Posts

Build in public has become one of those phrases that means everything and nothing. Some people use it to mean live-tweeting every todo list change. Others use it to mean polished case studies after a project is finished. Neither is quite right. The version that actually builds an audience and helps your career sits in between. What “Build in Public” Is Actually For Before you do it, know why you are doing it. The reasons people build in public and whether they are good ones: ...

FD interest is taxed at your income slab rate and TDS is deducted even before you receive it. Here is the full picture of how FD taxation works - and legal ways to reduce it.

If your FDs are sitting in SBI, HDFC, or Kotak, you are almost certainly leaving 1-2% per year on the table. For a Rs. 10 lakh FD, that is Rs. 10,000-20,000 in foregone income annually. Over 3-5 years, compounded, the gap is significant. Here is why large banks pay less - and where you can safely earn more. Why Large Banks Pay Lower FD Rates Large PSU and private banks have massive, low-cost deposit bases. SBI has over 50 crore savings account holders who are happy earning 2.7-3% on their balances. This cheap deposit base means the bank does not need to compete aggressively on FD rates to fund its loans. ...

Most employees never properly read their salary slip. Here is every line item decoded and the specific deductions you might be able to legally reduce.

Terraform changed its license in 2023, OpenTofu forked it, and Pulumi gained ground. Two years later, the landscape has settled and there is a clear choice for most teams.

Stopping in the middle of a flow state every 25 minutes to take a break is bad advice for engineering work. Here is what the research and experience actually support.

Salary negotiation in India has its own unspoken rules. Here is a word-for-word script that respects those rules while actually getting you more money.

The productivity industry loves selling you the idea that the right app will change your life. If you just found the perfect note-taking system, the ideal task manager, the optimal calendar setup - then you would be productive. It is a trap. The engineers who are most productive have boring systems. Simple ones. Systems that would not make a good YouTube video because there is nothing impressive to show. What a Second Brain Actually Is The concept, popularized by Tiago Forte, is simple: your brain is for having ideas, not for storing them. An external system that captures, organizes, and surfaces information when you need it extends your cognitive capacity. ...

Every time someone adds search to their app, the default advice is “spin up Elasticsearch.” That advice is expensive, operationally heavy, and usually wrong. PostgreSQL has had solid full-text search since version 8. For 90% of apps, it is genuinely good enough. How PostgreSQL Full-Text Search Works Postgres converts your text into a tsvector - a sorted list of normalized lexemes (words stripped of suffixes). Queries use tsquery to match against those vectors. The core functions are to_tsvector and to_tsquery. ...

The old vs new tax regime debate has a clear answer for most income levels in 2026. Here is the exact breakeven calculation and who should be on which regime.

A personal loan at 18% sounds expensive. Sometimes it is the smartest financial move. Sometimes it is a catastrophe. The difference depends entirely on what you are using it for.

When RBI cuts or hikes the repo rate, your home loan EMI should change. But it often does not move as expected. Here is the complete picture of how repo rate transmission actually works.

Eighteen months ago our AI tooling bill crossed $50,000 per year and kept growing. Commercial LLM APIs, a vector database with per-query pricing, an embedding service, an AI observability platform, and a prompt management tool. Each individually defensible. Together, unsustainable. We spent 90 days migrating to open source alternatives. The result: $11,000 per year (mostly infrastructure) and tooling we can extend without hitting API limits. The Before Stack Tool Purpose Annual cost OpenAI GPT-4 API Primary LLM inference $28,000 Pinecone Vector database $8,400 Cohere Embeddings $4,800 Langsmith LLM observability $3,600 PromptLayer Prompt versioning $2,400 Helicone API proxy + analytics $1,800 Total $49,000 Usage was growing 15% per quarter. Projecting forward: $80K+ by end of year. ...

Health insurance claim rejections often happen due to avoidable reasons. Here are the most common causes and the exact steps to prevent them before you ever need to file.

Most LinkedIn advice will turn you into someone you would unfollow. Here is how to build a professional presence that actually works without performing for the algorithm.

GraphQL’s promise is efficient data fetching - clients request exactly what they need. The reality is that flexible queries create performance problems that REST APIs rarely encounter. These are the specific issues that appear when traffic scales, and the fixes that work. The N+1 Problem Is Worse in GraphQL In REST, a poor implementation causes N+1 queries. In GraphQL, N+1 is the default behavior without explicit mitigation. query { posts { # 1 query: SELECT * FROM posts LIMIT 20 title author { # 20 queries: SELECT * FROM users WHERE id = ? name } } } The resolver for author fires once per post. The schema makes this natural to write. The database sees 21 queries. ...

The 40% jump is achievable and common. Here is the mechanics of how it works, why companies offer it, and how to position yourself to get it.

There is a specific kind of Obsidian vault that looks incredible in YouTube videos - perfectly organized folders, hundreds of linked notes, beautiful dashboards - and gets abandoned after two weeks because maintaining it is a full-time job. Many developers go through multiple versions of this. What works best is simpler and actually sticks. The Core Principle: Capture Fast, Organize Later (but Not Too Much Later) The enemy of a useful knowledge system is friction at capture time. If adding a note requires navigating to the right folder, applying five tags, linking to three existing notes, and writing a proper summary, you will not do it when you need to. ...

If you bought your term insurance through an agent or walked into an LIC branch, there is a reasonable chance you are paying 15-30% more in annual premium than someone who bought the exact same policy online. Over a 30-year term, that is Rs. 1-3 lakh in excess premium paid. Here is why the gap exists and how to buy online without sacrificing service or claim security. The Economics of Agent-Sold Insurance Insurance agents in India earn a commission on the first-year premium (typically 15-40% for term plans) and a smaller renewal commission in subsequent years. This commission is not shown to you as a line item - it is embedded in the premium you pay. ...

Most Python code is single-threaded and runs on one machine. ML workloads outgrew single machines years ago. Ray is the framework that makes distributed Python feel as natural as single-machine Python - and it has become the default infrastructure layer for serious ML teams. What Ray Is Ray is an open-source distributed computing framework from UC Berkeley’s RISELab (now maintained by Anyscale). The core idea: make a Python function run on any machine in a cluster as easily as it runs locally. ...