arXiv:2510.10807v3 Announce Type: replace 
Abstract: We examine whether regime-conditioned generative scenarios combined with a convex CVaR allocator improve portfolio decisions under regime shifts. We present MARCD, a generative-to-decision framework with: (i) a Gaussian HMM to infer latent regimes; (ii) a diffusion generator that produces regime-conditioned scenarios; (iii) signal extraction via blended, shrunk moments; and (iv) a governed CVaR epigraph quadratic program. Contributions: Within the Scenario stage we introduce a tail-weighted diffusion objective that up-weights low-quantile outcomes relevant for drawdowns and a regime-expert (MoE) denoiser whose gate increases with crisis posteriors; both are evaluated end-to-end through the allocator. Under strict walk-forward on liquid multi-asset ETFs (2005-2025), MARCD exhibits stronger scenario calibration and materially smaller drawdowns: MaxDD 9.3% versus 14.1% for BL (a 34% reduction) over 2020-2025 out-of-sample. The framework provides an auditable pipeline with explicit budget, box, and turnover constraints, demonstrating the value of decision-aware generative modeling in finance.

تقديم إطار العمل Multi-Agent Regime-Conditioned Diffusion (MARCD) يمثل تقدمًا كبيرًا في إدارة المحافظ، خاصة في التنقل بين التغيرات في الأنظمة. من خلال دمج السيناريوهات التوليدية مع مُخصص CVaR، يُحسن MARCD عمليات اتخاذ القرار، مما يسمح للمستثمرين بالتكيف بشكل أفضل مع ظروف السوق المتغيرة. هذه الابتكار مهم لأنه لا يُحسن فقط من قوة استراتيجيات المحافظ، بل يتناول أيضًا تعقيدات إدارة المخاطر في البيئات المتقلبة.

La introducción del marco Multi-Agent Regime-Conditioned Diffusion (MARCD) marca un avance significativo en la gestión de carteras, especialmente en la navegación de cambios de régimen. Al integrar escenarios generativos con un asignador CVaR, MARCD mejora los procesos de toma de decisiones, permitiendo a los inversores adaptarse mejor a las condiciones cambiantes del mercado. Esta innovación es crucial, ya que no solo mejora la solidez de las estrategias de cartera, sino que también aborda las complejidades de la gestión de riesgos en entornos volátiles.

L'introduction du cadre Multi-Agent Regime-Conditioned Diffusion (MARCD) représente une avancée significative dans la gestion de portefeuille, en particulier pour naviguer dans les changements de régime. En intégrant des scénarios génératifs avec un allocateur CVaR, MARCD améliore les processus de prise de décision, permettant aux investisseurs de mieux s'adapter aux conditions de marché changeantes. Cette innovation est cruciale car elle améliore non seulement la robustesse des stratégies de portefeuille, mais elle aborde également les complexités de la gestion des risques dans des environnements volatils.

The introduction of the Multi-Agent Regime-Conditioned Diffusion (MARCD) framework marks a significant advancement in portfolio management, particularly in navigating regime shifts. By integrating generative scenarios with a CVaR allocator, MARCD enhances decision-making processes, allowing investors to better adapt to changing market conditions. This innovation is crucial as it not only improves the robustness of portfolio strategies but also addresses the complexities of risk management in volatile environments.

Multi-Agent Regime-Conditioned Diffusion (MARCD) for CVaR-Constrained Portfolio Decisions

The Apple Watch Series 10 might be last year's model, but its specs prove it's basically as capable as the Series 11. The main difference? It's $120 off.

ساعة آبل Series 10، على الرغم من أنها ليست الطراز الأحدث، تقدم قدرات مثيرة للإعجاب تنافس الطراز الجديد Series 11، وهي متاحة حاليًا بخصم كبير قدره 120 دولارًا. وهذا يجعلها خيارًا جذابًا لأولئك الذين يبحثون عن ساعة ذكية عالية الجودة دون إنفاق الكثير. تضمن ميزاتها وأدائها أن يتمكن المستخدمون من الاستمتاع بتجربة متميزة، مما يجعلها صفقة ذكية.

El Apple Watch Series 10, aunque no es el modelo más reciente, ofrece capacidades impresionantes que rivalizan con la nueva Series 11, y actualmente está disponible con un descuento significativo de 120 $. Esto lo convierte en una opción atractiva para quienes buscan un reloj inteligente de alta calidad sin gastar demasiado. Sus características y rendimiento aseguran que los usuarios puedan disfrutar de una experiencia premium, lo que lo convierte en una compra inteligente.

La montre Apple Series 10, bien qu'elle ne soit pas le dernier modèle, offre des capacités impressionnantes qui rivalisent avec la nouvelle Series 11, et elle est actuellement disponible à une remise significative de 120 $. Cela en fait une option attrayante pour ceux qui recherchent une montre intelligente de haute qualité sans se ruiner. Ses caractéristiques et performances garantissent aux utilisateurs une expérience premium, ce qui en fait un achat judicieux.

The Apple Watch Series 10, although not the latest model, offers impressive capabilities that rival the newer Series 11, and it's currently available at a significant discount of $120. This makes it an attractive option for those looking for a high-quality smartwatch without breaking the bank. Its features and performance ensure that users can enjoy a premium experience, making it a smart buy.

One of the best Apple Watches you can buy isn't Apple's newest (but it's 30% off)

The proposal received more than 75% approval. Musk can earn the $1 trillion in shares over the next 10 years if Tesla hits a number of milestones.

وافق مساهمو تسلا بأغلبية ساحقة على حزمة تعويض غير مسبوقة بقيمة تريليون دولار للرئيس التنفيذي إيلون ماسك، حيث صوت أكثر من 75% لصالحها. يتيح هذا البرنامج الطموح لماسك كسب هذه المبالغ الضخمة من الأسهم على مدى العقد المقبل، بشرط أن تحقق تسلا أهدافًا معينة. تعكس هذه الخطوة ثقة المساهمين في قيادة ماسك وتبرز أيضًا الإمكانات الكبيرة لنمو الشركة في صناعة السيارات المتطورة بسرعة.

