The Core Thesis: Liquidity Over Fundamentals

Macro investor Dan Krausz argues the recent sell-off in AI infrastructure (semiconductors, credit spreads) is a liquidity crisis, not a demand crisis. AI fundamentals remain strong, but global capital availability is contracting.

The Mechanism: The Liquidity Trap Door

• The Waterfall Effect: Treasuries and Hyperscalers (top credits) consume capital first. With massive twin deficits (fiscal + AI CapEx), the volume of top-tier issuance is crowding out the next credit tier (AI infrastructure).
• Time Frames Shorten: Concerns that were “18 months out” (e.g., financing gaps) become immediate when liquidity weakens.
• From Cash to Market: Hyperscalers funded their CapEx from internal cash hoards for years. Now relying on market financing, their build-out is directly exposed to liquidity conditions.

The Policy Dilemma (The Trampolines)

• The Fed is stuck in a trilemma: Hiking rates to fight inflation kills housing/small business. Easing risks inflating the AI bubble.
• The U.S. government views “winning the AI race” as existential. Krausz expects policymakers to deploy trampolines (liquidity support) under the trap door, such as adjusting Treasury issuance (short vs. long end) or discussing a “Fannie Mae for AI.”

The Great Rotation (Fallen AI)

• The recent rally in “AI loser” stocks (e.g., Intuit, Adobe) is a mechanical short squeeze from the unwind of relative value trades, not a fundamental improvement.
• Disruption risk remains. Krausz predicts the fundamentals for these companies haven't changed, and the disruption narrative will resurface.

Macro Outlook & The Productivity Paradox

• Consumer Weakness: Discretionary wallet share is structurally declining. A material weakening in consumption is expected next year.
• The Productivity Paradox: “Productivity as a concept is extremely important. Productivity as a data point, almost useless.” AI productivity gains take years to show in data and are often strongest in recessions, not booms.
• Questionable Margins: AI is a moat destroyer. Efficiency gains may flow to consumers (commodity trap) rather than permanently boosting corporate margins, making the long-term margin story bearish.

Bottom Line

Equities are highly vulnerable to further liquidity tightening. Policy makers retain room to ease conditions (strong dollar, low term premium). If they do, the market holds. If not, a major correction (the trap door) is highly probable.

The Case of Jennifer Kesse

Jennifer Kesse disappeared without a trace from her apartment complex in Orlando, Florida, on January 24, 2006. The 24-year-old was considered extremely reliable, professionally successful, and very close to her family. To this day, her case remains unsolved.

Background & Personality

• Jennifer grew up in Florida, studied business administration, and worked as a project manager at Central Florida Investments.
• She was determined, cheerful, and protective – in her sorority, she was affectionately called “Mother Hen.”
• Her family placed great importance on safety: Jennifer practiced so-called “safe calls,” always staying on the phone when walking alone until she reached home safely.
• Shortly before her disappearance, she bought a condominium in the Mosaïque at Millenia complex, where numerous construction workers were still on site – many female residents felt harassed and watched by them.

The Day She Disappeared

• On January 23, 2006, Jennifer returned from a Caribbean vacation with her boyfriend Rob Allen and drove back to Orlando early Monday morning.
• That evening, she spoke with her family, a friend, and finally with Rob at 10 p.m. – the conversation ended in a disagreement about their long-distance relationship.
• The next morning, she never showed up for work. In her apartment, the shower was still damp, a towel had been used, but her purse, phone, keys, and also her brother’s forgotten phone were missing.

The Mysterious Car Discovery & Video

• Her black Chevrolet Malibu was found 1.6 km away in a parking lot – parked, locked, and without visible damage.
• A surveillance camera captured an unknown person parking the car around noon. Their face was hidden behind a fence post.
• The person was approximately 1.60 m tall, wearing light clothing and dark shoes with white socks. Whether it was a man or a woman remains unclear.
• Two witnesses also reported seeing a black car driving erratically out of the apartment complex that morning, as if someone was fighting over the steering wheel.

Investigation & Dead Ends

• Police interviewed many construction workers, but some could no longer be found.
• A later lead pointed to a maintenance worker named “Gino,” who had worked in Jennifer’s apartment shortly before her disappearance. He denied any involvement and passed a lie detector test.
• Investigators later found hand and wipe marks on the car’s hood and a shoe print near the gas pedal – but none could be matched to anyone.

What Happened Later?

• In 2016, Jennifer Kesse was officially declared dead.
• After years of legal battles, her family gained access to the investigation files in 2019.
• In 2022, the Florida Department of Law Enforcement took over the case.
• In 2025, additional DNA traces were found on previously untested material – they are now being analyzed using modern methods.

Conclusion & Personal Assessment

The most plausible theory is that Jennifer was overpowered and abducted on her way to her car – possibly by someone who knew her routines and had already been watching her. The case remains puzzling, but the new DNA evidence gives the family hope for answers.

