Welcome to Nofinity

Introduction: The Shift on OpenRouter

• 12 months ago, 72% of traffic on OpenRouter came from US AI models; today it's just 33%.
• Chinese models now account for over 50% – Deepseek alone is bigger than Google and OpenAI combined.
• In the week ending June 21, Chinese models processed 21.37 trillion tokens, US models only 5.76 trillion.

Why China Wins: Export Controls as a Catalyst

• US sanctions cut China off from top Nvidia chips, forcing a focus on efficiency over brute compute.
• Deepseek uses Mixture of Experts (MoE): only 5.5% of parameters are activated per query – drastically cutting costs.
• Price comparison: Deepseek V4 Flash costs $0.14 per million input tokens, comparable US models (GPT-5.5) cost $5 – a factor of 36x cheaper.

Real-World Switching Examples

• Lindy (US AI startup): Switched to Deepseek, saving 90% on inference costs – but hosts models on US soil.
• Coinbase: Uses Chinese open-weight models (e.g., GLM 5.2) for routine tasks – costs halved, hit rate jumped from 5% to 60%.
• Pinterest: Fine-tuned Alibaba's Qwen model – 90% cost reduction and 30% accuracy boost.

Volume vs. Value: The Critical Gap

• China dominates volume but not revenue: Anthropic captures 46% of OpenRouter revenue with only 12% of tokens.
• US models still lead in complex reasoning (lead of 3–12 months depending on the source).

Risks & Open Questions

• Data privacy: Queries to Chinese APIs route through servers in China – subject to China's National Intelligence Law.
• Distillation attacks: Alibaba allegedly tried to clone Claude's reasoning.
• Investment gamble: US hyperscalers are spending $725 billion on AI infrastructure – Sequoia flags a $600 billion gap between spending and actual revenue.

Conclusion: Two Stacks Emerge

• US luxury stack: High prices for top-tier reasoning – akin to a luxury brand at risk of commoditization.
• Chinese industrial stack: Nearly free AI as default infrastructure – soft power no embargo can reverse.
• The big question: Who pays back the trillion-dollar investment if the US lead is only months wide?

Introduction

The host, an experienced trader who lived through the dot-com bubble, discusses topics like the AI bubble, crypto proxies, Solana's burn mechanism, and future prospects. He emphasizes that early investments are crucial and warns against risks in proxies.

AI Bubble and Market Phase

• Dot-com Comparison: The dot-com bubble (1995-2000) saw 985% gains in 62 months; the current AI phase (47 months) is up 157% in the index, but the chip sector (e.g., Nvidia, Micron) has risen 345% since November 2022.
• Valuation: Despite the boom, many AI stocks are not overvalued – Nvidia trades at 14x forward earnings, Micron at 5.9x (vs. Cisco's 200x PE in 2000). There’s still $10.3 trillion in AI buildout ahead.
• Phase Estimation: The host believes we're in the middle of the cycle (comparable to 1998), with about 24 months of bull run left. He warns of excesses and advises watching debt, defaults, and slowing growth.

Proxies and Their Risks (Focus: DeFi Dev, Smarter Web)

• DeFi Dev (DFDV): Despite a low NAV premium (0.51), the company has high debt ($126M) and dilution of 130.76% since June 2025 – five times more than MicroStrategy. This led to an 89% drop from all-time highs. He advises not to buy leaps on proxies as they get 'completely cooked' due to implied volatility and dilution.
• Smarter Web vs. MicroStrategy: Smarter Web has a NAV premium of 1.37 (vs. 0.74 for MicroStrategy), little cash ($2M), and debt. MicroStrategy is more established (309x more Bitcoin) and safer. Rule: Buy proxies with low NAV premium and watch debt and dilution – not 'death and taxes' but 'debt and dilution'.

Solana's Burn Mechanism

• Burn Rate: With 60,000 SOL minted daily, only 434–990 SOL are burned currently. 10 billion transactions per day would burn 1.44% of supply annually, fully absorbing inflation and being bullish. Validators still earn priority fees. New rules (SMD-0550) double disinflation.
• Assessment: Solana becomes scarcer as adoption grows. Long-term, the burn mechanism needs adjustment, but it works for now. Key metrics: adoption (e.g., apps like Jupiter, Backpack) and simple valuations like market cap per transaction.

Additional Insights

• Space-Based Data Centers: Marvell lacks radiation-hardened technology, but Tesla/SpaceX work with Intel. Google already sent four TPUs to space – the future is here.
• Investment Strategy: Build 'retire-on-bags' (e.g., 655 SpaceX shares for $1M by 2032) and buy on dips (e.g., Palantir at $106, Nvidia at $88). Q4 will be hot, Q1 often sells off – so be patient.
• Critique of Funds: 550 holdings are nonsense – you lose returns through fees (2 and 20) and lack of transparency. Better invest in pure form (e.g., ETFs, direct stocks).