Los accionistas de Tesla han aprobado abrumadoramente un innovador paquete de compensación de 1 billón de dólares para el CEO Elon Musk, con más del 75% de los votos a favor. Este ambicioso plan permite a Musk ganar esta enorme suma en acciones durante la próxima década, siempre que Tesla logre hitos específicos. Esta decisión no solo refleja la confianza de los accionistas en el liderazgo de Musk, sino que también subraya el potencial de crecimiento de la empresa en la industria automotriz en rápida evolución.

Les actionnaires de Tesla ont largement approuvé un plan de rémunération révolutionnaire de 1 trillion de dollars pour le PDG Elon Musk, avec plus de 75 % des votes en faveur. Ce plan ambitieux permet à Musk de gagner cette somme colossale en actions au cours de la prochaine décennie, sous réserve que Tesla atteigne des objectifs spécifiques. Cette décision reflète non seulement la confiance des actionnaires dans le leadership de Musk, mais souligne également le potentiel de croissance de l'entreprise dans l'industrie automobile en pleine évolution.

Tesla shareholders have overwhelmingly approved a groundbreaking $1 trillion pay package for CEO Elon Musk, with over 75% voting in favor. This ambitious plan allows Musk to earn the massive sum in shares over the next decade, contingent on Tesla achieving specific milestones. This decision not only reflects the confidence shareholders have in Musk's leadership but also underscores the company's growth potential in the rapidly evolving automotive industry.

Tesla shareholders approve Elon Musk’s $1T pay package

يدعم مستثمرو تسلا إيلون ماسك بينما يقترب من الحصول على مكافأة بقيمة تريليون دولار، مما يبرز النمو الملحوظ للشركة وثقة المساهمين في مستقبلها. لا يسلط هذا الإنجاز الضوء فقط على تأثير ماسك في صناعة التكنولوجيا، بل يعكس أيضًا الأداء القوي لأسهم تسلا، التي ارتفعت بسبب التقدمات الابتكارية وزيادة الطلب على السيارات الكهربائية. مع استمرار تطور السوق، فإن هذا التطور مهم لكل من المستثمرين وقطاع السيارات بشكل عام.

Los inversores de Tesla están apoyando a Elon Musk mientras se acerca a un posible pago de un billón de dólares, destacando el notable crecimiento de la compañía y la confianza que tienen los accionistas en su futuro. Este hito no solo subraya la influencia de Musk en la industria tecnológica, sino que también refleja el sólido rendimiento de las acciones de Tesla, que han aumentado debido a avances innovadores y a la creciente demanda de vehículos eléctricos. A medida que el mercado continúa evolucionando, este desarrollo es significativo tanto para los inversores como para el sector automotriz en general.

Les investisseurs de Tesla soutiennent Elon Musk alors qu'il se rapproche d'un potentiel paiement d'un trillion de dollars, mettant en lumière la croissance remarquable de l'entreprise et la confiance des actionnaires dans son avenir. Ce jalon souligne non seulement l'influence de Musk dans l'industrie technologique, mais reflète également la forte performance de l'action de Tesla, qui a grimpé en raison des avancées innovantes et de la demande croissante pour les véhicules électriques. Alors que le marché continue d'évoluer, ce développement est significatif tant pour les investisseurs que pour le secteur automobile au sens large.

Tesla investors are rallying behind Elon Musk as he approaches a potential trillion-dollar payday, highlighting the company's remarkable growth and the confidence shareholders have in its future. This milestone not only underscores Musk's influence in the tech industry but also reflects the strong performance of Tesla's stock, which has surged due to innovative advancements and increasing demand for electric vehicles. As the market continues to evolve, this development is significant for both investors and the broader automotive sector.

Tesla Investors Back Musk’s Trillion Dollar Payday

Slingshots AI grants provide resources that would be unavailable in most academic settings. Its debut program has 15 startups focused on AI evaluation.

أطلق معهد لاود دفعة الأولى من منح 'Slingshots' للذكاء الاصطناعي، والتي تهدف إلى توفير موارد أساسية للشركات الناشئة التي غالبًا ما تفتقر إلى الوصول إلى مثل هذا الدعم في البيئات الأكاديمية التقليدية. يتضمن هذا البرنامج الأول 15 شركة ناشئة مبتكرة تركز على تقييم الذكاء الاصطناعي، مما يمثل خطوة مهمة في تعزيز التقدم التكنولوجي وريادة الأعمال في قطاع الذكاء الاصطناعي.

El Instituto Laude ha lanzado su primer lote de subvenciones 'Slingshots' en IA, destinadas a proporcionar recursos esenciales a las startups que a menudo carecen de acceso a este tipo de apoyo en entornos académicos tradicionales. Este programa inaugural cuenta con 15 startups innovadoras centradas en la evaluación de IA, marcando un paso significativo en la promoción del avance tecnológico y el emprendimiento en el sector de la IA.

L'Institut Laude a lancé son premier lot de subventions 'Slingshots' en IA, visant à fournir des ressources essentielles aux startups qui n'ont souvent pas accès à un tel soutien dans les environnements académiques traditionnels. Ce programme inaugural comprend 15 startups innovantes axées sur l'évaluation de l'IA, marquant une étape importante dans la promotion de l'avancement technologique et de l'entrepreneuriat dans le secteur de l'IA.

The Laude Institute has launched its first batch of 'Slingshots' AI grants, aimed at providing essential resources to startups that often lack access to such support in traditional academic environments. This inaugural program features 15 innovative startups focused on AI evaluation, marking a significant step in fostering technological advancement and entrepreneurship in the AI sector.

