ultrathink unwrapped

foreword: When we started this newsletter three months ago, we believed one thing: if we put a strong signal into the world, we’d find the people who actually want to use AI at work—not just talk about bubbles + whether AI is real yet.

Butttt we didn’t know how many of you were out there.

Over 40,000 people now read ultrathink. And the pattern is pretty clear: you’re AI-interested, but allergic to BS. And you want workflows you can drop into your day without turning “learning AI” into a second job.

So consider this your end-of-year packing list… just the carry-on essentials. If you only have time to try one thing before the year flips, start here.

These are the playbooks our readers used the most in 2025—designed in collaboration with top subject-matter experts. 👇

core idea: About a quarter of the people reading this came from our AI-generated video pipeline on Instagram. Ads and video content no longer need to be months-long, million-dollar productions with craft services and on-camera talent. You just need tight writing.

it’s human-verified: After a decade directing traditional commercials for brands like Toyota and Red Bull, PJ shifted his approach. This year alone, his AI-assisted content has reached more than 350 million views. His workflow is simple and repeatable: ChatGPT for shot lists, Reve for storyboards, Google Flow with Veo 3 for camera work, ElevenLabs for voiceover, and a human editor for final polish.

in a nutshell: If you have taste, you can make something worth watching. And if you’ve ever thought, “We should make an ad,” now you can. Traction, leads, distribution, credibility—it all starts by putting something into the world.

core idea: You can create small, clickable prototypes in about 30 minutes that your entire team can react to. Instead of alignment meetings, hypothetical debates, or waiting weeks for engineering time, everyone can see a working tool and give feedback.

it’s human-verified: Logan shared his team’s workflow—using Gemini 3 and Google AI Studio to turn competitive analysis, voice notes, and chicken-scratch notes into micro tools on the spot.

in a nutshell: For the first time, the “idea person” can go from quietly suggesting to confidently championing change. It could look like a data visualizer your eng team would never prioritize, a lightweight workflow, or even a new landing page for your marketing team.

the core idea: There’s a reason Gamma hit $100M in ARR this year. Making presentations sucks. It eats time, plus PowerPoint hasn’t meaningfully evolved in decades. But, AI can handle the two most time-consuming parts—drafting decks and wrestling with formatting—so you can focus on what actually matters: shaping the story.

it’s human-verified: For Jon, the hardest part is getting that story onto the page. So, he kills his writer’s block by dumping context into Claude, iterating until the outline and copy are sharp, and only then handing it off to Gamma for structure, images, and charts. From there, he tweaks and adds branding.

in a nutshell: When decks take minutes instead of days, they start to behave like scalable assets. You can serve more stakeholders, prospects, and clients faster.

➰ go deeper: Every playbook we write includes the exact prompts the experts use, step-by-step handholding, and concrete examples you can adapt to your own work.

From PJ reshaping how an entire category works, to Jon saving dozens of hours of busywork, to Logan turning weeks of engineering and meetings into minutes, the takeaway is simple: time is money.

The people getting the most out of these workflows have taste, curiosity, and the discipline to turn experimentation into repeatable systems, which gives them room to push further than everyone else.

This is the fastest way to get better ChatGPT responses.

The tool now has a ‘stop being so peppy’ dial. Go to Personalization (sidebar/lower-left) and flip warmth, enthusiasm, headers/lists, and emoji to Less.

Try these settings for more professional outputs:

Did you know that nearly one in ten U.S. refunds is fraudulent? It adds up to about $76.5 billion a year in direct losses for companies.

return fraud = when someone buys something $$$, asks for a refund, and sends back a cheap fake instead (leaving the retailer stuck with junk).

the solution: Happy Returns (owned by UPS) uses AI to decide which returns deserve scrutiny. The system analyzes customer behavior when a return is initiated, then compares photos of returned items—taken by human auditors at processing hubs—with reference images and product data. The goal is to surface subtle mismatches that are easy to miss at scale.

the shift: Decisions that were once made thousands of times a day by frontline workers are now handled in stages. Software does the first pass on every return, while humans stay in the loop—opening flagged packages, photographing items, and making final calls when the system is uncertain or needs correction.

the result: Less than 1% of returns are flagged as high risk. About 10% of those are confirmed as fraud, with an average value of $261 per case. And for retailers, that translates into hundreds of thousands of dollars in annual avoided leakage.

p.s. I highly recommend watching the video to see the AI robots zooming around.

Does using more AI actually lead to better ideas?

Not necessarily, according to science.

A new study led by a Ph.D. candidate at the University of British Columbia and published in the Journal of Experimental Psychology examined how people generate creative ideas when working with AI.

tldr: Across two experiments, more than 450 business pros + working adults were asked to generate a plausible, original business idea (think: a new product, service, or market opportunity).

Participants used ChatGPT at varying levels of dependence. Some used it sparingly. Others leaned on it heavily.

Then their ideas were evaluated for creativity and business value using three methods: ratings from human judges, assessments from trained evaluators with entrepreneurial experience, and AI-based creativity scoring.

the trend: Interestingly, the strongest creative performance came from moderate human–AI collaboration.

apply it: According to the study, the boost didn’t come from better problem framing or sharper strategy. It came from idea diversity. When people used AI in moderation, they generated a broader range of novel ideas during their brainstorming phase.

As one X commenter put it, “the best AI imitates past consensus. The best human intelligence knows when to break away from it.”

📧 dont forget: Reply with the highest-ROI AI use case you currently employ. If it’s a banger, we’ll slot you in as a guest on our weekly show. View our calendar here.

Enjoy the eggnog. See you next week!