Grant Lee

Everyone is sick of generic AI, including us.

Gamma was the first AI presentation platform to reach real scale. But now that AI tools are everywhere, everything looks the same.

Earlier this summer, we came to an uncomfortable conclusion: without a step change in visual variety, Gamma’s output would look and feel too similar to all the other AI tools. The entire point of a presentation is to help someone "get" your idea, to see it as unique and do something with it.

We started over. Today we're shipping Gamma 5, the biggest product overhaul in our company's history.

We rebuilt Gamma to create presentations that actually look like you and your brand. Your teams can design anything they dream up, following your brand's aesthetic or something entirely new, with every frontier and image model under the hood. And the same goes for your docs, social assets, and graphics.

We revamped our agent, design tools, editing, import, export, connectors, and more.

Get your ideas out there.

You can feel it in the first 3 seconds.

AI-created “slop” filled with the average of every pitch deck it has seen. No one remembers anything on that deck.

Producing anything is faster now: text, slide decks, music, images, video, everything.

But it does not create any actual impact. If the idea behind the communication isn’t from you, but a machine, it’s useless and might even be called out.

The most beautiful deck in the world is still mediocre if the idea inside it has no thought.

Recently @Clay put out some definitive rules on writing:

- Stand behind every sentence.
- If someone asks what a line means, you do not get to say the model wrote it.
- A short prompt that produces a long document hands the real work to the reader. That’s no good.

The direction is the same: ideas are to be owned by humans only. AI can't be held accountable, and that makes all the difference.

We are starting to see some acceleration in discovery rates due to AI, in areas that matter:

-Cybersecurity research is showing a clear jump.
-Math is moving.
-Algorithm work is still lagging, which is useful data on where current tools still struggle.

I think what's happening is a compression of the time taken between asking a question and getting a verified answer. AI is making this cycle shorter, and the effect is compounding now.

Founders who treat AI as just a productivity feature will miss the big shift, as those use cases are still based on the old ways of working.

The ones who understand this and act accordingly will build companies completely different from what we have today.

Here is an easy way for founders who are overwhelmed or have too much on their plate: get the focus back on product.

Most AI companies lean so heavily into research that the product work feels secondary.

Remember that a product only improves as fast as the team can ship, measure, and adjust. The quality of that loop depends entirely on the people inside it and the standards they share.

The founders at Cursor spend a large share of their time (30%-40%) on hiring and culture. This seemingly “soft” priority is the mechanism that lets a team iterate at a speed unmatched by research-heavy groups.

When hiring and culture are done right, the product loop stays tight. They build the foundation from which everything else compounds.

"The past is kind of there to fool you if you think this technology is similar to the other ones."

According to Bill Gates, AI is fundamentally different from other technologies like the microprocessor, the internet, or electricity.

A few thoughts I had on this:

1. As we have seen with technologies of the past, they first eat the obvious tasks, and then raise the value of the people who can handle the technology and steer it to show never-before-seen results. That's the new value that gets added.

2. When a capability gets extremely cheap or free, the world doesn't do less of it, but more. This happened with steam engines and steel, and is now happening with cheap intelligence. Tasks that were too expensive to attempt earlier, become normal. What's not on anyone's radar right now is all the new kind of work this technology makes possible.

3. Agreed that AI as a tech is way different than earlier technology. It can actually think. When coupled with what's happening in hardware, we would soon be seeing intelligent robotics all around us. But who manages them? Humans? Who designs them with real world sensibilities? Humans. The last mile of taste, conviction, and real judgement from people who have done the reps stands.

Past technologies have taught us one thing: Humans will always be valuable, resourceful, creative, and above any man-made tech.

I keep coming back to this idea that the best AI products make users more capable.

In 2005 playchess(dot)com had a tournament which let anyone take help of any chess supercomputer they had. Grandmasters and highly rated players showed up, ready to fire up their strongest chess engines. Even the most powerful chess-specific supercomputer, Hydra, an old rival of IBM, entered. It didn't even make it to the semifinals.

Grandmasters, with their chess engines, couldn't make it too. The grand tournament was, crazily enough, won by two amateurs, Cramton and Stephen, who hadn't played a competitive game in six years. They didn't have any supercomputer, and used a normal laptop with three ordinary chess programs. They were on their own for the most part.

It was unbelievable so the win was challenged. Garry Kasparov himself checked and verified that the winners were legit, and didn't take any secret help from a grandmaster or supercomputer.

How did they do it? They mainly used their judgement, and followed a process. They trusted one computer program to help them during the opening, and the other two programs to guide them through tricky middle-game positions when they needed help. Of course, this was over two decades ago, and chess engines have dramatically improved since then, closing the advantage.