Laude Institute announces first batch of ‘Slingshots’ AI grants

تعمل OpenAI بنشاط على معالجة المخاوف المتعلقة باستقرارها المالي في سعيها لطمأنة المستثمرين وأصحاب المصلحة. هذا مهم لأن الشفافية المالية ضرورية للحفاظ على الثقة والدعم في المشهد التكنولوجي سريع التطور، خاصةً لشركة في طليعة الذكاء الاصطناعي.

OpenAI está abordando activamente las preocupaciones sobre su estabilidad financiera mientras busca tranquilizar a los inversores y partes interesadas. Esto es significativo porque la transparencia financiera es crucial para mantener la confianza y el apoyo en el paisaje tecnológico en rápida evolución, especialmente para una empresa a la vanguardia de la inteligencia artificial.

OpenAI s'efforce de répondre aux préoccupations concernant sa stabilité financière afin de rassurer les investisseurs et les parties prenantes. Cela est important car la transparence financière est essentielle pour maintenir la confiance et le soutien dans le paysage technologique en évolution rapide, en particulier pour une entreprise à la pointe de l'intelligence artificielle.

OpenAI is actively addressing concerns regarding its financial stability as it seeks to reassure investors and stakeholders. This is significant because financial transparency is crucial for maintaining trust and support in the rapidly evolving tech landscape, especially for a company at the forefront of artificial intelligence.

OpenAI Races to Quell Concerns Over Its Finances

<ol>
<li>Stored Procedure</li>
</ol>

A stored procedure is a precompiled collection of SQL statements (and sometimes control logic) stored in the database. It’s like a function that can be called whenever needed. It can accept parameters such as input and output, and is also often used for complex logic, such as loops.

<a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F4wyd9942rpxzrwhosz5k.JPG" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F4wyd9942rpxzrwhosz5k.JPG" alt=" " width="644" height="238"></a>

<ol>
<li>CTE (Common Table Expression)
A CTE is a temporary named result set you can use within a single query. Think of it as a short-term, readable alias for a subquery. It exists only during query execution. CTE's make complex queries easier to read and maintain.</li>
</ol>

<a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fkg77hkg9jdbc751f22i3.JPG" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fkg77hkg9jdbc751f22i3.JPG" alt=" " width="640" height="167"></a>

<ol>
<li>Subquery
A subquery is a query nested inside another query; in other words, it is like a query within a query. It can appear within the SELECT, FROM, or WHERE clause, and it executes each time the main query is run.</li>
</ol>

<a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fjimps1x95onye5z63dfo.JPG" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fjimps1x95onye5z63dfo.JPG" alt=" " width="368" height="172"></a>

فهم الفروق بين الإجراءات المخزنة، والاستعلامات الفرعية، والتعبيرات الجدولية الشائعة (CTE) أمر ضروري لأي شخص يعمل مع قواعد البيانات. الإجراءات المخزنة هي مجموعات مسبقة التجميع من عبارات SQL يمكن تنفيذها حسب الحاجة، وغالبًا ما تستخدم للمنطق المعقد. تتيح الاستعلامات الفرعية إجراء استعلامات داخل استعلام آخر، بينما توفر CTE وسيلة لتعريف مجموعات نتائج مؤقتة يمكن الإشارة إليها ضمن عبارة SELECT أو INSERT أو UPDATE أو DELETE. معرفة متى وكيفية استخدام هذه الأدوات يمكن أن تعزز بشكل كبير من أداء وكفاءة قاعدة البيانات.

Entender las diferencias entre los procedimientos almacenados, las subconsultas y las Expresiones de Tabla Comunes (CTE) es esencial para cualquier persona que trabaje con bases de datos. Los procedimientos almacenados son colecciones precompiladas de sentencias SQL que se pueden ejecutar según sea necesario, a menudo utilizadas para lógica compleja. Las subconsultas permiten realizar consultas dentro de otra consulta, mientras que las CTE proporcionan una forma de definir conjuntos de resultados temporales que se pueden referenciar dentro de una instrucción SELECT, INSERT, UPDATE o DELETE. Saber cuándo y cómo usar estas herramientas puede mejorar significativamente el rendimiento y la eficiencia de la base de datos.

Comprendre les différences entre les procédures stockées, les sous-requêtes et les expressions de table communes (CTE) est essentiel pour quiconque travaille avec des bases de données. Les procédures stockées sont des collections précompilées d'instructions SQL qui peuvent être exécutées selon les besoins, souvent utilisées pour une logique complexe. Les sous-requêtes permettent de faire des requêtes à l'intérieur d'une autre requête, tandis que les CTE offrent un moyen de définir des ensembles de résultats temporaires qui peuvent être référencés dans une instruction SELECT, INSERT, UPDATE ou DELETE. Savoir quand et comment utiliser ces outils peut améliorer considérablement les performances et l'efficacité des bases de données.

Understanding the differences between stored procedures, subqueries, and Common Table Expressions (CTEs) is essential for anyone working with databases. Stored procedures are precompiled collections of SQL statements that can be executed as needed, often used for complex logic. Subqueries allow for querying within another query, while CTEs provide a way to define temporary result sets that can be referenced within a SELECT, INSERT, UPDATE, or DELETE statement. Knowing when and how to use these tools can significantly enhance database performance and efficiency.

Explain the difference between subquery,CTE and stored procedures

Nancy Pelosi’s trades over the years have been so good that a startup was created to allow investors to directly mirror her portfolio.

أعلنت نانسي بيلوسي، المعروفة بمهاراتها الاستثنائية في التداول، عن تقاعدها، مما يمثل نهاية حقبة للمستثمرين في وول ستريت. كانت تداولاتها ناجحة لدرجة أنه تم إنشاء شركة ناشئة لمساعدة الآخرين على تقليد استراتيجياتها الاستثمارية. هذه الأخبار مهمة لأنها تسلط الضوء على تأثير المستثمرين الأفراد على السوق وتثير تساؤلات حول مستقبل استراتيجيات الاستثمار في غيابها.

