Last Updated on 6 March 2026

This page is a personal experiment, a living log of investment ideas. These are thoughts-in-progress, brief market notes I capture as they arise. I share these notes to distill signals from noise- highlighting events and themes that, in my view, create asymmetry between projected risk and reward.

Each idea represents, in my view, a piece of information or reasoning not yet reflected in market prices.

Performance Summary

This is a summary of the investment positions listed in this page.

Ticker	Entry Date	Entry Price	Current Price	Return (%)	Std	CVaR	Exposure	Conviction	Status
INTC	22 Aug 2025	…	…	…	…	…	Long	High	Active
ITA	18 Sep 2025	…	…	…	…	…	Long	High	Active
PAVE	25 Sep 2025	…	…	…	…	…	Long	High	Active
HWAY	25 Sep 2025	…	…	…	…	…	Long	High	Active
SRVR	26 Sep 2025	…	…	…	…	…	Long	Low	Active
DTCR	26 Sep 2025	…	…	…	…	…	Long	Medium	Active
SPY	30 Sep 2025	…	…	…	…	…	Long	Medium	Active
AIPO	1 Oct 2025	…	…	…	…	…	Long	High	Active
ARGT	28 Oct 2025	…	…	…	…	…	Long	Medium	Active
REMX	25 Nov 2025	…	…	…	…	…	Long	High	Active
SETM	25 Nov 2025	…	…	…	…	…	Long	High	Active

Portfolio average return is calculated using an equal-weighted position return index. Each position starts at 100 on its entry date and then moves with its price (closed positions stay fixed). The portfolio average return is the simple average of these position indices, expressed as a percent relative to 100.

These returns are provided on a gross basis, before transaction costs, fees and tax.

Portfolio Risk Analysis

Utilizing Souppe‘s risk model, we can better understand the true exposures of the portfolio we’ve created. See the figure below- it summarizes our holdings and portfolio-level beta exposure to 3 market regimes: offensive, defensive and all weather. In this portfolio, we assume equal security weights.

^{Portfolio regime beta exposures. Source: Souppe. In ( ): security weight in portfolio. In [ ]: the betas’ confidence score (1-100). In the portfolio line, the ( ) value is the beta standard error.}

Beta is the slope coefficient of a linear regression: for each 1% change in the explanatory variable, the target (our asset’s price) is expected to change by beta percent. The Regime Capture ratio measures how much upside market sensitivity the asset captures per unit of downside market sensitivity- above 1.0 is favorable, below 1.0 means the asset amplifies losses more than it participates in gains.

Like people, markets have “moods”, manifesting as “regimes”. They are influenced by various forces that operate concurrently, be it the credit cycle, geopolitics or policy mistakes. Our goal is to first identify our portfolio’s weaknesses, and then tweak its constituents to have a portfolio that is robust to various market conditions. A regime market beta measures how much the portfolio is expected to move for each 1% move in the broad market, conditional on the current market environment. Markets cycle through three regimes:

• Stress- the market is in a significant drawdown. Think 2000-2003, 2007-2009, March 2020 or 2022. We want our portfolio to have a stress beta below 1.0, which indicates that our portfolio has elements that dampen losses when markets crash.
• Recovery- the market is climbing back but hasn’t reached a new high yet. This is the rebound phase after stress. We want our portfolio to have a recovery beta above 1.0, which indicates elements that amplify gains on the rebound.
• Normal- the market is at or near all-time highs. We want our portfolio to have a normal beta at or above 1.0, which indicates elements that participate in steady growth.

The ideal portfolio sensitivity shape is asymmetric: we seek to capture more upside than downside.

Our portfolio shows the following traits:

Regime	Beta	Interpretation
Stress	1.31 ± 0.04	For each 1% market drop during a crisis, our portfolio is expected to drop ~1.31%. It amplifies stress losses.
Recovery	1.07 ± 0.06	For each 1% market rebound, our portfolio is expected to gain ~1.07%. Solid market participation.
Normal	1.05 ± 0.09	For each 1% market rise in normal conditions, our portfolio is expected to gain ~1.05%. It generally tracks the market in normal times.

