June 2026

In How to Stretch Cursor Pro Further, I argued for treating Cursor as the execution layer and routing planning, research, and cheap thinking elsewhere. Ollama got a few paragraphs — enough to explain why local models belong in the pipeline, not enough to actually set them up. This post is the missing piece: install Ollama on Manjaro, pick models for your hardware, and connect local agents that can read your repo, edit files, and run commands without sending code to a cloud API. ...

I use Cursor every day across a lot of codebases — not just one repo. At work that is mostly a large Angular/Nx monorepo plus many smaller web projects. At home it is broader still: playground companion repos tied to omid.dev articles, browser demos on playground.omid.dev, Rust/WASM experiments, Linux tooling, and whatever the next post needs. Agent mode, multi-file refactors, and inline edits are genuinely faster than doing the same work by hand on any of them. ...

In The Zoom-Out, I wrote about the moment a senior developer realizes they don’t know what they don’t know — Corepack, Yarn, Orval, the whole peripheral toolchain that matured while we were busy shipping. That post was about naming the feeling and regaining perspective. Plenty of readers wrote back with the natural next question: “Okay, I get it — the map is huge and I can’t know everything. But how do I actually find the important stuff before I’ve been doing it wrong for five years?” ...

I’ve been using Linux for more than 15 years. I’ve administered servers, tuned kernels, experimented with filesystems, and read countless articles about swap, virtual memory, caching, and performance. I started with Ubuntu 4.x, spent years on Arch-based distributions, and today most of my work happens on Manjaro and Kubuntu. So I was surprised when I stumbled upon a feature that had apparently been helping my system for years without me realizing it: zswap. ...

Use this guide when you want to run a short session with Bio-Dynamics — a browser-only 3D lab linked from the probiotic articles on omid.dev. Each path below takes about 5 minutes; all three presets together fill a 15-minute workshop. Educational model — not medical advice. Population counts, strain effects, and biome shifts are illustrative. Do not use this lab for diagnosis or treatment decisions. Interactive lab Bio-Dynamics: Microbiome Sandbox Open the live lab before your session. No install required — works in a modern desktop or tablet browser. Open interactive lab Before you start Open the lab on a projector or share your screen. Leave the macro view (rotatable body) visible at first. Point out the dashboard on the right: integrity and inflammation meters, event log, stressor buttons, and catalog tabs (Strains, Prebiotics, and so on). Click a tissue hotspot on the body (or pick a region in the sidebar) to enter the micro view — zoomed tissue with moving microbes. Press Esc to return to the full body map. Tip: Hover catalog items before clicking — the action impact panel previews what will change. ...

Shipping an educational lab to a global audience means more than translation strings. Bio-Dynamics adds RTL layout for Persian, keyboard region shortcuts, ARIA live announcements, touch gesture hints, and URL-encoded lab checkpoints so teachers can share a mid-simulation state without a backend. Companion resource Bio-Dynamics: Microbiome Sandbox Try ?lang=fa or copy a lab link after running a scenario — source for i18n and labState.ts is on GitHub. github.com/omidfarhang/example-projects/labs/microbiome-sandbox Open live lab View source on GitHub Lightweight i18n without a framework Locales live in src/i18n/en.ts, de.ts, and fa.ts. A small t() helper resolves dot-path keys with parameter interpolation: ...

The Bio-Dynamics dashboard exposes a lot of buttons: regional stressors, inoculations, environment sliders, and four catalog tabs for strains, prebiotics, postbiotics, and products. The trick is not rendering HTML — it is keeping the catalog honest as content grows. This post explains the data-first layout and the action impact preview panel. Companion resource Bio-Dynamics: Microbiome Sandbox Hover strains and products in the live lab to see impact preview, or browse the catalog TypeScript files on GitHub. github.com/omidfarhang/example-projects/labs/microbiome-sandbox Open live lab View source on GitHub Data catalogs, not hard-coded panels Almost every control maps to a typed catalog under src/data/: ...

