👋 Howdy to the 613 new legends who joined since our last edition! You are now part of a 579,491-strong tribe outperforming the competition together.

LATEST POSTS 📚

If you’re new, not yet a subscriber, or just plain missed it, here are some of our recent editions.

🌊 Why Payroll Is Still Broken (And Fixing It). An interview with Ayush Sharma, Founder & CEO at Warp.
📔 The Fine Art Of Disagreeing & Committing. Why consensus is where good decisions go to die, and the seven-step fix.
🌏 How Product Teams Are Going Global In Days. An interview with Minh Cung, Co-Founder & CEO at Hyperlocalise.

PARTNERS 💫

You take the calls, Granola takes the notes. Granola transcribes your calls from your computer while you jot down a few bullets, then writes them up. At the end of the day, you have notes on all six, not two good ones and four you meant to finish.

Get 3 months off any paid plan with code OPENSOURCECEO.

Everyone’s talking about getting back to the office. Candidates are still searching for remote.

Ashby’s data from 1,200+ startups show that remote roles draw 42% more applications and 9% higher offer acceptance rates, even as remote jobs slipped from ~80% to ~60%. For more insights, read our free State of Startup Hiring report.

_{Interested in sponsoring these emails? See our partnership options here.}

HOUSEKEEPING 📨

As a busy (and often stressed) founder, I need to find tools to switch my brain off from time to time, and the RYUCAMP YouTube channel is one of my faves. This young South Korean guy takes his dog into the woods, usually in epic premium tents, fully decked out, and surrounded by the soul-soothing sounds of rain.

The best thing about the channel is it’s non-speaking. Just ASMR of camping, cooking, playing with your pet, and nature. Give it a shot the next time you’re feeling like you want to zone out and enjoy a bit of downtime. Also, great to put on in the background while doing easy work. Enjoy!

GUEST POST 💄

Why AI Pivoted To Cute

In 1894, brothers André and Édouard Michelin were at a trade fair in Lyon and noticed that a stack of tires looked a bit like a man without arms. Four years later, a French cartoonist called O’Galop (real name Marius Rossillon) drew that idea for a Michelin poster. The marshmallow-shaped corporate good guy known as Bibendum—and later the Michelin Man—was born.

Bibendum was the first real time a company turned an invisible, boring, hard-to-tell-apart product into a lovable character. You can’t see a tire’s puncture resistance or sturdy tread. So Michelin gave Bibendum a body, a personality, and in the early days, even a cigar and a fancy pair of pince-nez style spectacles.

Today, artificial intelligence has a bad wrap. The fear around job loss, proliferation of the towering data center, and general existential fears for the human race has the general population skeptical. Effectively, AI needs a rebrand (and no, not to ‘super intelligence’). For our big tech overlords that rebrand is coming in the way of mascots. Meta’s Muse has Jolly, OpenAI launched Dots, and xAI did the same with GrokBot.

Today, my buddy Isaac Peiris is back to break why companies are following Michelin’s century-old lead here. There is no better marketing psychology nerd I’d rather pick the brain of than him. Without further ado, here’s Isaac.

Three launches, one idea

Hey, Isaac here! In about a month, three of the biggest companies in AI shipped almost exactly the same product. Then all three did something nobody in a strategy meeting would have predicted a couple of years ago. They gave it a mascot.

Meta went first with Muse, its always-on personal agent. A couple of weeks after launch, Mark Zuckerberg walked out on stage with Jolly, a soft little character with beady eyes and a small upturned grin, while Meta's Chief AI Officer Alexandr Wang spent the day posting Jolly memes on X. Within a week, OpenAI had launched Dots, which lets you name your agent and kit it out with accessories like a beret or a bow tie. A month before either of them, xAI had released GrokBot, where every agent gets a shape and color you pick yourself.

These are companies with enormous compute budgets and some of the best-paid engineers on the planet. When it came time to put their most ambitious product in front of normal people, every one of them reached for one of the oldest tricks in advertising: a cute character.

I write about brand moves like this every week in Brand Chemistry, and this is the most interesting one I've seen in AI this year. Not because mascots are new, but because of what they reveal about the product sitting behind them. If you're building anything people can't see working, can't easily tell apart from the competition, or have to trust with something important, the same logic applies to you.