Bottom line: This portfolio amplifies market moves in all regimes, with the strongest amplification during stress. In a -20% drawdown, the model-implied portfolio loss is approximately -26.2%. During a +20% recovery, the portfolio gains approximately +21.4%. Our portfolio is pro-cyclical with no crisis dampening- it rides the wave in both directions, but the wave hits harder on the way down. This is a significant weakness.

When it comes to general statistical risk measures, Souppe‘s model quickly reveals the portfolio’s weaknesses:

^{Portfolio risk profile. Source: Souppe.}

This portfolio holds 11 equally-sized positions across 5 sectors, but that’s surface-level diversification. The Diversification Ratio scores just 11 (DR = 1.11), meaning the holdings’ risk exposures overlap so heavily that combining them reduces portfolio volatility by only ~11% compared to holding each position independently. Systematic Coverage is 56- the model explains about 56% of these positions’ price movements, leaving ~44% driven by security-specific risk unrelated to broad market forces. The money is spread evenly (Position Diversity 91) and liquid (Liquidity 89), but the underlying risk exposures are heavily overlapping. Our portfolio is diversified in allocation, but not in the risks the positions are actually exposed to.

Bottom line: This portfolio amplifies market moves, especially during crises. The regime market betas tell the story directly: a Stress beta of 1.31 means that for each 1% the market drops during a drawdown, this portfolio drops ~1.31%. Recovery (1.07) and Normal (1.05) betas are close to the market- the portfolio participates in upside but doesn’t meaningfully outpace it. The asymmetry runs the wrong way: losses are amplified more than gains. We tackle this in the next section: “Improving Our Portfolio”.

Risk Metrics Notes

See notes on the risk metrics shown in the performance summary table:

• Standard deviation of returns (Std)- this statistic measures the dispersion of returns around their mean. We calculate it using daily simple returns over the past 126 trading days and then annualize the result. We get the annualized standard deviation of returns, representing the asset returns’ typical volatility over a year, so higher values point at larger day-to-day price swings ex post, and assuming a stable behavior, also ex ante. While it is well-defined for the typical return distributions, it does not capture tail risk or asymmetry (3rd and 4th distribution moments) in non-normal return distributions, which is particularly severe in less diversified portfolios. From an investor’s perspective, Std reflects the market’s level of disagreement around asset valuation- whether due to insufficient information, competing interpretations of available information, or noise.

• Conditional Value at Risk 97.5% (CVaR, expected shortfall)- this statistic quantifies tail risk. It estimates the typical average daily loss in the left tail of the return distribution, given an asset’s historical return behavior over our selected window. We calculate CVaR from daily simple returns starting on 1 January 2020. We sort these daily returns from worst to best, select the worst 2.5%, and compute their arithmetic mean. The result summarizes the asset returns’ negative tail behavior as it appears in its worst historical daily returns over this period. “n/a” means the security did not trade on 1 January 2020.

Improving Our Portfolio

Once Souppe has identified our portfolio’s weak spots, we turn to act and improve it.

Generally, when improving a portfolio, given that we have high conviction for its current constituents, we have two decisions to make:

1. What securities to add to our portfolio- Adding securities into a portfolio must always be based on wisdom. Smart investors iteratively diversify until they run out of smart ideas.
   
2. What weights shall we give to each security in the portfolio- this is a more technical step that comes to adjust our portfolio’s risk exposures. Different weighting schemes reflect different investment philosophies: equal weight treats all positions as equally important, risk parity equalizes each position’s risk contribution, and Souppe-weighted allocates proportionally to model conviction.

We will discuss ways to improve upon our portfolio using Souppe’s recommendation model. With Souppe, security selection and portfolio weighting are not independent decisions- the optimal next security depends on how the existing portfolio is weighted.

Souppe provides security suggestions based on the gaps in systematic exposure it identifies in our portfolio. The recommendations are calibrated to my risk group (medium) and target the portfolio’s structural balance across market regimes. It’s important to emphasize: Souppe does not predict returns- that remains the investor’s edge. Souppe ensures the portfolio’s risk architecture is sound and offers multiple weighting schemes to shape position sizing, so the investor can focus entirely on their conviction about which securities will deliver returns.