The 8 Best AI Trading Repos on GitHub – Ranked

The author analyzed the top 20 repos and selected the best 8, which together have over 400,000 stars. He ranks them with tiers (C, B, A, S) and emphasizes that Trading Agents is the best repo.

The Repos in Ranking Order

LLM Trading Lab (Tier C)

• Description: An experiment where GPT-4 managed $100 of real money for 6 months to pick stocks. After 1 month, it was up 24% and beat the market. The repo evolved into a framework for LLM trading experiments.
• Pro: Transparent, real data, publicly available, great for learning.
• Con: No repeatable framework, no fully automated trading, no long-term profitability guarantee.

Freak Trade (Tier B)

• Description: Crypto trading bot since 2017, 56,000 stars, supports live trading and a practice mode. Controlled via Telegram, includes backtesting and ML strategy optimization.
• Pro: Proven, regular updates, easy Telegram integration, practice mode.
• Con: Only as good as your strategy, crypto-only (unless using perps), setup is more technical.

Daily Stock Analysis (Tier A)

• Description: Sends an AI report on stocks every morning with buy/watch/sell alerts, score (0-100), price levels, and risk alerts. 67,000 stars, quick to set up.
• Pro: Very easy setup, runs daily automatically, cost-effective (only needs an AI key).
• Con: Not deep research (mostly LLM inference), main interface in Chinese, needs tuning for real alpha.

Day Trader Skill (Tier B)

• Description: New repo (released a week ago) with 11 Claude skills for day trading: chart scanning, news tracking, entry/exit timing, and replay data for backtests.
• Pro: Fast to use, comprehensive features (trade scanner, risk management, broker integration).
• Con: Very new, not thoroughly tested, just skills – not a full hedge fund system.

AI Hedge Fund (Tier B)

• Description: Simulates investment legends like Warren Buffett, Munger, Michael Burry, and Cathie Wood. Provides diverse perspectives on stocks and can execute trades via APIs (e.g., Alpaca).
• Pro: Different investment styles, backtesting, flexible integration.
• Con: Analysis is generic, not real alpha, somewhat gimmicky.

Vibe Trading (Tier A)

• Description: AI research assistant from a university lab in Hong Kong. Accepts plain English queries, performs backtests, and can handle paper trading via Alpaca. 36,800 stars, regularly updated.
• Pro: Self-improving agent, multi-agent teams, supports backtesting and institutional data.
• Con: Still relatively new, but already very versatile.

Anthropic Financial Toolkit (Tier A)

• Description: Toolkit by Anthropic (creator of Claude) for banks and investment funds. Includes agents for meeting prep, market research, earnings review, valuation, etc. 40,800 stars.
• Pro: Professional-level analysis, free for Claude users, trained like real analysts.
• Con: Built for professionals, may require paid data sources (e.g., F&P, Quiver), no auto-trading.

Trading Agents (Tier S) – The Best Repo

• Description: The most-starred trading repo on GitHub (117,000 stars). It operates like a real investment firm with 12 agents: Analysts → Researchers → Trading Team → Risk Management → Manager. Can be connected to Alpaca for live execution.
• Pro: Combines research, analysis, and execution – like having a hedge fund in your pocket.
• Con: Complexity, requires an onboarding period.

Final Thoughts and Tips

• The author recommends trying all repos and combining the best parts (e.g., Trading Agents for research, others for execution).
• Setup guides are available in the Free School community, ready to plug into Claude or GPT.
• Since they are open source, you can create custom workflows and even contribute your own improved version.

Introduction: Two Breakthrough Updates in Chemical Reprogramming

• Two recent scientific advances suggest that biological age could be reversed with simple molecules instead of complex gene therapies.
• Dr. David Sinclair's lab presented preliminary data on SL100; Dr. Alejandro Ocampo published a study using a two-molecule cocktail.

Ocampo's Two-Molecule Formula: Tranylcypromine & Repsox

• Ocampo reduced an original seven-molecule mixture to only two compounds: Tranylcypromine (TCP) and Repsox.
• TCP inhibits LSD1, an enzyme that removes methyl groups from histones – this restores chromatin compaction and suppresses transposable elements.
• Repsox blocks the TGFβ receptor ALK5, reducing oxidative stress and inflammatory markers.

Results in Cell Cultures

• In dermal fibroblasts from older donors (56–83 years), treatment led to:
  • Normal cell division (unlike the seven-molecule mixture)
  • Reduced DNA double-strand breaks and senescence markers (p21, p53, IL-6)
  • Increased repressive chromatin marks (H3K9me3, H3K27me3), persisting 30 days after removal
• No collagen damage or fibrosis – cell identity was largely preserved.

Lifespan Extension in Nematodes (C. elegans)

• Continuous treatment increased median lifespan by 42.1% (from 19 to 27 days).
• Single compounds each gave 31.6% – the combination was slightly superadditive.
• Treated worms showed better movement, longer reproductive window, and increased resistance to oxidative stress.

