Everything in between.

Every content project used to start the same way: clicking through a client’s site page by page, copying text and saving images one at a time.

Free tools only got me part of the way: a page here, some text there, images left behind. Paid tools did the whole job, but it was a job I could build myself. So why pay for it?

So I built a scraper that crawls the whole site in a single pass. It snapshots every page, pulls down every image and every line of copy, and links all the pages together, so the result works as an active archive you can click through, mapped to the site’s real structure.

That local copy becomes the reference point for everything that follows. I can check what a page said before the rewrite, pull an asset without chasing the client for it, and feed the entire site straight into my competitive analysis.

• Full-page snapshots
• Every image, downloaded
• Every page linked: a clickable archive
• URL and site-structure map
• Dated local snapshots

Competitive audits used to mean days of reading rival sites and building spreadsheets that were out of date by the time they were finished.

This is a set of LLM-powered analysis and reporting skills, built on Claude, that measure a client’s site against its competitors, at two depths. For a quick read, it takes a fast snapshot of each site and reports how they position themselves. For the full picture, it runs on content from the scraper, so every page of every site is indexed and searchable.

With the whole field indexed, I can cross-reference anything, anytime: the topics competitors own that the client doesn’t, where messaging overlaps, which services are underexplained and where the openings are. Those gaps turn directly into a content strategy and a prioritized plan, backed by evidence.

• Quick snapshot or full-depth audit
• Cross-site content index
• Content gap analysis
• Strategy and priorities out

Every client has hundreds of files: sales decks, spreadsheets, research, brand guidelines, product docs. Most of what they know is buried in formats nobody has time to reread.

This retrieval-augmented generation (RAG) system turns that pile into a structured knowledge base. The core is an intake engine I built from scratch. It converts every deck, spreadsheet, PDF and document to clean text, verifies the conversion against the original, then breaks each file into structured fact cards: every claim, figure, quote and price on its own line, traced to the exact slide, page or row it came from. Each card passes a quality check before the next file is opened.

Those cards are tagged against a fixed set of themes and indexed for retrieval. Teams ask questions in plain language, retrieval pulls the exact evidence, and answers come back like a senior strategist who has read everything, with a citation behind every fact.

It’s built to be trusted. It never answers from memory, and it says so plainly when something isn’t in the files. Outdated, conflicting or unapproved content is flagged, and that flag follows the claim into every answer it touches. A person signs off on everything before it enters the index.

• Custom ingestion pipeline
• Every fact linked to its source
• Conflict and quality flags
• Human sign-off before anything goes live

The banner tools on the market couldn’t keep up, so every campaign meant rebuilding the same ad in a dozen sizes, then again every time a headline changed.

Banner Studio fixes that. You design once at a standard ratio, and it adapts the layout to every size the media plan calls for. It ships with ten standard sizes, and you can add as many custom ones as a campaign needs. Then you fine-tune only the sizes that need it.

It’s a full production tool, not a resizer. It has layers, a timeline for motion and flexible export, so the same design ships as static images, animated GIFs or HTML5, whatever the channel needs.

• Design at standard ratios
• 10 presets, unlimited custom sizes
• Layers and timeline animation
• Static, GIF or HTML5 export

Small businesses need real-looking photos of their people, food and products. A shoot costs time and money most of them don’t have, and stock never looks like them.

So I developed a set of prompt engineering and workflow methods that produce imagery that reads as real photography: people, food, products and spaces, in any pose, style or setting, matched to the brand and the audience it’s for.

It’s tuned for efficiency as much as quality. Every step is designed to land the shot in fewer attempts, which means fewer wasted renders and less spent on credits. Better images, at a lower cost per image.

• People, food, products and spaces
• On-brand style control
• Fewer wasted renders
• Lower cost per image

A full-time role, my own agency and a constant stream of everything in between. Sticky notes and calendar reminders couldn’t keep up.

So I built an agentic AI assistant around how I actually work. Each morning, before I’m up, it reads my email and calendar and has a brief waiting: what changed since yesterday, a prioritized list for today and replies drafted for me to review. It even volunteers for up to three jobs and does most of the work on them.

It’s one workspace for everything: to-dos sorted by client and priority, a hub for every project that it keeps current, a canvas for mapping notes, and a place to dump ideas and sort them later. Every correction I give it becomes a standing rule, so it gets sharper the longer I use it.

And nothing happens without me. It works overnight in read-only mode, never sends an email on its own and changes my priorities only when I approve them.

• Morning brief, ready on wake
• To-dos by client and priority
• Prompt-to-storyboard, frame by frame
• Learns from every correction