We ran Souppe’s model for 4 iterations for each weighting scheme, bringing our portfolio’s size to a manageable 15. In each iteration, we added the first security Souppe’s model suggested into our portfolio. Souppe will recommend adding securities that strengthen our portfolio’s weaknesses, and provide the rationale to its recommendation. Since the securities’ weights affect the portfolio’s risk exposure, every weighting scheme resulted in different recommended securities. The result is 10 portfolios with 4 added securities to our existing 11 listed above, each weighted according to a different scheme. All securities in this specific process are US listed.

The following table shows the constituent securities of each portfolio and their weights:

^{The 10 portfolio structures, as created by the Souppe model.}

The following radar plot shows the different risk profiles of the 10 portfolios we simulated:

^{The risk profiles of 10 portfolios created using our base 11 securities + 4 recommended by Souppe.}

Notice that the Souppe weighting scheme brought a tremendous improvement regime robustness, which measures how well a portfolio is expected to perform across all market environments- stress, recovery and normal.

The following table shows the risk scores of the 10 portfolios:

^{Comparing the risk profiles of the 10 suggested portfolios. Source: Souppe.}

In this simple use case of enhancing an existing 11-security portfolio’s risk profile, the minimum variance weighting scheme achieved the highest composite risk score of 77 out of 100, showing the strongest overall balance across all eight risk dimensions. All weighting schemes lifted the portfolio’s original composite score of 63.

The following table shows the 10 portfolios’ regime beta exposures:

^{Portfolio regime beta exposures. Source: Souppe. In ( )- beta estimation standard error.}

This table shows each portfolio’s market sensitivity across three regimes. For each 1% the broad market moves during Stress, Recovery, or Normal conditions, how much is each portfolio expected to move? A value above 1.0 means the portfolio is expected to amplify the market’s move, and below 1.0 means it is expected to dampen it. We use historical data to understand security and portfolio dynamics, and apply our model ex-ante.

The base portfolio of 11 equal-weighted US securities has a Diversification Ratio score of 11/100- the positions are highly correlated, offering minimal true diversification despite appearing spread across sectors. The stress beta of 1.31 means a 20% market drawdown implies a ~26% portfolio loss, with an RCR of 0.81 (81 cents of upside captured per dollar of downside).

The model recommended 4 additional securities and applied 10 different weighting methods. The best result was minimum variance: stress beta dropped to 0.97, RCR improved to 0.94, and composite risk score reached 77/100. HRP¹ was close behind (stress beta 1.04, RCR 0.89, composite 76). Both methods succeeded by reallocating weight away from the riskiest positions rather than relying solely on the new additions.

However, no portfolio achieved an RCR above 1.0. The fundamental constraint is dilution: with 11 original positions, each new security enters at ~6-7% weight. Four additions cannot materially shift the portfolio’s risk profile when the original holdings dominate. To achieve RCR above 1.0 (more upside capture than downside), two paths are available:

• Start with a smaller portfolio (e.g., 5 securities) and let the model run 10-15 recommendation steps.
• Keep the existing 11 securities and extend the model to recommend 9+ more, reaching 20 positions.

The weighting method matters as much as security selection: the same 15 securities produced composite scores ranging from 64 to 77 depending on the method. Risk-aware methods (minimum variance, HRP) consistently outperformed naive allocation (equal weight, confidence weighted).

Notice that throughout this analysis we completely ignored cumulative performance. There is no “this stock returned X% last year” anywhere in Souppe. Periodic returns serve only as statistical raw material- to measure how assets move, classify market regimes, estimate risk exposures, and quantify diversification across eight independent risk dimensions. None of that requires a view on which security will outperform next.

The portfolios shown above are notional 15-security options for enhancing our 11-security portfolio using Souppe’s model, each differing in how securities are added and weighted. The model operates entirely in the systematic risk space- factor exposures, regime sensitivities and diversification structure. It does not predict which individual security will outperform. That is another layer of analysis, which is left to the investor.

In other words, this model takes care of the logical layer of portfolio management, leaving the investor the complete autonomy to apply their wisdom and control their idiosyncratic risk exposure. Most investors believe they generate alpha, but much of their performance is driven by systematic risk they don’t see. By managing that layer explicitly, Souppe isolates the investor’s true edge- so when they’re right, it’s because of skill, not hidden bets on systematic risks.