Nancy Pelosi, conocida por sus excepcionales habilidades de trading, ha anunciado su jubilación, marcando el final de una era para los inversores de Wall Street. Sus operaciones han sido tan exitosas que se creó una startup para ayudar a otros a imitar sus estrategias de inversión. Esta noticia es significativa ya que destaca el impacto de los inversores individuales en el mercado y plantea preguntas sobre el futuro de las estrategias de inversión en su ausencia.

Nancy Pelosi, reconnue pour ses compétences exceptionnelles en trading, a annoncé sa retraite, marquant la fin d'une ère pour les investisseurs de Wall Street. Ses transactions ont été si réussies qu'une startup a été créée pour aider les autres à imiter ses stratégies d'investissement. Cette nouvelle est importante car elle souligne l'impact des investisseurs individuels sur le marché et soulève des questions sur l'avenir des stratégies d'investissement en son absence.

Nancy Pelosi, renowned for her exceptional trading skills, has announced her retirement, marking the end of an era for Wall Street investors. Her trades have been so successful that a startup was established to help others mirror her investment strategies. This news is significant as it highlights the impact of individual investors on the market and raises questions about the future of investment strategies in her absence.

One of the Greatest Wall Street Investors of All Time Announces Retirement

Three years ago, I didn't know what HTML stood for. Today, I have a portfolio with 5+ full stack projects and just landed my first freelance client.

My Portfolio - <a href="https://sahil-sahu-portfolio.vercel.app/" rel="noopener noreferrer">click here</a> 
Follow me on X - <a href="https://x.com/SahilSahu731" rel="noopener noreferrer">click here</a>

This is everything I wish someone told me when I started.
<h2>
 
 
 The Beginning (Age 15)
</h2>

It started with Minecraft mods. Seriously.

I wanted to create custom items for my server, found a YouTube tutorial, and copied some code I didn't understand. It didn't work. I spent 6 hours figuring out why.

When it finally worked, I felt like a wizard.

That's when I realized: I could BUILD things. Not just use apps - actually create them.

I was hooked.
<h2>
 
 
 Year 1: The Messy Start (Age 15-16)
</h2>

What I learned: HTML, CSS, JavaScript basics

What actually happened:

I jumped around. A LOT.

<ul>
<li>Week 1: "I'll learn Python!"</li>
<li>Week 2: "Actually, web dev seems cool"</li>
<li>Week 3: "Wait, should I learn C++ for games?"</li>
<li>Week 4: Back to web dev</li>
</ul>

My first website was... terrible. Like, Comic Sans terrible. Centered everything terrible. But I made it myself and put it on the internet. That felt incredible. 


<div class="highlight js-code-highlight">
<pre class="highlight html"><code>&lt;!-- My actual first website, no joke --&gt;
&lt;html&gt;
 &lt;body bgcolor="black"&gt;
 &lt;center&gt;
 &lt;font color="lime" face="Comic Sans MS"&gt;
 &lt;h1&gt;Welcome to my EPIC Website!!!&lt;/h1&gt;
 &lt;marquee&gt;This is so cool!!!&lt;/marquee&gt;
 &lt;/font&gt;
 &lt;/center&gt;
 &lt;/body&gt;
&lt;/html&gt;
</code></pre>

</div>



I thought I was a genius. Looking back... yikes.

<h2>
 
 
 The Turning Point
</h2>

Around month 4, I built a simple to-do app following a tutorial. Then I tried to add my own feature - a dark mode toggle.

I spent 2 days on it. Stack Overflow became my second home. But when that toggle finally worked?

That's when I learned the most important lesson: You learn by breaking things and fixing them.

Following tutorials is fine for learning syntax. But real learning happens when you go off-script.

<h2>
 
 
 Year 2: Getting Serious (Age 16-17)
</h2>

What I learned: React, Node.js, MongoDB, Git

This was the "grind" year. I was that kid coding during lunch break. My friends thought I was weird. My parents worried I was on my computer too much.

But I was building real stuff:

<h3>
 
 
 Project 1: Study Timer App
</h3>

<ul>
<li>Built with vanilla JS (before I knew React)</li>
<li>Pomodoro technique timer</li>
<li>Saved data to localStorage</li>
<li>Looked ugly but WORKED</li>
</ul>

What I learned: DOM manipulation, local storage, JavaScript events

<h3>
 
 
 Project 2: Weather App
</h3>

<ul>
<li>My first API integration</li>
<li>Used OpenWeather API</li>
<li>Learned about async/await the hard way (callback hell nightmares)</li>
<li>Actually looked decent with CSS Grid</li>
</ul>

What I learned: APIs, async JavaScript, responsive design

<h3>
 
 
 Project 3: Movie Database
</h3>

<ul>
<li>First React project</li>
<li>TMDB API integration</li>
<li>Search, filter, favorites</li>
<li>Deployed on Netlify</li>
</ul>

What I learned: React basics, state management, component lifecycle

<h3>
 
 
 Project 4: Chat App
</h3>

<ul>
<li>First full-stack project</li>
<li>Node.js + Express backend</li>
<li>Socket.io for real-time chat</li>
<li>MongoDB for storing messages</li>
<li>Completely broke in production multiple times</li>
</ul>

What I learned: Backend, databases, deployment, that production is DIFFERENT from localhost

<h2>
 
 
 The Struggles Nobody Talks About
</h2>

Imposter Syndrome Hit HARD

Every YouTube tutorial made it look easy. Every dev on Twitter seemed to know everything. I felt stupid constantly.

I'd see 12-year-olds building AI apps and think "I'm already behind."

Reality check: Those kids probably followed a tutorial too. Everyone's faking it a bit.

Tutorial Hell Was Real

For 3 months, I just followed tutorials. Built the same to-do app 15 different ways. Never built anything original.

Breaking free was hard. I had to force myself: "Build something without a tutorial."

It was messy. It took forever. But that's when I actually learned.