Cramton and Stephen were good chess players, but theoretically no match for the grandmasters in the tournament. What helped them win is that they were better operators of the tools (even inferior ones) they had with them.

It's not enough to just be productive due to the tools available. There should be permanent growth in skills and capacity.

An example I often use: make better photographers, not a more feature-rich camera.

It sounds counterintuitive maybe, but creating skill among users should be the north star.

It may sound valid in theory, but reality hits different.

I believe AI does not drain the "cognitive commons", but widens the highway on who can participate, while also speeding up human judgement.

Five points:

1) There are two kinds of mastery. Hard problems create a kind of internalized mastery, while distributed mastery comes from the craft of handling AI tools. The danger is being AI-centric on the surface without the internalized mastery. This is fixable by right recruiting.
2) Staying actively engaged with AI (not just blindly accepting its output) creates a more usable end result. People need to stay in the loop of judgement.
3) Codified knowledge is often more prone to be automated. First drafts, templates, syntax etc. However, tacit knowledge like taste, judgement, client trust, knowing which idea has legs, is all human domain.
4) With AI, each person carries more "surface area" as they keep making the right calls that build judgement along the way. This is the desired combination that compounds.
5) Junior talent is the pipeline for future senior hires, and is definitely not an item to shrink. AI is fast-tracking the training with automated smart work replacing hard work.

What we are seeing everyday is that a much larger set of people are super-productive faster than what was previously possible. There is no use worrying about getting a piece of the pie, as AI makes that pie larger.

The risk is in labeling AI as a replacement when it's actually a force multiplier.

Can you train AI to have taste? We think it's worth finding out.

I’m thrilled to announce our acquisition of Lica.

@pritopian and @purvanshi_mehta share our belief that visual communication should be easy and accessible. Together, we plan to build a new research team exploring what comes next:

How do design, layout, and visuals shape attention?

Can qualities like taste and trust become learnable signals?

And how can we help our customers make their ideas matter?

Visual intelligence remains remarkably jagged at the frontier. We’re investing in the research and talent needed to help AI truly understand design, taste, and communication. Our ambition is to create new visual formats that surprise and delight without becoming harder to use.

The team has already hit the ground running, and we can’t wait to share more in the coming weeks.

This is probably the best "Founder Mode" podcast I have listened to this year.

@travisk built one of the fastest-growing companies in history and is both reflective and intense while talking about his experience.

He is naturally drawn towards problems that most people would rather avoid.

The difficulty of a problem can be part of the opportunity. The thing is difficult for a reason, to hinder everyone from doing it. That 'difficult' toll gate is necessary, but it doesn't mean it's impossible. Sometimes this isn't fully understood by founders, who club "extremely difficult" with "impossible" almost by default.

I remember my friends that were early at Uber sharing all the insane obstacles they faced, as the company navigated local "mafia" like regulation, city politics and legal issues. Every city had a new set of problems, and Uber did this globally, even in China.

Travis compares entrepreneuring to a competition in pain tolerance. To him it's essentially proving that you can take more pain than the other guy.

There are periods when a company can feel like that, a huge collection of problems loosely held together by a bank account. A founder has to, of course, run towards the problem that hurts the most. It's important to remain effective under pressure.

For Travis, pain tolerance became the entire operating system. It's better not to be numbed by it, and actually let it bother you so that you take inspired action.

But it's fun to play the game. "Find your sport" is good advice, as you would be good at the game if you love it. Before you find the PMF for the world, you have to see how your brain works and find a PMF for yourself, the problem you will love dealing with, for years.

Bring order to the chaos within you and then implement the same in the world. That's what founders do, they find some corner of the world that feels chaotic and try to create systems that bring order to it. That's what Uber did, and what his new company Atoms is doing.

"Chaos is the law of nature. Order is the dream of man." - Henry Adams

Exceptional conversation. Shout out to @davidsenra for this.

"Building the tech is the easy part now, getting people to believe is the hard part."

@lulumeservey is at her usual best in this.

AI has made average communication infinite. Polished launch videos and "good enough" content is now the norm, to the point of being invisible.

But founder conviction cannot be generated with AI horsepower. Conviction happens when the founder has resolved their own belief, knowing exactly what is to be done.

A few other things:

>Go direct with your stakeholders and your customers. There is no other way. You don't want your ideas to get diluted or mutated.
>Compress the idea until even a stranger can repeat it. If the whole idea can be distilled in the simplest language and easily repeated, it spreads like anything.
>Templates can become traps. They remove the texture that makes any company distinct. They pull every founder towards the same jargon, and the same, homogenized shape.

"Go direct" is also founder intensity. Conviction and intensity overshadow everything else about the founder.

A good watch:

This is why I am optimistic about the future.

