TL;DR — A few weeks ago I tested four AI tools on a build job: a website for my son's cricket academy. This time the job had nothing to do with code. The coach just wanted a banner he could post. Same four tools, totally different result. ChatGPT made the best image, Grok made the best video, Gemini wouldn't make anything, and Claude tried to solve a graphics problem by writing HTML.

If you read the last post, you've met my son's cricket coach. He runs MMCA — Maverick Master's Cricket Academy. Started in 2020, based in Bengaluru, genuinely good with the kids.

The website is live now and parents have started messaging him on WhatsApp. So last weekend he came back with the next thing he needed, which is the thing every small academy actually runs on:

"Can you make me a weekend batch banner? Something I can post in the parent groups."

Now, this is a completely different job from the last one. That first experiment was design and development — agents writing real code, running tests, deploying to Cloudflare. This one is just graphics. No repo, no deploy, nobody reviewing a pull request. Just: here's my logo, here's a sample I like, make me something I'd be happy to send out.

So I figured I'd run the same four tools again and see what happened. Same brief, same logo, everything on the default model with no special settings: ChatGPT, Claude, Gemini, Grok.

Here's roughly what I typed, the way a normal client would brief you:

Similar to this banner, make one for MMCA Academy (since 2020, logo attached). Weekend batch Sat 4:30—7, Sun 7—9:30pm. Add a small phrase like the sample. Be creative, keep it simple, but don't copy the sample exactly.

The whole test really came down to one instruction: be creative, but don't copy. Whatever each tool did with that told me everything.

Round 1: the static banner

ChatGPT got it on the first go. "WEEKEND BATCH. TRAIN. PLAY. GROW." Logo top-left, the "Since 2020" bit kept, timings in clean little cards, an enrol number, three badges across the bottom for coaching, skill, discipline. It clearly understood this was a flyer a coach would hand out at a school gate, and that's exactly what it gave me.

Then I asked for one more version — smaller logo, add a human hero this time — and it came back with "DREAM. PRACTICE. PERFORM." and a photoreal batsman walking out under stadium lights. Looked like a film poster. Two prompts, two banners I'd genuinely use, no arguing with it.

Claude is the one that made me laugh. I asked for a banner. It told me it would "create the MMCA banner as an HTML file you can download," ran six commands, and gave me a dark navy web page. "TRAIN HARDER. Play Smarter. Win Together," with Saturday/Sunday cards and a Register button. And to be fair it looked nice — same eye for design that won me over on the website build.

But it's not a banner. It's a landing section. You can't drop a <div> into a WhatsApp group and call it a poster.

That's the part worth sitting with. The tool I actually shipped the website with, the one with the best taste when the medium is code, defaulted straight back to code the moment I asked for graphics. It's wired to build. Ask it to build something and it's brilliant. Ask it to draw something and it quietly turns your design job back into an engineering one.

Grok at least made an actual image, which already put it ahead of half the field. Problem was it threw everything at the wall — three overlapping player photos, mismatched fonts, text everywhere, faded cricketers bleeding through the background. The exact opposite of "keep it simple." It knew what a banner was. It just didn't know when to stop.

Gemini gave me three attempts and three different ways of saying it couldn't quite generate that. The image guardrails kept tripping over what was, I'll remind you, a children's cricket flyer. I gave up after the third try. For a test that's purely about making graphics, a tool that won't make graphics doesn't really place.

Round 1 goes to ChatGPT, and it's not close. Two prompts, two finished banners, on brand, logo respected, the restraint Grok didn't have and an actual image Gemini wouldn't produce.

Round 2: make it move

Static was only half of what I was curious about. The real 2026 question is whether I can turn a flat banner into a short clip with a voiceover. So I took the good banner and gave two tools a single line: "make this banner live."

Grok nailed it. One prompt, six seconds. The banner came alive — the batsman moving, the paint-splash colours animating in, the timings resolving on screen — and over the top, a clean Indian-accented voiceover reading the academy out. "Where practice meets purpose." Honestly it looked like something an agency would charge ₹15,000 for.

One caveat, and it's the CTO in me talking: check the details. The phone number that showed up in the video wasn't the one I'd put in the source banner. Motion tools will happily rewrite text you wanted left alone, so you proof it before it goes out. Gorgeous result, but you don't post it blind.

Gemini, the same tool that wouldn't make me a still image, decided it would make me a video. Ten seconds of abstract paint-splash motion, a "WEEKEND BATCH" title card, a voiceover — but disconnected from the actual banner, and the on-screen text kept breaking apart and reflowing into gibberish. The idea was there, the execution wasn't. A trailer for a banner that never got made.

Round 2 goes to Grok. For motion plus voice off a single prompt, nothing else came close.

The scorecard

Image goes to ChatGPT. Video goes to Grok.

What changed from the coding test, and what didn't

Last time, on a development job, Claude was the tool I actually wanted to keep working with. The taste that made its website the nicest in that lineup is the same taste that made its HTML "banner" pleasant to look at here. None of that's a knock — it was just answering a question I hadn't asked. Point a code-native model at a design brief and it reaches for the thing it's best at.

Switch the job from build to draw and the whole ranking flips. A few things I'm taking away:

The best coding agent isn't automatically the best graphics agent. Obvious once you say it out loud, easy to forget when your whole team has standardised on one tool. The model that shipped my website couldn't make me a poster.

Guardrails are a product decision you feel as a user. Gemini turned down a kids' cricket flyer three times and then made a broken video of it. Whatever the safety reasoning, what I experienced was a tool that wouldn't do the simplest creative thing I needed.

And "done" beats "impressive." ChatGPT's banners weren't the flashiest pixels I've ever seen. They were finished, and I could post them in two prompts. Grok's video was the flashiest thing in the whole test and still needed me to catch a wrong phone number. For graphics, I'll take the tool that hands me something I can ship today over the one that wows me in the demo.

So I'm not picking a winner overall. I'm picking one per medium. ChatGPT makes the still, Grok makes it move. That's the actual workflow now — not one model to rule them all, just the right one for the thing in front of you.

The coach got his banner. Two of them, really, plus a video. Cost me a handful of prompts and one careful read-through.

Last time the lesson was that taste decided a coding job. Turns out taste decides a graphics job too. It just lives in different tools.

Which one are you reaching for when the job is graphics and not code? And has your favourite coding agent ever quietly tried to turn a design task back into a dev task on you? Curious to hear it.

AI #DesignTools #BuildInPublic #GenAI #Startup