That covers the portfolio’s risk architecture. Now for the conviction behind it- why I chose these specific securities.

Idiosyncratic Ideas

These ideas focus on companies and industries where I see a clear need for more investment and offer a favorable balance of potential return against risk. They are based on information I come across that points to opportunities worth pursuing.

Taking on idiosyncratic risk adds wisdom to the system. Results may take time to appear but often materialize in the short to medium term.

Ticker	Date	Driver	Exposure
INTC (US)	22 Aug 2025	The US government will invest around $9 billion in Intel to strengthen the American semiconductor industry. We want the US government as a partner, as government support creates unique growth opportunities and diminished risk. It is a part of a larger process of reinforcing US production capacity and tech leadership, making this a solid turnaround bet.	Long
INTC (US)	18 Sep 2025	Nvidia announced that it will buy $5 billion worth of Intel shares, a sum equal to about 17% of Nvidia’s 2024 net income of about $30 billion. This move signals confidence from a key industry peer and strengthens Intel’s position within the AI supply chain. It reinforces Intel’s strategic importance and adds upside optionality with limited downside.	Long

Idiosyncratic investing is more suitable for individuals and fast-moving sophisticated investors such as hedge funds. These investors are mainly interested in exploiting short-term distortions in the aggregate investor’s perception. Institutional investors, on the other hand, employ patient capital. They are the ones quietly financing the economy, and are more focused on identifying the next sectors that need funding in order to meet future demand. This is where Exchange Traded Funds (ETFs) and their likes come in.

ETF Ideas

ETFs offer a bridge between the idiosyncrasy of individual assets, and the systematic exposure of sectors and industries. They make a balance between adding wisdom and logic into the system. Here are some investment ideas and their rationales:

Ticker	Date	Driver	Exposure
ITA (US)	18 Sep 2025	With rising global geopolitical tensions, governments are committing larger portions of their budgets to defense and security. Global military expenditure reached a record $2.7 trillion in 2024, up nearly 9.3% year over year, reflecting the rise of global geopolitical temperature. This sustained increase in spending supports the entire defense sector and stabilizes revenues across its key holdings.	Long
ITA (US)	25 Sep 2025	NATO’s new 5% defense spending goal formalizes military budgets as a structural, long-term component of member economies. It is more than double the previous goal of 2%. It is projected to lift the alliance’s annual defense budgets from $1.5 trillion in 2024 to about $4.2 trillion by 2035, a 60% increase in spending. This turns demand for defense into a predictable stream, supporting long-term sector growth.	Long
PAVE (US) | HWAY (US)	25 Sep 2025	The US President announced trillions of dollars in capital commitments for foreign direct investment into the US, which stood at $308 billion in 2024. This marks a significant projected annual increase. Part of this capital will flow into domestic infrastructure renewal, benefiting companies involved in construction, materials, and logistics. These ETFs provide diversified exposure to the development phase of infrastructure.	Long
SRVR (US) | DTCR (US)	26 Sep 2025	Data centers became an investable asset class in the 2010s as cloud and storage demand accelerated. Since 2022, generative AI has driven a new wave of investment in facilities that require far greater power and capital. These products provide access to the data-center infrastructure theme, ranging from real-estate owners and operators to the broader ecosystem of global infrastructure and service providers. Notice that data-center real-estate assets traded at implied cap rates of about 4.4% as at mid-2025, reflecting high valuations with a reasonable yield.	Long
SPY (US)	30 Sep 2025	The US publicly traded capital markets remain the world’s most reliable savings medium. As long as global economic growth continues and US credit risk stays low, capital inflows will persist, supporting long-term equity appreciation. The large scale adoption of AI is projected to increase productivity throughout the market. It is projected to contribute about 0.6-1.5 percentage points to annual productivity growth, providing an additional force of long-term support for the S&P 500.	Long
ISF (UK) | VUKE (UK)	30 Sep 2025	Rising political and social instability threatens the long-term attractiveness of the UK capital markets. Uncertainty around policy direction and investor confidence increases the risk premium and uncertainty around local equities.	Avoid
AIPO (US)	1 Oct 2025	The growing need for AI data centers is driving a parallel surge in energy and infrastructure demand. While data centers represent the visible frontend of this trend, the less visible backend- power generation, cooling, and transmission- requires substantial capital investment. As AI-focused facilities require 2–3× the power density and capital intensity of traditional ones, this ETF provides exposure to this infrastructure layer.	Long
ARGT (US)	28 Oct 2025	In 2023, Javier Milei won Argentina’s presidency, vowing to dismantle the failed economic system that had long forced the country to depend on international aid and debt financing. Many doubted whether the public would be willing to break from the false promises of the past and withstand the pain required for such systemic change. On 27 October 2025, Milei’s party won the midterm legislative elections with 40.8% of the vote, demonstrating broad and durable support for his agenda. As Argentine society endures this test, we gain greater confidence in the country’s path forward.	Long
REMX (US) | SETM (US)	25 Nov 2025	China dominates the rare-earth supply chain at both the extraction and processing stages, but its control is far more decisive in the chemical separation (processing) layer, where it holds nearly all global capacity. As demand for technologies that depend on rare-earth elements rises, particularly those requiring high-performance permanent magnets, this processing chokepoint becomes a strategic weakness for the Western bloc. The US and Australia are developing alternative processing hubs, but building a complete and scalable processing capacity takes years and requires multiple and complex technical stages to mature. Much of this capacity is likely to develop in countries with vast desert or tundra regions, where large-scale chemical processing faces fewer land-use and environmental constraints. Furthermore, new facilities remain exposed to Chinese price pressure, which undercuts competitors before they scale. ETFs with exposure to critical minerals, strategic materials, and industrial-policy beneficiaries may gain as Western governments and investors direct capital into rare-earth processing capacity.	Long