Burnout at 16

Around month 14, I hit a wall. Coded for 8 hours, made zero progress, felt like I'd never be good enough.

I took a week off. Just played games, hung out with friends. Came back refreshed.

Lesson: Breaks aren't weakness. They're necessary.

<h2>
 
 
 Year 3: Portfolio &amp; Real Projects (Age 17-18)
</h2>

What I learned: Next.js, TypeScript, PostgreSQL, Tailwind, Docker

This year was about leveling up. Making things that actually looked professional.

<h3>
 
 
 The Portfolio
</h3>

I rebuilt my portfolio 4 times. Each time I learned something new and thought "I need to redo everything."

Version 4 finally felt right:

<ul>
<li>Next.js for performance</li>
<li>Tailwind for styling (game changer)</li>
<li>Framer Motion for animations</li>
<li>Dark mode (obviously)</li>
<li>Actually mobile responsive (shocking, I know)</li>
</ul>

<h3>
 
 
 Real Projects I'm Proud Of
</h3>

*1. FileFlow *

<ul>
<li>User Authentication with Email Verification</li>
<li>Secure File Uploading and Storage</li>
<li>Preview and Download Files</li>
<li>Delete and Manage Uploaded Files</li>
<li>Responsive and User-Friendly Interface</li>
<li>JWT-based Authentication &amp; Authorization</li>
<li>Drag-and-Drop File Upload</li>
<li>Cloud Storage Integration with File Validation</li>
</ul>

2. JobTube

<ul>
<li>User &amp; Employer Authentication (JWT)</li>
<li>Role-based Dashboards (Job Seeker &amp; Employer)</li>
<li>Job Listings with Search &amp; Filter Functionality</li>
<li>Application Submission &amp; Tracking</li>
<li>Admin Panel to Manage Users, Jobs &amp; Applications</li>
<li>Responsive &amp; Accessible UI (Tailwind CSS)</li>
<li>Backend API with Secure CRUD Operations</li>
</ul>

*3. Framely *

<ul>
<li>User Authentication with JWT and secure cookies</li>
<li>Full Profile Customization (profile picture, bio, website)</li>
<li>Follow/Unfollow user system</li>
<li>Post Creation with image uploads to Cloudinary</li>
<li>Interactive Posts with likes, comments, and post deletion</li>
<li>Ephemeral Stories that automatically expire after 24 hours</li>
<li>Real-Time Notifications for likes, comments, and follows using Socket.IO</li>
<li>Real-Time Direct Messaging (Chat) between users</li>
<li>Dynamic Explore Page to discover popular content</li>
<li>Saved Posts feature for bookmarking content</li>
<li>User Search functionality to find and connect with others</li>
<li>Responsive and Intuitive UI with skeleton loaders for a smooth experience</li>
</ul>

4. Growth AI

<ul>
<li>AI-Powered Conversational Interface</li>
<li>Real-time Response Rendering</li>
<li>Multiple Chat Sessions</li>
<li>Clean and Modern UI</li>
<li>Responsive Design for All Devices</li>
<li>Secure API Integration</li>
<li>Persistent Chat History (Local Storage or DB)</li>
</ul>

<h2>
 
 
 What Actually Helped Me Learn
</h2>

1. Build Projects, Not Just Follow Tutorials

After basics, I learned by building. Every new concept = new project.

Want to learn React? Build something in React. Not another todo app. Something YOU want to use.

2. Document Everything

I kept a dev journal. Every time I solved a hard problem, I wrote it down.

Now when I hit the same problem, I have my own notes. Way better than googling the same thing 50 times.

3. Read Other People's Code

GitHub is free education. Find projects similar to what you're building, read the code.

I learned more from reading Next.js source code than any tutorial taught me.

4. Don't Learn Everything At Once

I wasted months trying to learn:

<ul>
<li>Frontend frameworks</li>
<li>Backend</li>
<li>DevOps</li>
<li>Databases</li>
<li>All at the same time</li>
</ul>

Learn ONE thing deeply, then add layers.

My path: HTML/CSS → JavaScript → React → Backend → Databases

5. Deploy Your Projects

Code on localhost isn't real. Deploy it.

Netlify, Vercel, Railway - they're free. Deploy everything.

You'll learn about environment variables, CORS, HTTPS, and all the stuff tutorials skip.

6. Join Communities

Discord servers, Reddit, dev.to - other developers are helpful.

I was scared to ask "dumb questions" for so long. Turns out, everyone has the same questions.

<h2>
 
 
 The Tech Stack I Use Now
</h2>

Frontend:

<ul>
<li>React / Next.js (for everything)</li>
<li>Tailwind CSS (never going back to vanilla CSS)</li>
<li>TypeScript (game changer for big projects)</li>
</ul>

Backend:

<ul>
<li>Node.js + Express (simple, works)</li>
<li>MongoDB (when I need a database)</li>
<li>Firebase (for quick projects)</li>
</ul>

Tools:

<ul>
<li>VS Code + extensions</li>
<li>Git + GitHub</li>
<li>Figma (for designs)</li>
<li>Postman (for API testing)</li>
</ul>

Deployment:

<ul>
<li>Vercel (frontend)</li>
<li>Railway / Render (backend)</li>
<li>Cloudinary (images)</li>
</ul>

<h2>
 
 
 What I Wish I Knew at 15
</h2>

You don't need a $3000 course

I learned everything from:

<ul>
<li>YouTube (Traversy Media, Web Dev Simplified, Fireship)</li>
<li>freeCodeCamp</li>
<li>docs</li>
<li>Building stuff and googling errors</li>
</ul>

You don't need a perfect portfolio

My first portfolio had 3 projects. That's enough. Just make sure they WORK and the code is on GitHub.

Age doesn't matter

I'm 18. Some people start at 30. Some at 50. Doesn't matter. Just start.

