Welcome to Nofinity

Dr. Irina Conboy (formerly UC Berkeley, now CSO at Generation Lab) discusses her latest study on selective targeting of cancer and senescent cells and the future of systemic rejuvenation.

• The triple combination (DMA) of dichloroacetate (DCA), metformin, and ABT263 (navitoclax) was tested in 18–20-month-old mice – equivalent to a 65–70-year-old human.
• From treatment start, mice lived ~41% longer; relative to total lifespan, the extension was ~10–12%.
• Importantly, the main goal was not lifespan extension but eliminating cancer and senescent cells together without harming healthy cells.
• Mice showed improved treadmill performance and no increase in frailty – despite a chemo-like intervention.

• Cancer and senescent cells have damaged mitochondria and heavily depend on glycolysis.
• DCA forces cells to push pyruvate into mitochondria and use oxidative phosphorylation.
• Metformin inhibits mitochondrial electron transport (complex I), worsening the energy crisis.
• ABT263 blocks BCL survival proteins, allowing cytochrome C release and triggering apoptosis.
• Crucially, healthy cells have enough functional mitochondria and survive; diseased cells die selectively.

• ABT263 was used at only one-tenth of the usual dose to avoid thrombocytopenia (low platelets).
• In vitro tests on various human cells showed: cancer and senescent cells were eliminated, while healthy liver and neuronal cells remained largely unaffected.
• The authors view DMA not as a classic senolytic or chemotherapeutic, but as a dual-capacity therapy against both pathogenic cell types.

• Generation Lab is developing an injectable substitute for therapeutic plasma exchange (TPE) that antagonizes age- and disease-elevated blood proteins and promotes tissue repair.
• The drug "One Generation" consists of two FDA-approved components and is being prepared for clinical trials.
• A potential indication is prevention and treatment of cardiovascular disease.
• Dr. Conboy advises against biohacking with the combination until safety and dosing are validated in clinical trials.

Summary follows...

Universal Music Group (UMG) – Stock Analysis at €14

Universal Music Group's stock has dropped 50% over the past year, from around €28 to €14. Bill Ackman (Pershing Square) previously offered €30 per share, which the board rejected as undervalued. The stock now trades well below that level. This summary covers the business, finances, Ackman's thesis, and current valuation.

• UMG owns the rights to most major music IPs and represents 9 of the top 10 global artists (e.g., Taylor Swift).
• The market is growing, especially streaming. Partnerships with Spotify, Tencent, and expansion in India.
• Only missing major artist: Bad Bunny.

• Revenue growth has slowed; profits stable but below expectations.
• Free cash flow is disputed: management reports ~€1.5 billion, but after investments (catalog purchases, acquisitions like Downtown), it shrinks to about €700 million.
• Debt increased from €5.4 billion to almost €13 billion; equity barely grew due to dividend payouts.
• Dividend yield currently ~3.7% (€1 billion payout).

• Plan: Merge UMG with Pershing Square, increase leverage, buybacks, move listing to the US (higher P/E).
• Price target: €74 by December 2030 (5x return).
• Criticism: Ackman essentially wanted to buy the stock using the company's own money – effectively only €1.50 equity per share, the rest from UMG itself.
• The board rejected the offer; Ackman sold his stake before the official rejection.

• UMG is using the Spotify sale proceeds (€2.7 billion) for share buybacks.
• True free cash flow growth is limited, as UMG cannot negotiate better streaming deals (split 70/30).
• The stock trades at a discount of over 50% to intrinsic value, typical for European holding structures (e.g., Vivendi).
• Relatively speaking, UMG offers a 3.7% dividend and modest growth (~4%) for an 8% total return – but not a classic value investment.
• Risk: further price decline; becomes interesting at €10 (10% dividend yield).

• Not an absolute bargain – cash flow is weaker than reported.
• Relatively attractive for income-oriented investors comfortable with stagnation.
• A catalyst like Ackman's plan (US listing, higher leverage) is off the table for now.

• The 10-year yield is approaching 5.2%, the 30-year near 5.12%. The market signals that the Fed’s rates are not high enough.
• Inflationary forces (tariffs, Middle East conflict, AI buildout) persist. Expectations drive prices higher.

• Historically, rising long yields eventually trigger a growth scare (e.g., 2023 banking crisis), which brings yields back down.
• Currently we are in the up phase: markets fear inflation more than recession.