Systematic forces shape the backdrop, while idiosyncratic ideas identify where those forces clearly express themselves in specific assets.

The line between “passive” and “active” investing is drawn by the individual’s thinking effort. If you think on a regular basis and act upon your thinking, then you are an active investor. Otherwise, you are passive.

Evolving Indicators

These are short to mid-term indicators that capture emerging shifts in the economic environment, which offer an opportunity for ongoing research and refinement.

Indicator	Meaning
The AI trade	AI adoption is beginning to reshape sectoral performance by widening the gap between industries that can utilize smart automation, and those that rely on manual, repetitive data-related tasks. Capital is flowing toward sectors where AI directly enhances productivity, such as semiconductors, cloud and data infrastructure, finance, healthcare, logistics, advanced manufacturing, and energy systems supporting the demand for computing power. At the same time, industries built on routine cognitive or labor-intensive work, such as administrative services, customer support, basic accounting and legal tasks, traditional media, retail, and warehousing, are showing early signs of pressure and role displacement. Tracking this indicator helps identify where AI is creating new competitive advantages and where structural headwinds are forming across the economy.
An AI bubble	There is a feeling that the valuations of companies operating in the AI field, from model developers to infrastructure providers, are far too high relative to what their eventual value might justify. This feeling reflects a common human tendency to exaggerate early potential: to overestimate excitement and overlook the impact of competition. FOMO is the basic ingredient for bubbles. However, AI model development displays winner-take-all dynamics, where expensive models can become obsolete overnight if a rival advances, creating a risky environment for investors. This is because when a model becomes obsolete, its parameters become useless, forcing the developer to go back to the architecture drawing board. This may help explain why AI companies often invest in each other, as a way to stabilize their positions and spread risk in this quickly-developing landscape. The strong competition in the field of model development may prove dangerous if overlooked by investors.
Bipolar global geopolitics	The post-Cold War era of uncontested US leadership is giving way to a bipolar system, with the Eastern bloc of China, Russia and other countries increasingly willing to contest US dominance. This shift carries deep economic implications for the coming decade, as the US accelerates efforts to re-anchor manufacturing at home, protect its technological lead, and fund a military posture that now requires far greater investment. Over time, this environment tests the US dollar’s role as the world’s sole reserve currency. For investors, this evolving indicator may point to a gradual move away from frictionless globalization toward “bloc-based” globalization- clusters of countries organizing trade, standards, and capital flows around competing security and technology spheres.

Structural Themes

The following table records systematic investment ideas, built on the slower, longer-term forces shaping the societies and economies.

By taking on systematic risk, we add logic to the system.