Consistency beats intensity

1 hour every day &gt; 10 hours once a week

I coded daily, even if just for 30 minutes. That adds up.

Your first projects will suck

That's normal. Make them anyway. You can't get better without making bad stuff first.

<h2>
 
 
 Where I Am Now
</h2>

<ul>
<li>12 projects in my portfolio</li>
<li>First freelance client ($200)</li>
<li>Starting to apply for junior dev positions</li>
<li>Still learning every day</li>
<li>Still googling "how to center a div" sometimes</li>
</ul>

I'm not a senior developer. I'm not a genius. I'm just someone who started at 15 and didn't quit.

<h2>
 
 
 If You're Just Starting
</h2>

Start today. Not tomorrow. Today.

Pick ONE thing to learn (I'd say HTML/CSS). Follow ONE tutorial. Build ONE thing.

Then build another. Then another.

In 3 years, you'll look back and be amazed at how far you've come.

You don't need:

<ul>
<li>Perfect conditions</li>
<li>Expensive courses</li>
<li>A CS degree</li>
<li>To be a "natural"</li>
</ul>

You just need:

<ul>
<li>A laptop</li>
<li>Internet</li>
<li>Consistency</li>
<li>Patience with yourself</li>
</ul>

<h2>
 
 
 What's Next For Me
</h2>

<ul>
<li>Keep building projects</li>
<li>Learn more about system design</li>
<li>Contribute to open source</li>
<li>Maybe start a YouTube channel?</li>
<li>Get my first dev job</li>
</ul>

I'm 18. I have so much to learn. But I'm not waiting until I "know enough."

I'm building, shipping, and learning as I go.

<h2>
 
 
 Your Turn
</h2>

If you're learning to code right now - what's holding you back from building something?

If you already learned - what do you wish you knew when you started?

Drop it in the comments. Let's help each other out.




My learning resources (all free):

<ul>
<li>freeCodeCamp</li>
<li>JavaScript.info</li>
<li>React docs</li>
<li>YouTube (Traversy, WebDevSimplified, Fireship)</li>
<li>MDN Web Docs</li>
<li>Stack Overflow (obviously)</li>
</ul>

Follow my journey: 
My PortFolio - <a href="https://sahil-sahu-portfolio.vercel.app/" rel="noopener noreferrer">click here</a> 
Follow me on X - <a href="https://x.com/SahilSahu731" rel="noopener noreferrer">click here</a>

If this helped you, hit that ❤️. If you're learning too, drop your GitHub - I'll check out your projects!

<h1>
 
 
 webdev #beginners #javascript #react #coding #100daysofcode
</h1>

في سن الثامنة عشرة فقط، تحول ساهيل ساهو من مبتدئ لم يكن يعرف ما هو HTML إلى مطور كامل المكدس مع محفظة تضم أكثر من خمسة مشاريع وعميله الحر الأول. تسلط هذه الرحلة الضوء على قوة العزيمة والتعلم الذاتي في صناعة التكنولوجيا، مما يلهم الآخرين لمتابعة شغفهم في البرمجة والتطوير.

A solo 18 años, Sahil Sahu ha pasado de ser un novato que no sabía qué era HTML a convertirse en un desarrollador full-stack con un portafolio de más de cinco proyectos y su primer cliente freelance. Este viaje destaca el poder de la determinación y el autoaprendizaje en la industria tecnológica, inspirando a otros a seguir sus pasiones en la programación y el desarrollo.

À seulement 18 ans, Sahil Sahu est passé d'un novice qui ne connaissait pas HTML à un développeur full-stack avec un portfolio de plus de cinq projets et son premier client freelance. Ce parcours met en lumière le pouvoir de la détermination et de l'auto-apprentissage dans l'industrie technologique, inspirant les autres à poursuivre leurs passions en codage et développement.

At just 18, Sahil Sahu has transformed from a novice who didn't know HTML to a full-stack developer with a portfolio of over five projects and his first freelance client. This journey highlights the power of determination and self-learning in the tech industry, inspiring others to pursue their passions in coding and development.

How I Went From "Hello World" at 15 to Full-Stack Developer at 18

arXiv:2510.10728v3 Announce Type: replace-cross 
Abstract: We tackle high-dimensional, path-dependent valuation and control and introduce a deep BSDE/2BSDE solver that couples truncated log-signatures with a neural rough differential equation (RDE) backbone. The architecture aligns stochastic analysis with sequence-to-path learning: a CVaR-tilted terminal objective targets left-tail risk, while an optional second-order (2BSDE) head supplies curvature estimates for risk-sensitive control. Under matched compute and parameter budgets, the method improves accuracy, tail fidelity, and training stability across Asian and barrier option pricing and portfolio control: at d=200 it achieves CVaR(0.99)=9.80% versus 12.00-13.10% for strong baselines, attains the lowest HJB residual (0.011), and yields the lowest RMSEs for Z and Gamma. Ablations over truncation depth, local windows, and tilt parameters confirm complementary gains from the sequence-to-path representation and the 2BSDE head. Taken together, the results highlight a bidirectional dialogue between stochastic analysis and modern deep learning: stochastic tools inform representations and objectives, while sequence-to-path models expand the class of solvable financial models at scale.

تقدم دراسة حديثة نهجًا مبتكرًا لتحسين المحافظ المالية باستخدام المعادلات التفاضلية الخشنة العصبية (RDE). تجمع هذه الطريقة بين التحليل العشوائي المتقدم وتقنيات التعلم العميق لتحسين إدارة المخاطر، مع التركيز بشكل خاص على مخاطر الذيل الأيسر. من خلال استخدام بنية فريدة تدمج التوقيعات اللوغاريتمية المقطوعة، تهدف الأبحاث إلى تقديم تقييمات وتحكمات أكثر دقة في السيناريوهات المالية عالية الأبعاد. هذه التطورات مهمة لأنها قد تؤدي إلى استراتيجيات استثمار أفضل وقرارات مالية محسنة.

