Ads that pull,
shipped in 2 to 3 days.
UGC testimonials, product films, cinematic brand spots, and the static image ads that run alongside them. Scripted, voiced and cut in-house, which is why testing five angles costs less than filming one.
Watch one, then decide.
UGC testimonials, product films and cinematic spots. Script to finished vertical ad in 2 to 3 days, made in-house.
Not everything has to look real.
The format decides what the ad does. The look is a separate choice, and it runs on the same pipeline with the same turnaround as everything else.
See the styles →Numbers that stop the scroll.
Image creative for Meta: typography baked into the artwork rather than overlaid, and every line of copy checked against ad policy before it renders.
$9.83 to $16.71
CPL our leading concept held in live testing, against roughly $30 to $36 in the same niches.
The systems behind the ads.
Beyond the creative, we build the pipelines that run the operation behind it: data that stays accurate, competitors tracked daily, content at volume, and code audited on every deploy.
Data & Compliance
ksa-data · CasinoGuard
Your regulatory data is scattered across spreadsheets and CMS fields that nobody trusts. When requirements change, you hear it from an audit rather than from your own stack.
One source of truth your systems can trust. Every record verified against official sources and live operator sites, published content checked against it automatically, and an alert the moment something shifts upstream.
Competitive Intelligence
MetaMiner · Vizier
You find out about competitor moves from your boss, not your tools.
You learn about a competitor's move the morning it ships, not the week after. Gaps, velocity, and promo changes tracked daily.
See it liveContent Automation
ContentSmid · WoordSmid
Your content team can't keep up with the volume, and freelancers don't know your voice, your compliance rules, or your data.
Publication-ready content at volume, in your voice, across English, Dutch and German. Built from verified data instead of guesswork, then run through a layered review that strips the machine-written tells before anything reaches your CMS.
Site Quality & Code Audits
KeurSmid · Crucible
Broken pages and SEO regressions go unnoticed until traffic drops. Worse: silent code-level bugs in data, payment, or compliance flows can sit for months before becoming a regulator's problem.
Broken pages, regressions, and silent code bugs caught before users or regulators do. Every deploy, automatically.
About
Built in a regulated market.
Engineered for scale.
Maikel Slomp spent years doing content and compliance work by hand in a regulated market. Then he automated all of it.
MAD IT is the result: autonomous systems for data, intelligence, content, and monitoring. Every system we sell to clients, we run ourselves first.
How we work
Open source
What we publish openly
Beyond client work, we publish the pipeline itself. The ComfyUI workflows our creative runs on are public, and so is the agent tooling we build with.
Creative pipeline
MacMax Turbo
MiniMax H3 video with native stereo audio, running on Apple Silicon. Two ComfyUI workflows, measured cost tables, and the dead ends written down so nobody repeats them.
200+ downloads on Civitai
h3edit
The same video model used as an image editor: a one-frame reference-to-video pass. MPS port, compatibility node and a CLI.
150+ downloads on Civitai
ReDetail
Generative video re-detailer. Runs a clip back through LTX-2.5's pixel spatial upscaler, handling sizing, cut-based chunking and audio. It synthesizes detail rather than restoring it.
900+ downloads on Civitai
ltx-faceid-lipsync
Identity-locked talking video from one photo and one voice. The workflow was featured by the CEO of ComfyUI.
Featured by ComfyUI
krea2-turbo-mac
Krea 2 Turbo fully local on Apple Silicon, with in-graph model downloads and the locked 16-step recipe.
ideogram4-mac
Ideogram 4 locally on Apple Silicon: two core-node workflows on the official recipe, plus the caption grammar that keeps text legible.
Agent tooling
Crucible
Codebase-level adversarial review by a panel of frontier models, with findings verified against the source.
Rival
Adversarial multi-model code review via OpenRouter, as a Claude Code plugin.
Chippy
Crafts high-quality task prompts before spawning a sub-agent.
Sextant
Plan-and-execute multi-agent orchestrator. ~820 LoC, bounded replanning, parallel DAG execution. No framework lock-in.
Lede
Open-source editorial AI imagery: synthesizes an art-director's brief from an article body and bakes it across three frontier image models in parallel. The brief layer is the product, not the model.
ScrapeProbe
Open-source recon CLI for scraping work. Profiles a target URL's anti-bot stack, tech stack, bulk-data endpoints, and GDPR posture into a proposal-ready report in five minutes.
Contact
Let's talk.
Whether you need ad creative that performs or systems that run your operation without you: send us the brief, or your niche and your current cost per lead. We'd like to hear from you.