Hey p(ai)rty people,

You know that feeling when you finally master something right before it becomes completely irrelevant? That's my recent career in a nutshell, and this week is no exception.

But first, the important stuff: We had a great week as a family gearing up for our Paris trip next week. (No newsletter next week—I'll be eating croissants while Maddie, who's been grinding Duolingo tokens, tries to order us dinner in her best Parisian accent.)

James experienced his first Yom Kippur breakfast and chose a plain bagel over everything. I'm trying not to read too much into what this says about his future taste.

Linus almost caught a squirrel. A nearby guy said, "I didn't know they could catch them!" Brother, have you ever met a 60% cattle dog, 30% pit bull, 2% lab, 2% bear, 2% lion, 4% “oh my!” … didn’t think so.

Anyway, I've been deep in the trenches testing Sora 2 and Veo 3, and I've come to a profound realization: I've been working like it's the 21st of September. Which, in AI years, is basically the Stone Age.

Sora 2 = The "just make something cool" button
Veo 3 = The "I know exactly what I want" tool

Choose your fighter based on how much you want to micromanage your AI.

Massive copyright infringement aside, here's a sentence I never thought I'd write: Sora 2 works better when you're lazy.

I'm serious. I spent hours crafting the perfect prompt. Detailed shots, specific lighting conditions, exact pacing notes. You know, like a professional.

The result? Not absolute garbage, but something that really just went off and did its own thing.

The biggest thing here, is your lack of control.

Note: the quality of the GIFs below are worse than the videos themselves!

The lesson here?

Sora 2 is the first AI tool that actively punishes you for trying. It's like it knows when you're overthinking and just... decides to mess with you.

✓ Product rendering so good I've started questioning if I've ever seen a real product
✓ Automatically creates multiple shots in one video (because apparently it knows better than you)
✓ No start frames, end frames, or pretending you understand cinematography
✓ The shorter your prompt, the better it works (finally, a tool that rewards my attention span)
✓ Clients will 100% think you hired a production team

✗ You have very limited zero creative control (embrace the chaos)
✗ Can't specify shots, which is great until the client asks "can we change that one part?"
✗ I have a 847-word prompt document I can never use
✗ No API yet, so I'm clicking buttons like a caveman
✗ That watermark bounces around like it's mocking me

The Verdict: Sora 2 is what happens when AI gets so good it doesn't need your input. It's like hiring someone way more talented than you and then realizing you're not even sure what your job is anymore.

Having said that, once the API comes out, I think EVERYTHING will change.

Meanwhile, Veo 3 is over here validating my entire existence by actually caring about my input.

It's the tool for control freaks (hi, it's me [didn’t Taylor Swift say that or something?]) who spent years developing skills and would like those skills to remain relevant for at least another fiscal quarter.

Note: the quality of the GIFs below are worse than the videos themselves!

Please reference this my previous post should you want to explore this workflow in more depth!

✓ Your prompt engineering skills are still useful (for now)
✓ You can actually explain to clients what you did
✓ Works in a workflow with other tools (job security!)
✓ Predictable outputs (no surprises, like a healthy relationship)
✓ Start frames give you control (the illusion of relevance)
✓ 9:16 aspect ratio finally exists (only took them forever)

✗ Requires you to actually know what you're doing
✗ Each asset takes 6x longer than Sora 2
✗ Still only one frame to start (greedy)
✗ You need to learn Midjourney, Nano Banana, Veo 3, AND CapCut
✗ Clients will ask "why does this take so long?" (because I MATTER, KAREN)

The Verdict: Veo 3 is for people who went to design school and would like that to mean something. It's the participation trophy of AI tools. And I love it.

Okay, deep breath. Let me try to end this on something resembling optimism.

Here's my current coping mechanism: Both tools have their place. For now.

Sora 2 is perfect for:

Veo 3 is perfect for:

The strategy I'm going with:

It's not much, but it's honest work. For now.

Until the API drops and we all become prompt supervisors.

Or whatever job title we'll use to describe "person who asks AI to do things."

Here's what actually matters:

Sora 2 is the best tool for rapid ideation. When you need to wow clients fast or test ten different directions, nothing comes close. The quality is absurd, and it gets better when you stop overthinking.

Veo 3 is what you use for precision work. When brand guidelines matter, when you need repeatable processes, when the client needs to understand what you did—this is the tool. It rewards expertise instead of punishing it.

Use both. They're solving different problems. Sora 2 for discovery. Veo 3 for execution. It's that simple.

The API release will change everything, sure. But that's always been true in this space. The people who win are the ones who figure out how to use these tools faster than everyone else catches up.

So yeah, I'm building workflows that might be obsolete in six months. I'm also building the instinct for knowing which tool to reach for and when. That instinct—understanding the trade-offs between control and speed, precision and iteration—that's what actually transfers.

The tools change every week. The strategic thinking doesn't.

I'm not worried about becoming a "prompt supervisor." I'm focused on being the person who knows which prompts to write, when to write them, and when to just get out of the way and let the AI cook.

Until next week (or two weeks, because Paris),

Best,

Will