The unfashionable tactic that kept working

Here's the funny thing about mascots. Marketers spent the best part of three decades walking away from them, while the data quietly kept saying they worked. System1, the ad-testing firm, went back through every long-term campaign entered into the IPA Effectiveness Awards since 1998. Those are the UK's big awards for advertising that actually moved the numbers, so it's a solid dataset of what works over time. Campaigns built around a character were far more likely to report very large market share gains: 41% of them, against 30% of everything else. On profit, it was 34% against 26%.

Over roughly the same stretch, the share of those long-term campaigns built on a character or a slogan fell from 41% in 1992 to 12% by the mid-2010s. Brands chased sleek, minimal, grown-up identities, and mascots got shoved back into the cereal aisle where serious marketers assumed they belonged.

Duolingo was the loud exception. Its unhinged green owl became bigger than the app itself, and people followed the bird for the chaos rather than the Spanish lessons. So when Duolingo said this year it would dial the owl back on social, I argued it was a big mistake. A character hands people two things most products can't: something to instantly recognize, and something to actually feel something about.

What insurers figured out decades ago

There's one category that never gave up on characters, and it isn't cereal. It's insurance. Think about what an insurer actually sells. There's nothing to hold, just a commitment that only matters after something has gone wrong. There's nothing to compare either, because one policy reads exactly like the next. And the whole thing rests on the company picking up the phone after a car crash or a flooded kitchen. Invisible, interchangeable, and completely dependent on trust. That's why GEICO has a gecko, and why Compare the Market built an entire universe around a meerkat with a Russian accent.

Now run a personal AI agent through the same test. It does its work where you can't see it. It does roughly what its competitors do. And it asks for more of your private life than any insurance policy ever has. Agents don't just share insurance's three problems; they've got a heavier version of every one. Let's take them one at a time.

Problem #1: The work happens off-screen

Muse, Dots, and GrokBot all run on a cloud computer you never look at. You hand over a task, go make a coffee, and come back to a finished result. Everything in between happens out of sight. That's great for convenience and terrible for building any kind of relationship with a product. It's hard to feel anything about a process you never get to witness.

A character fills the gap by giving the invisible work a body. X's design team built GrokBot's icon in code, so its idle and working states melt into one another, and because each agent gets its own shape and color, a quick glance tells you which one is busy. OpenAI is already talking about teams of Dots working together. Meta went furthest, building a Tamagotchi-style wearable for Muse so the agent physically sits with you, away from any screen. At that point, the character stops being decoration. It's the only part of the agent you can actually see and touch.

Problem #2: Everyone ships the same features

Read the three launch posts side by side, and the agents are almost indistinguishable. They run around the clock, log into your accounts, and string together multi-step jobs while you get on with your day. Whatever one ships this month, the other two will likely have by next quarter.

Jolly is running that experiment at internet scale. People have adopted it and remixed it into their own memes (some of them very much not safe for work). Partners have joined in too: when GitHub and Spotify announced their Muse integrations, each posted its own take on Jolly. Nobody remixes a logo for fun. They'll happily remix a character, and every version points straight back to Meta. As a founder, this is the bit I'd pay most attention to. Features are a race that never finishes. Jolly, and whatever affection people build up for it, can't be copied in a two-week sprint.

Problem #3: You're asking for the keys

This is the big one. Muse can read your email, manage your calendar, and check out with your credit card. That's a far bigger ask than any insurance policy makes, and it's where these characters are working hardest. They do it in two ways. The first is disarming you. A 2009 study in the journal Ethology showed people photos of baby faces with the ‘baby schema’ features turned up and down: rounder head, bigger eyes, smaller mouth. The stronger those features, the cuter people rated the face and the more they wanted to look after it. Now look at Jolly, Dots, and GrokBot. Big round shapes, dot eyes, barely any features, the sort of thing you'd win at a fairground. The software that can spend your money has been drawn to trigger the same instinct as a puppy.

The second is making it yours. Researchers call it the IKEA effect, after a 2012 study where people valued storage boxes they'd assembled themselves more than identical ones that arrived pre-built. Put effort into something and you value it more. Naming your agent, picking its color, giving it a bow tie: each one is a tiny bit of effort, and together they add up to attachment.

Notion got there before any of them. When I broke down how Notion made AI cute, the standout decision was refusing the sparkle icon every other AI product uses. They drew a small face instead, just eyes, eyebrows, and a nose, and later let people name and accessorize their agents. Their Head of Creative said at the time that the first job wasn't proving the technology; it was getting people comfortable with being helped.

Clippy already ran this experiment

We've actually got a 30-year-old case study for all of this, and it's a paperclip. Clippy shipped with Microsoft Office 97, and most people remember it as the assistant that popped up uninvited to ask if you were writing a letter. Microsoft announced its retirement in 2001 and switched it off by default in Office XP.

