Piyush Mehta - Blog

Sonnet 5.5 beats Opus 5.5 by 4.2 points on Terminal-Bench at half the token price.

Context window entropy, state drift, and unconstrained decision loops cause 84% of multi-step agent tasks to fail.

How I built a browser extension that turns YouTube into a strict learning feed using Jev, TypeSafe Noul probabilities, a Cloudflare Worker, and local decision caching.

Twelve reproducible fixtures for an AI agent action gate: missing approvals, cross-project access, changed arguments, duplicate retries, and conflicting request keys.

A practical guide to choosing between retrieval-augmented generation and long-context prompting, with tradeoffs, a decision framework, and patterns for using both.

A first-person account of migrating this site from Vercel to Cloudflare Workers: the adapter swap, the Upstash Redis client that had to change, the native binary that could not run in a Worker, and the blank social-card bug it caused.

Herdr is the runtime your coding agents live on.

A Seagate IronWolf developed 52 bad sectors.

Meet OmaSwitch: a Windows-style Alt+Tab overlay for Omarchy with MRU ordering, type-to-search, live previews, and no daemon or extra packages.

Why time-based scheduling quietly produces a class of bugs that intent-based scheduling eliminates, and a 6-month migration playbook for moving from cron to an event bus without breaking production.

Coding agents now read your repository's instructions before they touch your code.

What one million tokens of context, DeepSeek's prefix caching, and a sub-cent per run price actually mean when you build an AI coding agent.

Moonshot AI released a 2.8-trillion-parameter model with a 1.56 TB checkpoint.

Entry-level engineering postings are down 20-35% globally while AI/ML roles are up 163%.

Why 75% of resumes get auto-rejected by bots - and the format, keywords, and bullet rewrites that get yours to a human at Google, Stripe, or the companies you want to work for

A guide to the questions that separate senior SWE candidates from mid-level, including a deep dive into HyperLogLog for counting unique users at scale, 15+ real interview questions, and the mistakes that silently disqualify you

The real architecture behind scaling an AI chat application — WebSockets, streaming, LLM cost control, caching strategies, and the three production incidents that taught me more than any system design interview ever did.

The 12 patterns interviewers actually ask about - load balancing, caching, message queues, sharding, consistent hashing, CQRS, event sourcing - with back-of-envelope math, concrete numbers, and the C4 diagram language

Why STAR+R (Situation, Task, Action, Result, Reflection) beats STAR every time — and the 5 key stories every senior engineer needs ready before their first loop

Why 30% of new hires fail in their first 90 days - and the week-by-week playbook that turned me from 'new joiner' to 'the person they can't live without' at three different companies

A production-grade walkthrough of the Model Context Protocol, the SDK almost nobody is using right, and why every full-stack dev will build at least one MCP server before 2027.

AI coding agents are moving from autocomplete into tool-calling workflows.

A detailed analysis of how mass tech layoffs ripple through housing markets, consumer spending, startup ecosystems, and the broader economy — and what it means for your career as a developer.

How I switched from macOS to Arch Linux using Omarchy and dramatically improved my development productivity.

A comprehensive case study of migrating a 50TB PostgreSQL database to support 10x user growth with zero downtime, improved performance, and 40% cost reduction.

How Bloom Filters power Google, Netflix, and Instagram with 99.9% accuracy while using almost zero memory - and the shocking reason they can lie to you

After 5 years as a developer, AI coding assistants did something unexpected - they made me realize I was actually good at my job.

Discover why I migrated from Nginx Proxy Manager to SWAG by LinuxServer.io, the challenges faced, and the transformative benefits that made this switch a game-changer for my self-hosted infrastructure.

Discover practical strategies for integrating AI coding assistants into your development process to boost productivity, improve code quality, and accelerate learning while avoiding common pitfalls.

Discover how Astro 5.9 introduces experimental Content Security Policy support, making it easier than ever to secure your websites against XSS attacks while maintaining framework flexibility.

A comprehensive deep-dive into migrating a legacy portfolio site to modern Astro 4.x, covering architectural decisions, performance optimizations, and lessons learned from a decade of web development.

Embark on a tech adventure like no other with "Building a Hare-raisingly Good Queue-based Messaging System"!

Creating a ChatGPT-like bot using Azure Bot SDK and Typescript can provide many benefits for your business

Ever wanted to make your own blog?

What exactly is hacktoberfest❓

JavaScript Event Loop is one of the most fundamental concepts that every developer must aspire to understand.

"Ever heard the word Micro-Frontend and wondered what the hell is that?

How to Monitor any website uptime using GitHub Actions for free

How is it that I stop feeling that I’m not good enough?

All about Kubernetes and Docker for a normal Man.

How we are maintaining ProtoBuf (Protocol Buffers) at Nuclei!

Why http(s) is not secure?