I'm Benny. Creative Director by title, Art Director by trade. More than twenty years at agencies, including TBWA\Media Arts Lab, where Apple was the only client, on brands like Heineken, Samsung, Lincoln, GE, Vitamin Water, and State Farm. I've worked every rung from Junior Art Director to GCD, and the part I loved most never changed: making the thing. The thread through all of it is ideas that feel discovered, not manufactured.
I concept it, design it, shoot it, cut it, and get it out the door. Lately I've added AI to that list, as a proficient learner rather than an expert, which turns out to be a handy thing to be: curious, hands-on, and honest about what the tools get wrong. Open to freelance, contract, and full-time roles with teams that want someone who ships.
The proof is Masters Of Sight, a living archive of taste I founded and built with AI: a curated record of the photographers, filmmakers, designers, and game makers who taught me how to see. When I'm not working on brands, I'm usually behind a camera, in the woodshop, or developing original story worlds.
Different mediums. Same pursuit: meaningful ideas, thoughtful craft, and work that leaves a mark.
Masters Of Sight started as a private list: the photographers, cinematographers, directors, designers, and game makers who rewired the way I look at things. Twenty years in advertising turned that list into an obsession. The obsession needed a home.
I'm a learner, and a proficient one. I directed Claude Code to build the whole product, one prompt at a time, and I kept the veto. This page covers how it works, how I work with the AI, and the places it confidently got things wrong.
Live
A working product at mastersofsight.com. Not a mockup, not a side gallery.
Findable
A real, crawlable page for every master, built to be found.
Free accounts
Sign in with a magic link or Google, save masters to your own Board, and delete the account yourself, cleanly, anytime.
The archive is the database. AI is the navigation layer.
A quick walkthrough of the live site.
The list that got out of hand
Every master is hand-picked. No algorithm, no committee, no submissions inbox. Roger Deakins shares a wall with Hayao Miyazaki. Paul Rand sits across from Hideo Kojima. Thelma Schoonmaker gets a room of her own, because a cut is a fingerprint too. Different tools, same nerve. The connective tissue was never the medium. It's intent.
The archive keeps widening. Editors and colorists now stand beside the directors and cinematographers whose films they shape. The Stack, The Moment, and The Encounter carry the argument onto Instagram and YouTube, one scene, one frame, one obsession at a time.
The problem with a list
A list can't be browsed by feel. It can't be saved, filtered, compared, or argued with. It tells you who, never why, and the why is the entire point. I wanted something where a creative who knows how a piece of work should feel could find the person whose work would teach them how to get there.
So the job was to turn a private canon into a product. I considered Instagram, but an entire archive living inside someone else's feed felt wrong. Notion worked for a while, then I hit the ceiling on design and functionality. I wanted to control every pixel, and I didn't love the idea of years of work depending on a service that could change its terms or disappear overnight.
What it does today
The archive. Photographers, cinematographers, directors, designers, game worldbuilders, editors, and colorists, all in one room, filterable by discipline.Search that talks like you do. Ask it the way you'd ask a friend: "Who shot...?" Results update as you type, and a "Follow the thread" row points you to neighboring masters.
Real pages for every master. Not a name and a thumbnail. A tight bio, five specific traits, major works, and a gallery that runs edge to edge."If you like..." A hand-picked row that crosses disciplines on purpose.A Board of your own. Sign in, save the masters who haunt you, and the Board sorts them by discipline. Free, and you can delete your account yourself, cleanly, anytime.
How it's built
Masters Of Sight is a static site on Eleventy, deployed automatically from GitHub, with images served through Cloudinary and Supabase handling free accounts, by magic link or Google. Every master has a real, crawlable page. It's fast because almost nothing about it is dynamic.
Claude Code is my development partner. I direct, it drafts, and I decide what survives. I run the curation, the writing, and the video, and I direct the build, which means every decision has one owner and nobody to blame.
How I work with the AI
I'm still learning this, and I've learned it by doing. Here's the system I landed on, built to catch my mistakes as well as the AI's: a brief, a QA gate, and a veto. It keeps evolving.
A living rulebook. One working document holds the editorial law, the design precedents, and the reasons behind every call. When a decision gets made, it gets written down once, so the AI never re-litigates it.
One step at a time. I give the AI one task, read what comes back, then give the next. Layered prompts hide mistakes.
A gate before anything goes live. Every change is built and checked locally. Nothing ships until I say so, and an automated validator refuses to deploy data that breaks the rules.
A human makes the final call. Always.
Where the AI got it wrong
This is the part most case studies skip. Mine includes it, because directing an AI well is mostly about catching what it confidently gets wrong. Each mistake below became a rule.
The design had no soul. First pass: generic fonts, a boring layout, no point of view. It was competent and forgettable. I tested a light mode and killed it because it lost all impact. The site stays dark, and I own the look.
It felt clunky. Buttons were off. Things shifted on click. Mobile behaved like a different site. None of that shows up in code review. It took a human eye and a thumb. Now every visual change gets checked on desktop and mobile before it goes anywhere.
It couldn't see the ripple. Change one thing and something else quietly broke, and the AI would never flag it. A video sitting in the wrong spot in one gallery taught me that. It became a rule, then an automated check that now covers every profile. The AI also writes prompts from stale memory of the file, so every instruction now starts by re-reading the live code instead of trusting last session's line numbers.
It built things that worked and weighed too much. Pages loaded fine and carried far more image data than they needed. I caught it, and we capped image sizes at render time.
AI is fast. Knowing what's good, and catching what isn't, is still the job.
The intake workflow
Adding a master isn't a form, it's a process. I research the person and go deep into their work, using their own sites, interviews, articles, and film credits to write a bio and traits that stick to a strict editorial standard. Film stills come from Shotdeck and IMDb, and photography comes from the artists' own sites and public-facing partner sites. Images live on Cloudinary. Videos link directly to YouTube or Vimeo, so everything points back to the original source.
Early on, a single profile took a long time because my process wasn't refined. Today the research is still entirely mine, but a quick intake form hands the details to Claude, which turns them into code. Claude Code builds the page locally, I check it, and only then does it go live. Every master clears the same bar, and a human makes the final call.
What's live, and what's next
Live: the full archive, master pages, live search, Boards, and an ongoing run of video series on Instagram and YouTube.
Next, in this order: first the content layer, with deeper entry points and a "why they matter" paragraph for every master. Then a metadata schema that describes each master's work in structured, searchable terms. Then Find Your Vision, an AI layer that searches my editorial canon instead of the open internet.
The order matters, because the AI is only as good as the taste underneath it.
