// stack / tools / rankings

Tech Stack & State of AI

Disclaimer. Rankings here are my personal opinions — not universal truth or a live snapshot of the current state of the field. I still review anything that touches production; models speed exploration but do not replace ownership.

Models

01
Claude Opus 5

Claude Opus 5

frontendbackenddistributed systemsagentic coding

My default when the work is hard. I use Opus 5 for frontend product work alongside Grok, and for complex backend and distributed-systems problems when I need real judgment — architecture, tradeoffs, and follow-through across a messy codebase.

+ Pros

  • Trusted for hard frontend and backend work, not just one side of the stack
  • Best partner when the problem is ambiguous and I need strong reasoning
  • Pairs well with Grok: Opus for judgment, Grok for fast iteration

− Cons

  • Slower and more expensive than I want for every small edit
  • I switch to Grok 4.5 when I just need to move quickly in the editor
02
GPT 5.6 Sol

GPT 5.6 Sol

coding agentstool callinglong-horizonmodern software

I reach for Sol when I want a strong coding pass outside my Opus default — shipping a feature, driving tools hard, or working against newer APIs and libraries. It often one-shots work when I give it a clear prompt.

+ Pros

  • Great when I need reliable tool use and a long coding loop
  • Often one-shots a feature or feedback pass with a strong prompt
  • Feels current on modern tooling and patterns

− Cons

  • Opus still wins for my hardest architecture and systems reasoning
  • Grok is what I use for everyday in-editor iteration
03
Grok 4.5

Grok 4.5

cursorgeneral purposefast iterationvalue

My general-purpose developer agent in Cursor for fast work. I use Grok 4.5 for day-to-day implementation — layout, components, cleanup, and keeping momentum — and pull in Opus when the problem needs deeper judgment.

+ Pros

  • Fast enough that I stay in flow for day-to-day coding
  • Strong general-purpose default when I don’t need Opus-level reasoning
  • Feels native to the editor loop I already live in

− Cons

  • I still escalate to Opus or Sol when the problem gets hard
  • Not my pick for deep architecture or systems-level work
04
Gemini 3.6 Flash

Gemini 3.6 Flash

general assistantmultimodalfastimage gen

I use Gemini 3.6 Flash in the Gemini app for everyday, non-coding help — travel, drafting, general chat, and image generation. It stays out of my coding harnesses on purpose.

+ Pros

  • Quick for everyday questions and drafts
  • My go-to in-app for image generation and other multimodal one-offs

− Cons

  • Not where I go for coding agents
  • Less integrated into my editor workflow than Cursor / Codex

Honorable Mentions

  • Kimi K3

    frontendon-device

    Frontend and on-device work when I want a capable model outside my usual Cursor / Anthropic / OpenAI loop.

Harnesses

01
Claude Code

Claude Code

codingcliagentic

My top harness. Claude Code is where I reach for deep, agentic work — real ownership of a task across a messy repo, not just autocomplete. It pairs naturally with Opus and holds up well over long, multi-step sessions.

+ Pros

  • Best agentic follow-through of anything I use, especially on long-horizon tasks
  • Terminal-native, so it fits directly into my existing workflow and tooling
  • Strong default pairing with Opus for hard problems

− Cons

  • Less of an all-in-one editor experience than Cursor for fast, visual UI iteration
02
Cursor

Cursor

codingeditorcloud agents

My main editor harness. Cursor wins for me because I can use all the models I care about, Grok usage is generous, cloud agents are a huge plus for real, ad-hoc development work, and the overall UX, speed, and reliability are strong enough that it just works and just makes sense.

+ Pros

  • Cloud agents are a major strength
  • Best overall mix of model access, UX, speed, and reliability
  • Multi-surface clients (agents UI, editor, CLI) that all consume the same models and agents on my subscription

− Cons

  • Platform churn still happens from time to time, especially for research tasks
03
Codex

Codex

codingagentsopenai

My number three harness. I reach for Codex often because it lives in ChatGPT — I can do knowledge work and coding in one interface, then kick off agent loops when I want to implement something without switching tools.

+ Pros

  • ChatGPT + coding in one place, so knowledge work and implementation stay together
  • Strong when I want an agent to take a clear task and run with it
  • Nice complement to Claude Code and Cursor when I’m already in the OpenAI surface

− Cons

  • Claude Code and Cursor are still my default for editor-centric and deep agentic work
  • I miss Cursor’s full IDE loop when I’m deep in a repo

Honorable Mentions

  • Pi

    codingopen

    A minimal, MIT-licensed terminal coding harness from Mario Zechner — a lean open agent harness I keep an eye on when I want something outside Claude Code, Cursor, and Codex.

Tech stack

Languages

What I write most often — not the same thing as runtimes or frameworks below.

  • TypeScript JavaScript Go Python

    TypeScript · JavaScript · Go · Python

    TypeScript and JavaScript for application code; Go for microservices and control planes; Python for tooling, notebooks, and ML-adjacent glue.

Runtimes

Where JS/TS executes — distinct from Hono or any web framework.

  • Bun Node.js

    Bun · Node.js

    Bun is my usual Node-compatible runtime for new JS services and scripts. I still reach for Node.js when tooling, libraries, or deployment targets expect it.

Web UI & client stack

Product-facing work: frameworks, styling, and build tooling.

  • React TanStack Svelte Tailwind CSS Vite Astro

    React · TanStack Suite · Svelte / SvelteKit · Tailwind CSS · Vite · Astro

    React with the TanStack Suite (Query, Router, and whatever else fits the feature); Svelte and SvelteKit when I want lean reactivity. Tailwind everywhere; Vite for builds; Astro for content sites like this one.

Backend & APIs

Services and HTTP boundaries — framework choice is separate from which runtime hosts it.

  • Go Hono GraphQL Socket.io

    Go (Gin, Chi) · Hono · GraphQL federation · REST · WebSockets

    Go for most service logic. Hono is my go-to small web framework for TypeScript HTTP APIs; I usually run it on Bun, but that is a hosting choice — not the same dimension as “what language” or “what framework.” GraphQL (including federation), REST, and WebSockets for APIs and realtime surfaces (dashboards, control planes, live updates).

Data, infrastructure, observability & ML

The resume-shaped layer: storage, delivery, signals, tests, and production ML patterns.

  • PostgreSQL Redis MongoDB

    PostgreSQL · Redis · MongoDB

  • Kubernetes Docker Terraform Helm

    Kubernetes · Docker · Terraform · Helm

    Multi-cluster and resilient deployments; GitOps and infra-as-code.

  • GitHub Actions

    GitHub Actions

    CI/CD pipelines and release automation.

  • Datadog Prometheus Grafana OpenTelemetry

    Datadog · Prometheus · Grafana · OpenTelemetry

    Metrics, traces, and dashboards aligned with on-call reality.

  • Vitest Jest

    Playwright · Vitest · Jest · React Testing Library

  • LangChain

    LangChain · prompt engineering · model APIs

    Plus evaluation and guardrail patterns for control planes in front of real models and traffic.