Most microbiome diagrams are flat. Bio-Dynamics tries the opposite: a single Three.js scene that starts as a rotatable body map, then animates into a tissue cross-section when you pick a region — with microbe meshes, SCFA particles, and fog density tied to live simulation state. This post covers the visualization layer. Companion source and the full architecture write-up are in the repository. Companion resource Bio-Dynamics: Microbiome Sandbox Open the live lab, zoom into gut or nasal tissue, then inspect the Three.js scene code on GitHub. github.com/omidfarhang/example-projects/labs/microbiome-sandbox Open live lab View source on GitHub One scene, two modes SceneManager owns a single WebGL scene with two interaction modes: ...

Bio-Dynamics is an educational 3D lab, not a clinical simulator. That constraint shaped every decision in the simulation layer: reproducible ticks, scalar tissue state, capped agent counts, and inflammation that emerges from pressure instead of jumping on every button press. This post walks through the engine design. Companion resource Bio-Dynamics: Microbiome Sandbox Run the live lab or inspect the simulation source — tick engine, golden tests, and full docs in the repository. github.com/omidfarhang/example-projects/labs/microbiome-sandbox Open live lab View source on GitHub Why deterministic? Interactive demos fail the educational test when the same button sequence produces different outcomes on refresh. Readers cannot build intuition from noise. ...

I write a lot about Angular platforms, monorepos, and production frontends. Bio-Dynamics is different: a browser-only educational lab where you rotate a 3D body map, zoom into tissue, and run deterministic probiotic scenarios tied to health articles on omid.dev. It started for three reasons — a human one, a developer-story one, and a career one. This post is the anchor for that project. Deeper technical posts follow in a short series linked at the end. ...

Most frontend teams do not have a testing problem because they lack tests. They have a testing problem because nobody can explain why a specific test exists. The result is familiar: hundreds of unit tests that prove implementation details; a few end-to-end tests that fail whenever timing changes; component tests that duplicate what unit tests already cover; slow CI pipelines that people stop trusting; high coverage numbers with very little confidence. This is especially common in large Angular codebases. Angular gives teams a serious testing toolbox: TestBed, standalone components, dependency injection, router testing, HTTP testing utilities, harnesses, and good compatibility with tools like Jest, Vitest, Cypress, and Playwright. The tooling is not the hard part. ...

Flaky tests are worse than failing tests. A failing test tells the team something broke. A flaky test teaches the team to negotiate with reality: “Run it again.” “CI is weird today.” “It passes locally.” “That test always fails on Mondays.” “Merge it, the failure is unrelated.” That is how a test suite loses authority. The first few flakes feel harmless. Then people stop reading failures carefully. Then a real regression hides inside the noise. ...

I wanted an excuse to build a small real Linux app. Not a shell script. Not a giant desktop application. Not a kernel module. Just a focused program that talks to Linux through the interfaces the system already exposes, gives that data a shape, and presents it as something a normal desktop user can run. Desktop stutter turned out to be a good excuse. My own machine is not slow: modern CPU, fast NVMe storage, plenty of RAM, KDE Plasma on Wayland, and a current kernel. Most of the time it feels excellent. Then, once in a while, the pointer hesitates, a window animation misses a beat, audio gets a tiny crackle, or the browser pauses while a package update or build is running. ...

I have not switched back to Kubuntu yet. I am still daily-driving Manjaro KDE on my ASUS Vivobook Pro 15, with Plasma 6.6.5, Wayland, btrfs, and a hybrid Intel Arc + NVIDIA RTX 3050 setup. But I have been running Kubuntu 26.04 LTS in VirtualBox, comparing the two side by side, and asking a question I did not expect to ask again after several happy years on Arch-based rolling release: ...

Open almost any mature Angular screen and you will find the same shape: 1 2 3 4 5 loading = false; error: string | null = null; data: Account[] | null = null; retrying = false; submitted = false; The template then becomes a negotiation: 1 2 3 4 5 6 7 8 9 @if (loading) { <app-spinner /> } @else if (error) { <app-error [message]="error" /> } @else if (!data?.length) { <app-empty-state /> } @else { <account-table [rows]="data!" /> } This looks fine. It ships. It passes review. And then production teaches you that the screen was never modeled as one thing. It was modeled as five independent switches that sometimes agree and sometimes do not. ...