Conclusion

You are not too late to the party, but much 'juice' has already been squeezed. Proxies are risky – buy them only as lottery tickets. Focus on fundamental strength, low debt, and clear adoption. Be ready for dips and keep cash!

Important: Not financial advice, do your own research.

Summary: The Murder of Becky Bliefnick

Introduction

• On February 23, 2023, Becky Bliefnick, a 41-year-old mother of three, was found dead in her home in Quincy, Illinois.
• Her father discovered her lifeless body in the bathroom – a devastating scene.

The Night of the Murder

• The perpetrator entered through a forced window in a child's bedroom, kicked in the bedroom door, and shot Becky 14 times with a 9mm pistol.
• She tried to call 911 at 1:11 a.m. but couldn't complete the call.
• A security system shows the front door opened at 1:12 a.m. – the killer fled.

Background

• Becky and her husband Timothy (Tim) Bliefnick had a troubled marriage; Tim filed for divorce in 2021.
• The separation was marked by intense conflict and mutual accusations.
• Becky feared Tim and texted her sister: “If something happens to me, Tim is the first person to look at.”
• The final divorce hearing was scheduled just days after the murder; Becky wanted to prevent Tim's father from having unsupervised contact with the children.

The Investigation

• Surveillance cameras captured an unknown person lurking around Becky's house on nights before the murder, and a cyclist on the night of the crime.
• The bicycle had no reflectors on the wheels.
• Searches of Tim's home yielded incriminating evidence:
  • 27 spent cartridge cases from the same gun used in the murder.
  • Google searches like “How to open a window from outside?”, “How to build a silencer?”, “Can gunshot residue be washed off?”
  • A fake Facebook account used to buy a blue Schwinn bicycle without reflectors.
  • Gaps in his fitness tracker data during the relevant nighttime hours.
  • A crowbar and Aldi plastic bags matching fragments found at the crime scene.

The Trial

• Tim was charged with first-degree murder and home invasion.
• The prosecution argued that Tim killed his wife out of anger over the divorce and upcoming custody issues.
• The defense pointed to a lack of direct evidence (no DNA, no murder weapon) and suggested the Google searches were made after the murder out of curiosity.
• The jury reached a unanimous verdict after only 4 hours and 15 minutes of deliberation.
• Sentence: Life in prison without parole.

Conclusion and Thoughts

• This case illustrates how a toxic relationship can escalate into deadly violence.
• Particularly chilling: Tim deliberately sent Becky's father to find her body while his own children slept in his home.
• Becky's failed attempt to call for help highlights her final terrifying moments.
• The chain of circumstantial evidence was overwhelming – I believe the verdict was just.

What's your opinion? Let me know in the comments!

Core Message: Stress Relief Enables Healing and May Reverse Aging Processes

Dr. Martin Picard, director of the Mitochondrial Psychobiology Lab, explains that stress isn't inherently harmful; rather, the capacity for recovery is the key to longevity. The body has an innate healing ability that is activated through rest periods.

Key Insights from the Conversation:

• Gray hair is reversible: Graying linked to stress can actually reverse – a visible example of the plasticity of aging.
• Allostatic load: Chronic anticipation of stress consumes energy, leading to wear and tear. The solution isn't avoiding stress but finding intentional rest and regeneration times.
• Focus on mitochondria:
  • Stress (e.g., cortisol) increases cellular energy consumption by up to 60%.
  • Chronic stress diverts energy from repair processes like DNA repair and telomere maintenance, accelerating aging.
  • Hypermetabolism in aging cells drains energy from the body, similar to cancer or sepsis.
• Mind-body connection: Emotions and thoughts directly affect mitochondria – via cortisol, blood sugar levels, and mitochondrial DNA leaks that trigger inflammation.
• Practical steps: Mindfulness, meditation, and exercise help to be present in the moment, enabling healing. The goal isn't a stress-free life, but balancing stress with recovery.
• Marker for mitochondrial stress: GDF15 (growth differentiation factor 15) indicates energetic stress – ideally low at rest.

Conclusion:

Dr. Picard encourages proactive rest, not passive stress avoidance. The body heals itself when given the chance – this is the foundation for a healthy lifespan.

Stargate Project in Trouble

The Stargate project – a joint venture by Oracle, OpenAI, and SoftBank with up to $500 billion in planned investment – aims to build massive AI infrastructure in the US. However, its flagship site Project Jupiter in New Mexico faces delays, not from chip shortages, but from a natural gas pipeline.