Progress enables us to dream bigger and push the boundaries.

I really like Pieter's take. You can use the tools to speedrun work for an easy, duplicated product, or use them to build products you couldn't build before.

It's now possible to ship the impossible.

Naval’s no-calendar approach might be overkill, but it hits different when you’re building.

Rigid schedules quietly drain the energy and serendipity.

I’ve learned to protect unstructured time as fiercely as any meeting. That’s often where the best ideas show up.

Follow the energy, and guard your empty space.

Now that @JensenHuang is finally on X, this story needs to be told again.

In 1996, Huang flew to Tokyo to tell his biggest customer three things, each more humiliating than the last.

-The tech we promised you does not work.
-Cancel the contract so that we stop wasting what is left of your money.
-We still need the final $5M or the company will disappear in thirty days.

The customer he was talking to was Shoichiro Irimajiri of Sega.

Irimajiri took a few days, and agreed. The $5M was converted into equity, which bought Nvidia six more months.

Huang didn’t even waste a minute. The chip was shipped in about 12 months, and sold a million units in four months.

Nvidia is now worth ~five trillion dollars. If Sega hadn't sold their equity, it would be worth one trillion dollars.

Most companies lose the feeling of scarcity the moment the bank account gets comfortable.

The trick is to treat urgency as a permanent operating constraint.

Big fan of the Afore team. Congratulations on 10 years.

A great place to start if you're exploring something new: afore.vc/residence

The products that last are built around selling a belief about how the world should work, and that starts with a compelling story.

A story that injects itself in the brain.

In 1900, fewer than three thousand cars existed on all of the French roads. André and Édouard Michelin ran a tire company in a country that had no reason to go out and buy tires.

They printed a small red book filled with maps, mechanic listings, fuel stops and travel recommendations, and gave it away for free.

The first edition didn’t even mention any tire features.

The guiding principle was that a tire company prospers when people travel more. They also built thousands of road signs across France, at their own expense, and branded them. Better signage produced longer trips. The founders even wrote newspaper columns about scenic drives and opened a free trip-planning office in Paris.

They built a story about exploring, finding a reason to own a car.

Michelin was already on everyone’s minds when cars arrived in large numbers. They had seen the name for years in the guides, the road signs, and the maps. The little red book evolved into the Michelin star, the prestigious restaurant-rating publication, renowned all over the world.

With the right story the product sells itself.

The market remembers the story before it remembers the specs.

The belief comes first, the trust builds around it, and the product becomes the proof once people already want to believe it.

Write the story that makes a stranger feel the problem before you write the one that describes your solution, and the rest of the pitch takes care of itself.

More in the article below:

I used to think ambition was the thing holding founders back.

I was wrong.

The thing holding most founders back is the ceiling they have quietly built for themselves, one unexplored skill, one avoided decision, one unchallenged assumption at a time.

Phil Knight built Nike from the trunk of his car, importing Japanese running shoes and selling them at track meets. Before any of that, he flew to Japan in 1962, visited the factory, walked away with a distribution deal and a conviction. When the opportunity arrived to build something real, he had the judgment to move fast and the experience to not break the thing he was building. Nike became a fifty-billion-dollar company.

Experience is like that. It rarely looks like progress while it is happening.

The founders I respect most are the ones who have developed what I can only describe as a high resolution map of their domain. They have been in the room when the deal fell apart. They have managed the person who was brilliant but wrong for the culture. They have launched the feature that users did not want and had to rebuild the roadmap from scratch. Each of those moments deposits something into a mental model that cannot be replicated by reading a book or listening to a podcast. The model compounds and becomes the edge.

Jeff Bezos ran the Amazon warehouse operations himself in the early years. He wanted to understand the constraint at a granular level before he hired someone to solve it. He was a founder building the judgment to see exactly where the business breaks. The pattern recognition he built in those years informed every supply chain and logistics decision Amazon made for the next two decades. Amazon became the most efficient distribution network in human history.

Most founders want to skip to the part where they are running strategy. They want to be in the room where the big decisions get made. The hard truth is that the quality of your big decisions is determined entirely by the quality of the experience you built before you were in that room.

Your outcome has a ceiling. You built it, decision by decision, every time you chose the comfortable read over the difficult room, every time you delegated the hard problem instead of sitting with it long enough to understand it yourself.

The ceiling is movable.

You can raise it.

- You raise it by putting yourself in situations that stretch your operating range.

- You raise it by hiring people who are genuinely better than you in specific domains and paying close attention while they work.

- By reading with the specific intent of challenging what you already believe.

- By asking the honest question every quarter: where is the business waiting on me to become someone different?

The founder is the ceiling. Change the founder, and you change the outcome.