Un estudio reciente presenta un enfoque innovador para la optimización de carteras utilizando ecuaciones diferenciales rugosas neuronales (RDE). Este método combina un análisis estocástico avanzado con técnicas de aprendizaje profundo para mejorar la gestión de riesgos, centrándose especialmente en los riesgos de cola izquierda. Al emplear una arquitectura única que integra log-firmas truncadas, la investigación busca proporcionar valoraciones y control más precisos en escenarios financieros de alta dimensión. Este avance es significativo, ya que podría conducir a mejores estrategias de inversión y a una toma de decisiones financieras más efectiva.

Une étude récente présente une approche innovante pour l'optimisation de portefeuille utilisant des équations différentielles rugueuses neuronales (RDE). Cette méthode combine une analyse stochastique avancée avec des techniques d'apprentissage profond pour améliorer la gestion des risques, en se concentrant particulièrement sur les risques de queue gauche. En utilisant une architecture unique qui intègre des log-signatures tronquées, la recherche vise à fournir des évaluations et un contrôle plus précis dans des scénarios financiers de haute dimension. Cette avancée est significative car elle pourrait conduire à de meilleures stratégies d'investissement et à une prise de décision financière améliorée.

A recent study introduces an innovative approach to portfolio optimization using neural rough differential equations (RDEs). This method combines advanced stochastic analysis with deep learning techniques to enhance risk management, particularly focusing on left-tail risks. By employing a unique architecture that integrates truncated log-signatures, the research aims to provide more accurate valuation and control in high-dimensional financial scenarios. This advancement is significant as it could lead to better investment strategies and improved financial decision-making.

Rough Path Signatures: Learning Neural RDEs for Portfolio Optimization

arXiv:2508.03159v2 Announce Type: replace 
Abstract: Drug toxicity remains a major challenge in pharmaceutical development. Recent machine learning models have improved in silico toxicity prediction, but their reliance on annotated data and lack of interpretability limit their applicability. This limits their ability to capture organ-specific toxicities driven by complex biological mechanisms. Large language models (LLMs) offer a promising alternative through step-by-step reasoning and integration of textual data, yet prior approaches lack biological context and transparent rationale. To address this issue, we propose CoTox, a novel framework that integrates LLM with chain-of-thought (CoT) reasoning for multi-toxicity prediction. CoTox combines chemical structure data, biological pathways, and gene ontology (GO) terms to generate interpretable toxicity predictions through step-by-step reasoning. Using GPT-4o, we show that CoTox outperforms both traditional machine learning and deep learning model. We further examine its performance across various LLMs to identify where CoTox is most effective. Additionally, we find that representing chemical structures with IUPAC names, which are easier for LLMs to understand than SMILES, enhances the model's reasoning ability and improves predictive performance. To demonstrate its practical utility in drug development, we simulate the treatment of relevant cell types with drug and incorporated the resulting biological context into the CoTox framework. This approach allow CoTox to generate toxicity predictions aligned with physiological responses, as shown in case study. This result highlights the potential of LLM-based frameworks to improve interpretability and support early-stage drug safety assessment. The code and prompt used in this work are available at https://github.com/dmis-lab/CoTox.

يعد التطور الأخير لنموذج CoTox، الذي يعتمد على سلسلة من الأفكار في تحليل وتوقع سمية الجزيئات، تقدمًا كبيرًا في مواجهة تحديات سمية الأدوية في تطوير الأدوية. على عكس نماذج التعلم الآلي التقليدية التي تعاني من صعوبة في التفسير والاعتماد على البيانات، يستخدم CoTox نماذج اللغة الكبيرة لتقديم تحليل خطوة بخطوة، مما يعزز قدرته على توقع السميات الخاصة بالأعضاء. هذه الابتكار مهم لأنه قد يؤدي إلى عمليات تطوير أدوية أكثر أمانًا ونتائج أفضل للمرضى.

El reciente desarrollo de CoTox, un modelo de razonamiento y predicción de toxicidad molecular basado en una cadena de pensamiento, marca un avance significativo en la resolución de los desafíos de toxicidad de medicamentos en el desarrollo farmacéutico. A diferencia de los modelos de aprendizaje automático tradicionales que luchan con la interpretabilidad y la dependencia de datos, CoTox aprovecha los grandes modelos de lenguaje para proporcionar un razonamiento paso a paso, mejorando su capacidad para predecir toxicidades específicas de órganos. Esta innovación es crucial, ya que podría conducir a procesos de desarrollo de medicamentos más seguros y mejores resultados para los pacientes.

Le développement récent de CoTox, un modèle de raisonnement et de prédiction de toxicité moléculaire basé sur une chaîne de pensée, marque une avancée significative dans la lutte contre les défis de la toxicité des médicaments dans le développement pharmaceutique. Contrairement aux modèles d'apprentissage automatique traditionnels qui ont du mal avec l'interprétabilité et la dépendance aux données, CoTox utilise de grands modèles de langage pour fournir un raisonnement étape par étape, améliorant ainsi sa capacité à prédire les toxicités spécifiques aux organes. Cette innovation est cruciale car elle pourrait conduire à des processus de développement de médicaments plus sûrs et à de meilleurs résultats pour les patients.

The recent development of CoTox, a chain-of-thought-based molecular toxicity reasoning and prediction model, marks a significant advancement in addressing drug toxicity challenges in pharmaceutical development. Unlike traditional machine learning models that struggle with interpretability and data reliance, CoTox leverages large language models to provide step-by-step reasoning, enhancing its ability to predict organ-specific toxicities. This innovation is crucial as it could lead to safer drug development processes and better patient outcomes.