• Issue: The 27 km "Green Chile Project" pipeline for fuel cell power is stuck in regulatory battles. State land office denied rights, pushing completion from August 2025 to February 2027.
• Oracle's response: Oracle issued a force majeure notice to operator Stack Infrastructure, seeking the right to halt lease payments for up to 3 years if the site isn't ready by 2028.
• Financial ripple effect: The site relies on $18 billion in construction loans; Oracle itself carries $156 billion in debt with negative free cash flow. Its credit rating was downgraded to BBB (S&P), just above junk.
• Founder risk: Larry Ellison pledged over 430 million Oracle shares as collateral for personal loans – more than a third of his stake. Oracle's stock dropped 56% from its peak, and a planned sale of 50 million shares was abruptly canceled.

Conclusion: Stargate was meant to validate the AI boom, but it highlights how fragile the massive debt structures are when facing mundane obstacles like a pipeline. A single local delay could unravel the entire chain from lenders to founder wealth.

1. Build a Financial Foundation

• Pay off expensive debt: Overdrafts (~9.4%) and consumer loans (~8.7%) first – risk-free, tax-free return.
• Emergency fund: 3–6 months' salary in a savings account for unexpected expenses.

2. Split Your Savings Rate for Maximum Subsidies

• Up to €30/month into the retirement deposit (AVD): 50% state subsidy (€15 bonus). Even better if you have children (up to €65 bonus) or are under 25 (one-time €200). Minimum: €10/month.
• €30–150/month into AVD: 25% subsidy. Example: With €100 contribution you get €32.50 bonus. Maximum €150/month for full subsidy (€45 bonus). Plus possible tax benefits.
• Important trade-off: AVD money is locked until retirement; withdrawals are fully taxed. Only worthwhile if costs are low and you truly save for retirement.
• Free ETF savings plan for everything above €150/month: Flexible for other goals or early retirement.

3. Check Employer Subsidies

• Employee savings scheme (VL): Employer bonus + possibly 20% state bonus (max €80/year) if income below €40,000. Transferable to a free depot after 7 years. Only few providers offer VL with ETFs (e.g. comdirect, Fidelity, Finvesto).
• Company pension scheme (BAV): Often inflexible and expensive. Only worth it if employer adds at least 20–30%. Anything beyond better in AVD or free depot.

4. Special Case: Saving for a Property

• You can withdraw from the AVD tax-free for a self-occupied home. But don't invest in stock ETFs – use short-term government bonds or money market ETFs to avoid market volatility before purchase.

Conclusion: First pay off debt and build emergency fund. Then invest up to €150/month in AVD for maximum subsidies, everything else in a free depot. Only take employer subsidies if they truly benefit you.

What is the Lelit Elizabeth 3?

• Dual boiler espresso machine with a 0.2 L espresso boiler and a 0.6 L steam boiler, weighing 15 kg, with a 2.5 L water tank and 58 mm portafilter.
• Price: €1,650 (€1,700 in black/white); it is just about to hit the market.
• Included: wooden-handled tamper, various filter baskets, blind basket, cleaning kit and water filter.

Updates and build quality

• Beautiful new wooden details on the portafilter, steam knob and Pagaia paddle – the overall feel is noticeably more premium.
• Very quiet in use: 56 decibels; the testers describe the brewing acoustics as genuinely pleasant.

The core feature: pressure and flow control

• Electronic pressure profiles with up to five programmable stages, set directly via the display – no app needed.
• Additionally, there is the Pagaia lever for manual flow control during extraction.
• No flow meter; instead, time-based control with a pressure sensor – remarkable at this price point.
• Example from the test: espresso at 3 bar for the first 8 seconds, then 8.5 bar until 45 g – cleanly repeatable.

Espresso quality and steam

• Tested with an Indian robusta: strong, sweet, with dark, almost peaty notes – overall espresso potential is rated very high.
• Steam power and milk texture are much better than the predecessor; 30 seconds for steaming, good microfoam quality.
• Weakness: temperature stability during consecutive shots. From the third shot onward, the temperature drops by about 2–3 °C.
• With a 2-minute gap between shots, temperature settles around 91 °C – stable values with only 1 °C offset.

Daily use and energy consumption

• Heat-up time: 20 minutes – not particularly fast, but efficient once up to temperature.
• Energy use: approx. 0.075 kWh for heating plus one espresso; 0.154 kWh with steam – about a third of a GS3.
• Large drip tray with 1200 ml capacity; the manual is solid but not particularly deep.

Verdict

• An absolute value-for-money wonder: for €1,650 you get electronic pressure profiles usually found only on much more expensive machines.
• Ideal for home use, less for heavy café duty; if entertaining guests, plan for 2-minute gaps.
• Clear recommendation from the testers if you like the cube-shaped design.