Sinclair's SL100: Preliminary Mouse Data

• In rapidly aging 'LAKI' mice (progeria model), SL100 prevented weight loss.
• Median lifespan increased by 15.4% (not significant), maximum lifespan trended above 33%.
• SL100 is a proprietary oral candidate; results are still unpublished.

Pharmacological Challenges & Safety Concerns

• Tranylcypromine irreversibly inhibits monoamine oxidase (MAO) – interaction with tyramine-rich foods (cheese, red wine) can cause hypertensive crises.
• Repsox systemically blocks TGFβ – this can damage heart valves and blood vessels.
• Thus, these molecules are research tools, not immediately applicable therapeutics.

Outlook & Next Steps

• Ocampo's work shows that two molecular nodes suffice – this enables development of safer second-generation compounds.
• Gladyshev's study (2026) warned of lipid accumulation in liver and kidney in mice with seven-molecule mix – even the two-molecule combo induced lipid droplets in mouse cells.
• Translation to humans requires selective LSD1 inhibitors without MAO activity, and tissue-specific ALK5 inhibitors.

Conclusion

• Chemical reprogramming is a promising but still early field of research.
• Readers are invited to share their opinion: Local gene therapy vs. targeted small molecules.

Insider Selling in AI Stocks: A Warning Sign?

This video analyzes massive insider stock sales at leading AI companies since June 2025. It reveals a pattern that goes beyond routine financial planning.

Key Figures and Sales Volume

• Nvidia (Director Mark Stevens): Sold approximately $1.35 billion worth of stock on the open market (virtually all insider sales at the company).
• Broadcom (Chairman Henry Samueli): Two sales packages totaling around $500 million (each ~$250 million).
• Arista (CEO Jayshree Ullal): Sales of about $145 million, with $131 million occurring within four days in October.
• CoreWeave (CSO Brian Venturo): Since the IPO, sales exceeding $1.2 billion, primarily after the lock-up period expired.

Why Are They Selling?

• Planned Sales (10b5-1 Plans): Most sales occur through pre-arranged trading plans, protecting insiders from insider trading accusations. The decision to sell is often made months in advance.
• Timing of Sales: Many sales happen right after quarterly earnings reports or during price strength.
• Lack of Purchases: The key signal: At Nvidia, Arista, and CoreWeave, there has been not a single insider purchase on the open market in the last 12 months. At Broadcom, one small purchase was dwarfed by $500 million in sales.

Conclusion

The video presents two interpretations:

1. Bearish View: Insiders are cashing out near the peak of the AI hype, believing valuations are unsustainable.
2. Neutral View: This is routine portfolio diversification, tax planning, and estate management, planned long in advance.

Key Takeaway: The absence of insider purchases combined with massive selling is a noteworthy indicator that warrants attention.

Introduction: Payment for Orderflow Ban

• Since July 2026, brokers are no longer allowed to accept fees for forwarding orders to specific exchanges. This Payment for Orderflow Ban was feared to increase costs, but surprisingly some areas have even become cheaper.

Depot Fees and Basics

• Out of 56 tested brokers, 31 offer a completely free depot with no conditions.
• Comdirect has peculiar terms: new customers get it free, but existing customers pay €1.95/month.
• Overall, 11 depots receive the top rating, all with free ETF savings plans.

ETF Savings Plans and Details

• Only 22 depots fulfill both basic conditions (free depot + free ETF savings plans).
• Smart Broker Plus, ING, and Consorsbank currently offer free plans, but without permanent guarantee.
• New filter: Search for your specific ETF and check if it can be invested in for free.
• Execution days vary: Trading 212 offers flexible dates, Scalable Capital has 9 dates.

Order Fees and Taxes

• Orders: Trading 212 is completely free; Neobrokers often free from €500 or charge €1.
• Traditional banks (ING, Consorsbank) charge percentage fees (e.g., 1.75% at ING).
• Xetra trading is becoming cheaper (Trade Republic €2, Scalable €1.99).
• Taxes: All Top11 automatically handle capital gains tax. Beware of Revolut, XTB, eToro – you must handle taxes yourself.

Interest on the Cash Account

• Trading 212 promotes 4.2% promotional interest (new customers, 4 months), then 2.8%. However, partly invested in money market funds.
• Trade Republic offers 2.5%, Scalable Capital 2.6% (on a separate savings account).
• Interest often needs to be activated.

Test Winner and Alternatives

• 1st place: Traders Place – low fees, sub-accounts for FIFO trick, €60 bonus via Finanzfluss.
• 2nd place: Scalable Capital, 3rd place: Trading 212 – own depot management, solid conditions.

App Overview

• Quick look at the apps of Traders Place, Scalable Capital, and Trading 212: portfolio overview, savings plan setup, cash accounts.

Conclusion

Despite the Payment for Orderflow Ban, the market remains competitive. Choosing the right depot depends on individual behavior – the Finanzfluss depot comparison helps with the decision.