Theme	Thesis
American Production Renewal	Even before the CHIPS and Science Act and Inflation Reduction Act of 2022, the US government has been reinforcing domestic technology and manufacturing leadership. These large scale incentives exceed $1 trillion over 10 years. This policy direction favors industries that combine advanced production capabilities with innovation, supporting a long-term reindustrialization.
Water scarcity	Climate change and population growth are intensifying global water demand while diminishing supply, particularly across the Middle East and Africa. With global water demand projected to exceed supply by nearly 40% by 2030, scarcity is becoming permanent. Energy, utilities and water-technology firms will require increasing investments to create a steady supply of water.
High sovereign debt levels	Various governments borrowed heavily In the 2020 COVID emergency, and these high debt levels still persist to this day (See the US, UK and France). Most importantly, the US government debt to GDP stands at 124% (2024). With increasing deficits, the focus is quickly shifting to avoiding financial distress and the costs it brings, and even bankruptcy, at all costs. The US government is acting to achieve this through various means of keeping value within the country. This environment drives more aggressive diplomacy and deeper government involvement in the economy, harming globalization and fueling a “race to the bottom” between countries.
Aging US infrastructure	Much of the US infrastructure was built largely in the 20th century. Older facilities require higher maintenance costs and possible replacements. The American Society of Civil Engineers estimates a $3.7 trillion funding gap through 2035 to modernize national infrastructure. Together with high national debt levels, this translates to higher value for private-sector capital providers, as well as construction companies.
Increased demand for energy	Humanity’s evolution from the Information Age to the Knowledge Age, driven by artificial intelligence, will require vast investments in infrastructure such as energy generation, transport, and data centers. Global electricity demand is projected to rise by 75% by 2050, underscoring the need for large-scale energy expansion across advanced and emerging economies.
AI reshaping societies	Artificial intelligence is restructuring value creation by automating routine cognitive work, and lifting productivity across the economy. As adoption rises, AI is set to generate trillions of dollars in incremental value output, driven by significant investment in computing capacity, data infrastructure, and automated workflows. These gains accrue primarily to societies capable of deploying AI at scale and controlling proprietary data. Labor-intensive industries will experience significant disruption. The result is a structural reorganization of society in which economic value increasingly depends on AI-driven systems. This theme is already taking place, as junior hiring in tech has mostly stopped, since these young employees’ work is easy to automate. The next stage is not a wave of layoffs, but a gradual shift: more experienced workers and managers will see parts of their workflow absorbed by AI systems, reducing the volume of human effort required even in higher-skilled roles. This will require major changes in how our societies distribute value.
Low birth rates	For roughly two decades, most developed societies have faced birth rates well below the 2.1 children per woman replacement level. As populations age and the share of working-age citizens shrinks, governments that want to preserve living standards face three broad paths: They can increase immigration, which may create social and political strain. They can try to raise birth rates, which so far has proved difficult in secular societies. Or they can invest heavily in robotics and automated processes, using capital and technology to raise productivity in a dwindling workforce. The more societies choose this third path, the stronger and more persistent the structural demand for automation, advanced manufacturing, and productivity-enhancing infrastructure.
The “Jewish Trade”	Throughout history, the safety, stability and well-being of Jewish communities have closely tracked the underlying health of societies. Rising antisemitism or social scapegoating often signals institutional stress, social fragmentation, and deteriorating rule of law, while episodes of Jewish exodus typically coincide with broader economic decline as human capital and financial networks weaken. Conversely, environments where Jews feel secure and integrated tend to reflect openness, innovation capacity, and resilient governance. Tracking these shifts provides an early indicator of long-term structural societal strength or fragility, which eventually surface in economic performance.

These notes evolve over time as logic and data develop. They are not recommendations, but personal observations and pieces of ongoing research.

Disclaimer: this page and website are for informational and educational purposes only and do not constitute investment advice, financial advice, or any other type of professional advice. The analysis presented is based on publicly available information. Readers must conduct their own independent research and due diligence before making any investment decisions. Past investment activity is not indicative of future performance. Consult with a qualified financial advisor before acting on any information presented here.

1. López de Prado, M. (2016). Building Diversified Portfolios that Outperform Out of Sample. Journal of Portfolio Management. ↩︎