But when Notion asked Steven Sinofsky, who ran Office at Microsoft, whether he'd give an assistant a face again, he pointed to a Stanford study called The Media Equation. People tend to assume software is right and blame themselves when something goes wrong. A friendlier face shifts that: people stop blaming themselves and start questioning it, treating it more like a collaborator than a tool. And the character outlived the product by a mile.

The lesson cuts both ways. A character can build affection that lasts decades. It can also become the punchline if the product behind it is annoying or doesn't deliver.

The trust is on credit

Not everyone's charmed. Ed Zitron, the PR founder and well-known AI critic, told Fortune the mascots are “entirely marketing.” Fairplay, a children's advocacy group, accused Meta and OpenAI of selling AI with children's iconography. Some of that lands, because the characters can blur what's really going on underneath. GrokBot lets you give each agent its own name, shape, and color, so each one feels like its own little creature. But xAI's own documentation says all your agents share a single cloud computer, including its files, browser sessions, and app logins. Lots of cute separate characters, one shared set of keys.

Muse has a similar gap. It's flying, with 5 million downloads in its first three weeks. But its Confidential VM, the feature that would lock your data behind a key only you hold so that not even Meta can read it, hasn't launched yet. It's slated for ‘later this year.’ So the characters are building trust ahead of the product. People hand over access because the thing asking looks friendly, before the safeguards are fully in place. The mascot opens the door. Only the product gets to decide whether people leave it open.

Playbook

• Find your trust gap first: Before sketching anything, ask whether people can see your product working, tell it apart from competitors, and trust it with what it touches. A character earns its place by fixing one of those, not by being cute.

• Make the invisible visible: If the value happens in the background, give it states people can watch: working, waiting, done, broken. GrokBot's morphing icon and Notion's face that falls apart on errors both turn back-end activity into something you can read at a glance.

• Let people make it theirs: Names, colors, small accessories. A little customization creates ownership, and ownership creates an attachment competitors can't easily poach.

• Design it to be remixed: Keep the character simple enough that fans and partners can redraw it. Every meme and partner version of Jolly is free distribution a logo would never get.

• Don't let the mascot write checks the product can't cash: Ship the safeguards alongside the face and be upfront about what the product really does. Otherwise, you've built a Clippy: remembered fondly, trusted by nobody.

Future

Within a year, I'd expect most personal agents to do roughly the same jobs at roughly the same quality. What won't converge is everything sitting on top: the character people recognize, the one they named themselves, the one their friends turn into memes. That's the part that compounds.

The real test comes later, though. Right now Jolly is winning the attention game. Next year, the conversation will be about breaches, mistakes, and what these agents did with people's inboxes, and no amount of fluff will carry a product through that. The winners will be the companies whose products grow into the trust their characters ran up. If you're a founder thinking about giving your product a face, that's the order of operations: earn the right to be trusted, then make it lovable.

Fun facts

• Notion binned the wizards: Early sketches for Notion's AI included wizards in different hats and a wand that turned into a pencil or a brush. They scrapped it because ‘magic’ made the AI harder to understand, not easier.

• 141,000 likes for a paperclip: Microsoft asked for 20,000 likes to bring Clippy back as an emoji in 2021 and got seven times that.

• Fastest to the feed: Muse passed 5 million downloads in its first three weeks, with Jolly doing a lot of the heavy lifting on social.

Extra reading

• The trust economy - July, 2026

• The Future New Media: a16z - June, 2026

And that’s it! You can follow Isaac on LinkedIn and X (Twitter), and make sure to subscribe to his newsletter, Brand Chemistry, as well!

BRAIN FOOD 🧠 

TWEETS OF THE WEEK 🐣 

TOOLS WE RECOMMEND 🛠️

Every week, we highlight tools we like and those we actually use inside our business and give them an honest review. Today, we are highlighting Airwallex—Open multi-currency business accounts, collect payments globally, save time and cost, and automate bill pay.

See the full set of tools we use inside Athyna & Open Source CEO here.

HOW I CAN HELP 🥳

P.S. Want to work together?

1. Hiring global talent: If you’re hiring tech, business, or ops talent and want to do it 80% less, check out my startup, Athyna. 🌏

2. See my tech stack: Find our suite of tools & resources for both this newsletter and Athyna here. 🧰 

3. Reach an audience of tech leaders: Advertise with us if you want to get in front of founders, investors, and leaders in tech. 👀