WhatTheStack 2026
who it's for
software engineers, AI and machine learning engineers, frontend developers, platform engineers, observability engineers, security engineers, engineering managers
topics
local model inference without NVIDIA; LLM internals and context windows; agent memory and caching; model quantization and GGUF; AI-assisted code generation agents; observability as a product with SLOs; Angular integrations for AI and WebMCP; TypeScript operator overloading layer; portable GPU kernel compilation (tinygrad to Metal); AI deployment in hospitals; graph algorithms for voting analysis; performant TypeScript shaders
| title | author | length |
|---|---|---|
| Legacy in IT: Breaking Stereotypes, Understanding Reality, and Embracing the Uncool | Goran David | 30m |
Argues that “legacy” is often misused and gives concrete evidence: COBOL handles ~80–95% of US ATM/card transactions, governments and transit systems still… | ||
| How to not use AI. Or staying sane in a frenzied world | Bogdan Ilie | 28m |
Argues that low-cost experimentation from transformer models reduces the cost of being wrong but increases decision fatigue and burnout, citing a… | ||
| My 30 Years of JavaScript: How I Learned to Stop Worrying and Love JavaScript | Wekoslav Stefanovski | 32m |
A 27–28 year personal history traces JavaScript from a 10-day experiment to the dominant runtime, with concrete episodes: a subtle bug that costs two days… | ||
| Building for the Unknown: Scaling Observability as a Product | Riste Oreshkovski | 19m |
Claims that treating observability as a product and applying platform engineering turns telemetry into consistent, self‑service insight for a portfolio of 85… | ||
| Artful Defense: Let's Sketch Web Security | Ramona Schwering | 30m |
Live sketch notes map front-end risks to concrete fixes: Vue auto-escapes templates but CVEs have hit i18n prototype pollution so frameworks are not a full… | ||
| Building Angular Apps With AI | Sam Vloeberghs | 34m |
Angular ships concrete tools: repository-level llm.ext and llm.full.ext files that fix LLM knowledge, plus official system-prompt files and an Angular CLI MCP… | ||
| From Portable AI Lab to Frontier Research | Nino Risteski | 34m |
Builds a complete, inspectable research workflow on a MacBook + tinygrad and follows a tensor op from Python through compiler optimization to a generated Metal… | ||
| What Do You Do When NVIDIA GPUs Aren't Available for Inference? | Bojan Jakimovski | 32m |
Claim: far more inference runs without NVIDIA than expected — engineered a small Mixture‑of‑Experts “Trinity mini” to avoid huge clusters while training still… | ||
| Inside GPTs: Understanding Large Language Models | Igor Trajkovski | 36m |
Full pipeline from raw web to deployed assistant with concrete numbers: raw filtered Common Crawl becomes 44 terabytes of training text, tokenized into ~50k… | ||
| Context Is Their Reality: How AI Agents Perceive the World | Elena Kolevska | 33m |
Claim: context that an agent receives is its only reality and managing it is the builder's responsibility. Tooling demonstrated: agent memory server backed by… | ||
| I’m a Software Engineer at Microsoft. I Don’t Write Code. | Gerald Versluis | 33m |
Claim: no meaningful code is written since November 2025 because work moves through GitHub Copilot and Copilot CLI and autonomous agents. The .NET MAUI team… | ||
| JavaScript Doesn't Support Operator Overloading... Right? | Iwo Plaza | 35m |
Claims a practical operator-overload layer for TypeScript that keeps full type inference and normal JS ergonomics by applying surgical, procedural patches to… | ||
| Harnessing the Model: What Coding Agents Are Really Made Of | Dushko Klincharov | 38m |
Walks through the Open Code codebase and proves a coding agent is mostly harness not magic: TypeScript async/generator loops glue an LLM to tools, streaming… | ||
| What makes AI work in a hospital, and what makes it harder | Dragan Shahpaski | 20m |
Real hospital AI needs deep integration: radiology voice recognition runs as two-stream real-time transcription tied to PACS and EHR so models can be… | ||
| One Project, Three Ways I Changed How I Write Kotlin | Andjelina Maksimovic | 14m |
Built a turn‑based card game exposes three concrete lessons: a single flexible Hero class with nullable effect fields produces invalid states, replaced by a… | ||
| Friendly Borders: Graph algorithms reveal Eurovision voting patterns | Domagoj Marić | 35m |
Claims neighborly favoritism is real and quantifies it using 2008–2024 Eurovision voting data, ingesting a community-maintained JSON dataset, building graphs… | ||
| I Stopped Throwing Errors and My Code Finally Works | Andrej Acevski | 23m |
Effect is a TypeScript library that makes errors part of the type so functions declare exact failure types (e.g., BoardNotFound | DBError | NotifyFailed), uses… | ||
| A Bouncer Gang in the (Vibe) Coding Party | Bozidar Spirovski | 32m |
Claims a practical pipeline that pairs AI with deterministic scanners to speed security reviews and surface real faults. It builds three open tools: PR Bouncer… | ||
| (When) Do We Finally Have a Unified Toolchain? | Alexander Lichter | 24m |
Claims a single CLI called V+ (vp) replaces stitching: VP migrate auto-converts ESLint/Prettier/tsup setups, runs installs and scaffolds projects. It bundles… | ||
| Thinking in qubits, using code | Serena Sensini | 32m |
Hands‑on Python demos use Qiskit emulator and state‑vector inspection to show concrete claims: a single Hadamard produces near 50/50 collapse (Hadamard ->… | ||
| A Brief History of Code Review (And What's Next) | Santosh Yadav | 32m |
History traces manual inspections to email patches to GitHub pull requests and now AI agents claim to remove the human bottleneck by auto-generating PRs and… | ||
| Performant TypeScript shaders: breaking down the Jelly Slider | Konrad Reczko | 32m |
Shows TypeScript compiled to GPU code using TypeGPU compiles TypeScript to GPU shaders, not by running JS on the GPU, and uses that to build a jelly slider… | ||
| Building Cleaner Angular Applications with Signal Forms | Kiril Zafirov | 23m |
A working booking form with organizer fields, a dynamic attendee list and nested company billing is built with Angular Signal Forms and a JSON schema; the demo… | ||
| How TanStack AI made the hidden complexity of AI simple for you | Alem Tuzlak | 31m |
TanStack AI claims to replace dozens of pages of hand‑rolled SSE and provider‑specific parsing with a few lines: a client useChat hook, a server chat method… | ||
| Reverse-Engineering Seniority, From Bootcamp Grad to Staff Engineer in Four Years | Faris Aziz | 35m |
Claims a four‑year jump from bootcamp grad to staff engineer and a promotion to engineering manager by age 22, backed by specific actions not platitudes: a… | ||
| Stop Engineering Features. Start Learning Graphs. | Stefan Andonov | 32m |
Relational tables become a heterogeneous graph using foreign keys as edges and many‑to‑many tables as weighted edges, then a GNN learns order embeddings to… | ||
| nothing matches | ||