CoTox: Chain-of-Thought-Based Molecular Toxicity Reasoning and Prediction

arXiv:2511.00064v2 Announce Type: replace 
Abstract: Clustering algorithms often rely on restrictive assumptions: K-Means and Gaussian Mixtures presuppose convex, Gaussian-like clusters, while DBSCAN and HDBSCAN capture non-convexity but can be highly sensitive. I introduce EVINGCA (Evolving Variance-Informed Nonparametric Graph Construction Algorithm), a density-variance based clustering algorithm that treats cluster formation as an adaptive, evolving process on a nearest-neighbor graph. EVINGCA expands rooted graphs via breadth-first search, guided by continuously updated local distance and shape statistics, replacing fixed density thresholds with local statistical feedback. With spatial indexing, EVINGCA features log-linear complexity in the average case and exhibits competitive performance against baselines across a variety of synthetic, real-world, low-d, and high-d datasets.

تقديم EVINGCA، وهو خوارزمية جديدة للتجميع، يمثل تقدمًا كبيرًا في تحليل البيانات. على عكس الطرق التقليدية التي تعتمد على افتراضات صارمة حول توزيع البيانات، يتكيف EVINGCA مع الطبيعة المتطورة للبيانات، مما يجعله أكثر فعالية لمجموعات البيانات المعقدة. هذه الابتكار مهم لأنه يعزز القدرة على تحديد الأنماط في مجالات متنوعة، من التعلم الآلي إلى العلوم الاجتماعية، مما يؤدي في النهاية إلى رؤى وقرارات أكثر دقة.

La introducción de EVINGCA, un nuevo algoritmo de agrupamiento, marca un avance significativo en el análisis de datos. A diferencia de los métodos tradicionales que dependen de suposiciones rígidas sobre la distribución de datos, EVINGCA se adapta a la naturaleza evolutiva de los datos, lo que lo hace más efectivo para conjuntos de datos complejos. Esta innovación es crucial, ya que mejora la capacidad de identificar patrones en diversos campos, desde el aprendizaje automático hasta las ciencias sociales, lo que lleva a percepciones y decisiones más precisas.

L'introduction d'EVINGCA, un nouvel algorithme de clustering, représente une avancée significative dans l'analyse des données. Contrairement aux méthodes traditionnelles qui reposent sur des hypothèses rigides concernant la distribution des données, EVINGCA s'adapte à la nature évolutive des données, ce qui le rend plus efficace pour des ensembles de données complexes. Cette innovation est cruciale car elle améliore la capacité à identifier des motifs dans divers domaines, de l'apprentissage automatique aux sciences sociales, conduisant finalement à des insights et des décisions plus précis.

The introduction of EVINGCA, a new clustering algorithm, marks a significant advancement in data analysis. Unlike traditional methods that rely on rigid assumptions about data distribution, EVINGCA adapts to the evolving nature of data, making it more effective for complex datasets. This innovation is crucial as it enhances the ability to identify patterns in diverse fields, from machine learning to social sciences, ultimately leading to more accurate insights and decisions.

EVINGCA: Adaptive Graph Clustering with Evolving Neighborhood Statistics

arXiv:2511.03620v1 Announce Type: cross 
Abstract: CLAX is a JAX-based library that implements classic click models using modern gradient-based optimization. While neural click models have emerged over the past decade, complex click models based on probabilistic graphical models (PGMs) have not systematically adopted gradient-based optimization, preventing practitioners from leveraging modern deep learning frameworks while preserving the interpretability of classic models. CLAX addresses this gap by replacing EM-based optimization with direct gradient-based optimization in a numerically stable manner. The framework's modular design enables the integration of any component, from embeddings and deep networks to custom modules, into classic click models for end-to-end optimization. We demonstrate CLAX's efficiency by running experiments on the full Baidu-ULTR dataset comprising over a billion user sessions in $\approx$ 2 hours on a single GPU, orders of magnitude faster than traditional EM approaches. CLAX implements ten classic click models, serving both industry practitioners seeking to understand user behavior and improve ranking performance at scale and researchers developing new click models. CLAX is available at: https://github.com/philipphager/clax

CLAX هي مكتبة مبتكرة تجمع بين نماذج النقر الكلاسيكية والتحسين الحديث القائم على التدرج في JAX. هذا التطور مهم لأنه يسمح للممارسين باستخدام تقنيات التعلم العميق المتقدمة مع الحفاظ على قابلية تفسير النماذج التقليدية. من خلال سد هذه الفجوة، تفتح CLAX آفاقًا جديدة للباحثين والمطورين في هذا المجال، مما يسهل تنفيذ نماذج النقر المعقدة بشكل فعال.

CLAX es una biblioteca innovadora que une modelos de clic clásicos y optimización moderna basada en gradientes en JAX. Este desarrollo es significativo porque permite a los profesionales utilizar técnicas avanzadas de aprendizaje profundo mientras mantienen la interpretabilidad de los modelos tradicionales. Al cerrar esta brecha, CLAX abre nuevas posibilidades para investigadores y desarrolladores en el campo, facilitando la implementación efectiva de modelos de clic complejos.

CLAX est une bibliothèque innovante qui combine des modèles de clic classiques et une optimisation moderne basée sur le gradient dans JAX. Ce développement est important car il permet aux praticiens d'utiliser des techniques avancées d'apprentissage profond tout en préservant l'interprétabilité des modèles traditionnels. En comblant cette lacune, CLAX ouvre de nouvelles possibilités pour les chercheurs et les développeurs dans le domaine, facilitant ainsi la mise en œuvre efficace de modèles de clic complexes.

CLAX is an innovative library that brings together classic click models and modern gradient-based optimization in JAX. This development is significant because it allows practitioners to utilize advanced deep learning techniques while maintaining the interpretability of traditional models. By bridging this gap, CLAX opens up new possibilities for researchers and developers in the field, making it easier to implement complex click models effectively.

Multi-Agent Regime-Conditioned Diffusion (MARCD) for CVaR-Constrained Portfolio Decisions

Multi-Agent Regime-Conditioned Diffusion (MARCD) for CVaR-Constrained Portfolio Decisions

Was this article worth reading? Share it