The stack wall.

Clojure

Functional language we use for large, long-lived web platforms

ClojureScript

Clojure for the browser: rich, data-driven single-page apps

JavaScript (Node.js)

The web's language, used on both server and browser

TypeScript

Typed JavaScript for safer front ends and services

Python

Scripting, data work, machine learning and automation

SQL

Querying and modelling relational data

Bash

Shell scripting for deploys, jobs and server chores

Swift

Native Apple code for small macOS and iOS utilities

HTML/CSS

Semantic, accessible markup and hand-tuned layouts

Reagent

Minimal React wrapper for ClojureScript interfaces

re-frame

Predictable state management for complex ClojureScript apps

React

Component library for interactive web interfaces

Next.js

React framework for server-rendered sites and apps

three.js

3D in the browser: configurators and product viewers

React Three Fiber + drei

Declarative 3D scenes built as React components

postprocessing

Post-processing effects (bloom, depth, tone) for WebGL scenes

Motion

Animation library for fluid UI transitions

TanStack Table

Headless tables: sorting, filtering and large data grids

MapTiler SDK + geocoding

Interactive maps and address search

Turf.js

Geospatial maths: distances, areas, regions

CesiumJS

3D globe and geospatial visualisation

Vanilla JS PWAs

Installable web apps that work offline on phones

Cloudflare Workers SSR (no bundler)

Server-rendered pages at the edge, with no build step

CodeMirror 6

Embeddable code and text editor for in-app editing

Paged.js

Paginates HTML into print-ready book and document layouts

Monaco Editor

The VS Code editor, embedded in web tools

Stripe Elements

Secure, embedded card payment forms

HEIC decoding in browser

Opening iPhone photos directly in the browser

i18n (EN/HU)

Bilingual English/Hungarian interfaces

http-kit

Fast, lightweight HTTP server for Clojure services

Ring + Reitit

Routing and middleware for Clojure web APIs

Malli

Data schemas and validation for Clojure systems

Integrant

Wiring and lifecycle for Clojure system components

core.async

Concurrent pipelines and background workers in Clojure

Hiccup

Generating HTML from plain Clojure data

Buddy (JWT/crypto)

Authentication tokens and cryptography for Clojure

Express

Minimal web server framework for Node.js

FastAPI + Uvicorn

Fast, typed Python APIs

Flask

Lightweight Python web apps and internal tools

bcrypt sessions / auth

Password hashing, sessions and login flows

PDF generation (pdf-lib, PDFKit, OpenHTMLtoPDF)

Generating invoices, labels, reports and print PDFs

PyMuPDF

Reading, splitting and rendering PDF documents

openpyxl (Excel)

Reading and writing Excel spreadsheets

Nodemailer + IMAP (imapflow, imap-simple, mailparser)

Sending, receiving and parsing email in automations

Cheerio

Parsing HTML pages for data extraction

sharp / Pillow image processing

Resizing, converting and optimising images

PostgreSQL

Our default relational database for serious workloads

SQLite

Embedded database for small services and local tools

Cloudflare D1

SQLite at the edge for serverless apps

Cloudflare R2 (S3-compatible)

Object storage for files, images and backups

Neon (serverless Postgres)

Serverless Postgres that scales to zero

MinIO

Self-hosted S3-compatible file storage

ClickHouse

Column database for fast analytics over large tables

Redis

In-memory cache, queues and rate limiting

MongoDB

Document database, used where a schema is still settling

HDBSCAN + scikit-learn

Clustering and classic machine learning on tabular data

Anthropic Claude API

Claude models for reasoning, writing and document work

Claude Code (headless agents)

AI coding agents running unattended background tasks

OpenAI GPT-4o / GPT-4o-mini

General-purpose LLMs for classification and extraction

OpenAI Realtime (voice agent)

Real-time speech-to-speech voice agents

Google Gemini 2.5 Flash

Fast, low-cost multimodal model for high-volume tasks

Ollama (local LLM serving)

Running open models on our own GPUs

Gemma 4 26B (local)

Open model we run locally for private workloads

Qwen (2.5 / 3.6, local)

Open models run locally for text and coding tasks

Qwen2.5-VL 7B (local VLM)

Local vision-language model for reading images

PyTorch (CUDA 12.8)

Training and running neural networks on GPU

Hugging Face Transformers

Pretrained models for text, vision and audio

DINOv2 embeddings

Image embeddings for similarity search and clustering

NVIDIA NeMo

NVIDIA toolkit for speech models

NVIDIA Canary STT

Speech-to-text transcription

faster-whisper

Fast local speech-to-text transcription

Orpheus TTS 3B / Magpie TTS / Higgs Audio

Text-to-speech voices generated on our own hardware

LLM lead assessment & classification

Scoring and sorting leads and businesses with LLMs

LLM message triage & reply drafting

Sorting incoming messages and drafting replies for review

Tesseract OCR

Reading text from scans and photos

Dataset tooling (icrawler, ddgs, imagehash)