• High yields pressure long-duration assets — including cryptocurrencies.
• Bitcoin has not outperformed the S&P 500 for years; its valuation is at 2021 levels.
• Altcoins suffer more: they are a zero-interest-rate phenomenon. Liquidity is tight (global net liquidity ~$25 trillion vs. $30 trillion in 2021).

• The analyst admits being perplexed: Bitcoin at ~$84,000. Technically a golden cross appears, but a drop below the May high would be bearish.
• Opportunity cost: If Bitcoin yields similar returns to the S&P over five years, investors question the extra risk.

• The Fed must hike aggressively to calm long yields — but it hasn’t.
• Current policy is too accommodative (Fed funds rate 4% vs. 2-year yield 4.9%).
• Without a real growth scare or crisis, bond vigilantes will keep revolting.

• Watch yields: They dictate direction for risk assets.
• A crisis (like the pandemic) is needed to boost liquidity and crypto sustainably. Until then, crypto remains an underperformer.

• AI experts consistently underestimate the speed and capabilities of AI models. Adoption is not yet optimal, but value creation is enormous.
• AI agents are revolutionizing business models by attacking inertia-based companies (e.g., bank fees, subscriptions). They are fearless, not lazy, and not confused.
• Meta's Muse is a personal assistant for social networks, already reaching 300,000 downloads. Developed under Alexander Wang, who increased Meta's market cap by $200 billion.

• Oracle's Force Majeure for Project Jupiter (New Mexico) due to missing permits for gas pipelines. This affects the entire Stargate plan ($500 billion).
• High credit default risks for Oracle, Blue Owl, Bloom Energy, etc. – bonds are being rated junk level.
• Data center construction is severely hindered: 300 US communities have imposed moratoriums, but only 1.5 GW out of 38 GW are affected.

• Memory is sold out through 2027, prices rising. Large customers like Elon Musk receive shipments, others don't.
• Nvidia B200 continues to generate high revenue (~$7.88 per GPU hour) and pays off quickly.

• Google's Project Sun Catcher launches a prototype satellite with TPUs in 6 days – first actual data centers in space.
• SpaceX will connect these with Starship and Starlink – sole provider for space compute.
• xAI's Grokbot saves SpaceX 200 customer service staff – cost per inquiry drops from $28-65 to 20 cents.

• Bernie Sanders wants to ban AI, 20 years prison for AI development. This drives companies out of California.
• China is advancing – either you lead or fall behind. Singapore is training 80,000 finance workers in AI.
• AI saves lives: A patient with a rare blood cancer was saved by an AI-repurposed drug combination.

• Solana is recruiting institutional experts to become the native currency in the AI agent era.
• Next show with Aaron Bernett on SpaceX valuation – TAM is huge, moat unbeatable.

This video shows how to build an extremely fast AI trading bot using Jev and Claude. The author (Miles) demonstrates a live high-frequency trading bot running on Hyperliquid.

• Jev is not a traditional LLM, but a System One model.
• It doesn't generate text; instead, it evaluates states and returns structured decisions with probability scores.
• This enables fast classification at a fraction of the cost of Claude.

• Claude (the LLM) handles strategy development, backtesting, and open-ended analysis.
• Jev takes care of fast decision-making during execution – ideal for high-frequency trading with low latency (~300–500 ms).

• The full setup prompt is available for free in the school community.
• Jev is deployed via Vercel, with Claude Code acting as the control center.
• The bot can run on paper trading or live accounts – the architecture is identical.

• There is no magic strategy: 99% of strategies lose money.
• Use backtesting (e.g., custom engines, Quantedia, NFX) and test strategies on paper first.
• Only switch to a live account after successful forward tests.

• Jev can read live headlines (e.g., Bloomberg, Reuters, Yahoo Finance) and classify them as bullish, bearish, or neutral within milliseconds.
• This makes it possible to build automated news trading systems.

• Alpaca: Simple, supports stocks and crypto, paper and live trading via API.
• Bybit: Better for crypto with lower fees; connect via VPS, MCP, or Trigger Trade.
• Important: set risk rules (max position, daily loss, kill switch) and use alerts or manual approvals for large trades.

Jev is not a replacement for an LLM, but a powerful complement for fast decisions. The technology is useful not only for trading, but also for business workflows (e.g., data classification). Early adopters can gain a real edge.