to AI supervision, prompt engineering, and output validation. Establish an internal "AI Research Quality" function (2-3 people) that audits AI outputs, tracks hallucination rates, and builds prompt libraries. Target operating model change: Junior analysts become "AI-augmented analysts" whose primary skill is directing and validating AI agents, not building spreadsheets from scratch. Year 2 (2027): Scale — Restructure Teams Actions: Reduce junior analyst hiring by 50%. Remaining hires must demonstrate AI fluency as a core competency. Promote the strongest AI-augmented analysts to a new "Research Engineer" role that designs agentic workflows, builds custom tool integrations, and maintains the firm's proprietary prompt and data pipeline. Expand coverage universe by 2x without adding headcount, using AI agents to maintain "passive coverage" models that auto-update and flag anomalies for human review. Begin publishing AI-assisted research to clients with appropriate disclosure. Establish the firm's AI research brand as a competitive differentiator. Negotiate data licensing agreements with alternative data providers to feed AI agents, recognizing that proprietary data inputs are now the primary source of differentiation. Target operating model change: The research team shifts from a labor-intensive pyramid (many juniors, few seniors) to a diamond shape (few juniors, more mid-level research engineers, same number of seniors). Year 3 (2028): Differentiate — AI-Native Research Actions: Launch "living research" products: client-facing dashboards powered by agent teams that continuously monitor and update company models, thesis trackers, and risk indicators. Reduce total research headcount by 40% from 2026 baseline. Remaining team is smaller, more senior, and higher-paid. Senior analysts and PMs spend 80% of time on judgment-intensive work: thesis development, management access, client relationships, and contrarian analysis that AI struggles with. Build or acquire proprietary datasets that create defensible advantages when processed by AI agents. Evaluate whether sell-side research can be replaced entirely by internal AI capability, potentially saving millions in research procurement costs. Roles That Disappear, Transform, or Emerge Category Roles Disappear Junior research associate (data gathering), model maintenance analyst, report formatting/production Transform Senior analyst (becomes AI-augmented strategist), portfolio manager (direct AI interaction replaces analyst intermediary), compliance reviewer (shifts to AI output auditing) Emerge Research engineer (designs agentic workflows), AI quality analyst (validates outputs, tracks hallucination), data pipeline architect (manages proprietary data feeds into AI systems) The Bottom Line The firms that treat AI as a headcount replacement tool will underperform. The firms that treat it as a capability multiplier—enabling smaller teams to cover more ground, think more deeply, and serve clients with living, continuously-updated research—will define the next era of financial analysis. The window to build this capability is approximately 18 months before it becomes table stakes.
73 Reds Posted February 5 Posted February 5 19 minutes ago, gfp said: As a marker of a general reprieve in the software rout, Microsoft has hit 397 which is basically where I've been expecting it to go this entire time. Now it can bounce for long enough for people to forget LOL, calling Greg Abel.
Marco Van Basten Posted February 5 Posted February 5 42 minutes ago, gfp said: As a marker of a general reprieve in the software rout, Microsoft has hit 397 which is basically where I've been expecting it to go this entire time. Now it can bounce for long enough for people to forget Why have you been expecting MSFT to go to $397? Thank you.
gfp Posted February 5 Posted February 5 34 minutes ago, Marco Van Basten said: Why have you been expecting MSFT to go to $397? Thank you. I was visited by ghosts in a dream. One was Dinar and he whispered it in my ear and then I woke up before I could get the "why"
Castanza Posted February 6 Posted February 6 17 hours ago, gfp said: I was visited by ghosts in a dream. One was Dinar and he whispered it in my ear and then I woke up before I could get the "why"
gfp Posted February 6 Posted February 6 I am buying the Sprott physical uranium trust, or "SPUT" another worthless "asset" with no cash flows that can't be valued so obviously the intrinsic value is zero
vakilkp Posted February 6 Posted February 6 42 minutes ago, gfp said: I am buying the Sprott physical uranium trust, or "SPUT" another worthless "asset" with no cash flows that can't be valued so obviously the intrinsic value is zero
cubsfan Posted February 6 Posted February 6 1 hour ago, gfp said: I am buying the Sprott physical uranium trust, or "SPUT" another worthless "asset" with no cash flows that can't be valued so obviously the intrinsic value is zero Do you find it interesting because of the Ai concern that power will be unavailable to many? I've been mulling an effective way to capitalize on the potential electricity shortage. It's interesting in California to watch them close hydro electric and fossil fuel plants - as well as not allowing no new large nuclear facilities. But certainly hearing a lot about the promise of smaller modular reactors. Something has to give.
Paarslaars Posted February 6 Posted February 6 2 hours ago, gfp said: I am buying the Sprott physical uranium trust, or "SPUT" another worthless "asset" with no cash flows that can't be valued so obviously the intrinsic value is zero You might find STND interesting.
Spekulatius Posted February 6 Posted February 6 3 hours ago, cubsfan said: Do you find it interesting because of the Ai concern that power will be unavailable to many? I've been mulling an effective way to capitalize on the potential electricity shortage. It's interesting in California to watch them close hydro electric and fossil fuel plants - as well as not allowing no new large nuclear facilities. But certainly hearing a lot about the promise of smaller modular reactors. Something has to give. The small reactors are many years (perhaps a decade) away from production. The startups working on them will tell you otherwise, but when you look at what they’d are working on, it’s clear they have ways to go. One thing about the small modular reactors is security? How do you make sure l that mo malicious actors will take control of one and uses it to make a dirty bomb. Could be quite a mess. Large central units are early to protect but if you have many small ones then it could become an issue. I guess that’s something to worry about later down the road,
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now