Gathering, deduplicating and cleaning image training sets

3D product configurators (WebGL)

Interactive 3D product configurators in the browser

glTF pipeline (gltf-transform)

Optimising and converting 3D models for the web

Tripo3D (image-to-3D)

Generating 3D models from photos

Blender headless GPU rendering

Scripted, GPU-rendered product images and animations

BiRefNet background removal (fal.ai)

Cutting products out of photos automatically

SAM 3 segmentation (fal.ai)

Segmenting objects in images for masks and measurements

Google Lens via SerpApi

Visual search to identify products from a photo

OpenCV

Classic computer vision: detection, measurement, cleanup

GPT-4o Vision classification

Sorting and tagging photos with a vision model

Label and nameplate reading (vision models)

Reading part numbers and labels from product photos

Visual condition assessment

Assessing physical condition and damage from photos

QR codes

Generating and scanning QR codes to track items

Google Photorealistic 3D Tiles

Photoreal 3D city and terrain maps in the browser

eBay APIs (Sell, Trading, Browse)

Listings, orders, fulfilment and buyer messaging on eBay

FedEx REST (rate/ship/track/pickup/ETD)

Rates, labels, pickups and customs paperwork

Multi-carrier shipping aggregators (GLS, DPD, UPS)

Multi-carrier shipping labels and tracking in Hungary

Foxpost

Parcel locker shipping in Hungary

szamlazz.hu invoicing

Hungarian invoicing and receipts, issued automatically

NAV Online Számla v3

Real-time invoice reporting to the Hungarian tax office

EU VIES VAT validation

Checking EU VAT numbers automatically

Revolut Business API

Bank transactions and payments pulled into bookkeeping

Stripe

Online payments, subscriptions and checkout

Book marketplaces

Listing and order sync with a rare-book marketplace

Classifieds marketplaces

Listing sync with a Hungarian classifieds marketplace

Gmail (IMAP/SMTP)

Reading and sending mail from automations

Postmark

Reliable transactional email delivery

Instantly

Cold email sending and inbox rotation

MillionVerifier

Verifying email addresses before sending

Quo (OpenPhone) calls + texts

Business phone calls and texts logged into a CRM

Google Calendar API

Booking calls and syncing schedules

Outscraper

Business data from Google Maps at scale

gosom Google Maps scraper

Self-hosted Google Maps business scraping

Apify

Hosted web scrapers for data collection

Yelp Fusion

Business listings and reviews data

SerpApi / Serper

Search engine results for research and monitoring

MapTiler

Maps, geocoding and map tiles

Cloudflare API

Automating DNS, tunnels and edge configuration

Puppeteer / Playwright browser automation

Headless browsers for automation, testing and screenshots

Docker

Everything we run is containerised

Nix flakes + nix2container

Reproducible builds and minimal container images

Cloudflare Tunnel

Publishing services securely without opening ports

Cloudflare Workers / Pages

Serverless code and static sites at the edge

Caddy

Web server with automatic HTTPS

systemd services + timers

Long-running services and scheduled jobs on Linux

Tailscale

Private network between servers and devices

Coolify

Self-hosted platform for deploying apps

n8n

Visual workflow automation, self-hosted

Prometheus + node_exporter

Metrics collection for servers and services

Grafana

Dashboards and alerts for everything we run

Grafana Loki logging

Centralised, searchable logs

Own GPU servers + worker fleet

Our own GPU hardware for AI and rendering, plus worker nodes

Automated backups (rsync to second node)

Nightly backups copied to a second machine

sops + age secrets

Encrypted secrets kept safely alongside code

Google Analytics 4 (gtag)

Site analytics and conversion tracking

Meta Pixel + Conversions API

Ad tracking with server-side conversion events

Plausible (self-hosted CE)

Privacy-friendly analytics we host ourselves

Google Ads API

Managing and reporting on Google Ads campaigns

Meta Marketing API

Managing and reporting on Meta ad campaigns

DataForSEO (Ads Transparency / SERP)

Competitor ad research and search rankings data

SEO market map + on-page audits

Mapping local search competition and auditing pages

Git + GitHub

Version control, code review and CI

Bun

Fast JavaScript runtime and package manager

uv (Python)

Fast Python package and environment manager

shadow-cljs + nREPL + clj-reload

Clojure builds with a live REPL during development

esbuild / Vite / LightningCSS

Fast bundling and CSS processing

ESLint + Prettier

Linting and consistent code formatting

Renovate

Automated dependency updates

node:test / unittest suites

Automated test suites

Remotion (programmatic video)

Programmatic video: demos and explainers built in code

EPUBCheck + fontTools

Validating e-books and working with font files