I didn't build Agora because multi-AI within the same conversation was cool. I built it because of something I kept noticing when I used AI for my own work.

No matter how "best" the model was, its answer was never complete. So I started copying the same question across different AIs, and something I didn't expect happened. Because they each think differently, I'd get better answers when the AIs reviewed each other's responses. They would reflect on their own answers, and rather than always agree with one another the way they sometimes do with humans, they would actually rethink, refine, or defend their perspectives. That collaboration led to a kind of answer no single AI could produce on its own.

That's the premise Agora is built around. Not the multi-AI hype. Not "more is better." It's the opposite, actually — you almost never need the most powerful model. You need the right shape of help for the question in front of you. I spent months refining how to make the models reflect on and build from each other's answers, so anyone can experience this synthesis in one click, instead of wasting tokens going back and forth across tabs.

So here's where I stand: we don't always need the best model. We need to work smarter, not harder. A small model that's right for the task is better than a frontier model used out of habit. Synthesis across a few efficient models is often better than one giant one running alone. Knowing when you actually need the heaviest tool is a skill — and it's one this industry has every incentive not to teach you.

What "more" actually costs

Right now, all the major AI players are chasing models that pass higher benchmarks, demanding more and more and seemingly endless compute. Some even aim for AGI (Artificial General Intelligence) — a concept so abstract that nobody can actually define it. And speaking as someone with a psychology PhD: we can't even define human intelligence. Historically, how intelligence has been defined hasn't been the most glorious story either — it has always carried its own agenda, particularly around race and supremacy.

And now we are chasing this abstract AGI.

But here's what's real, and what we know for sure. The energy draw is real. The water is real. The labour that goes into making these models safe and usable, much of it underpaid, much of it traumatic, is real. The communities living next to data centres they didn't vote for are real. If you want to understand the scale of this, not in vague terms but in concrete, reported detail, read Karen Hao. Her book Empire of AI, alongside her years of reporting on the AI industry's environmental footprint, its labour practices, and the way data centres are being pushed through communities against their will, is some of the most important work being done on this subject. I won't try to summarize it here. Go read her.

The pattern across all of these is the same. The compute is concentrated. The capital is concentrated. The decision-making is concentrated. The upside flows to a small number of companies and the people who own them. The downside — the water draw, the energy bills, the displaced workers, the disrupted communities, the scraped creative labour — flows outward to everyone else.

Every time the industry bypasses something — environmental review, labour protections, copyright, public consent, the wishes of the community next door — the justification is the same: trust us, the end result will be worth it. The benefits will reach everyone.

I want people to actually sit with the question that promise depends on: will it? Or will the benefits mostly accrue to the small group already building, owning, and profiting from this technology, while the costs get distributed to everyone else?

We can doesn't mean we should

There's one more thing I want to name, because it used to be obvious and somehow stopped being so.

The fact that we can build something does not mean we should. This used to be a normal sentence. Engineers said it. Scientists said it. It was understood that capability is not its own justification. Somewhere in the last few years, in this industry specifically, "we can" started being treated as a complete argument. It isn't. It never was.

Some of the largest companies in this space have stated their goal is AGI. I want to ask, plainly: for what? Whose future is that? Who voted on it? I don't think wanting to question this makes anyone anti-technology. I think it makes them a person who noticed that "we are building something that will change everything" is not, by itself, an argument that it should be built.

The question of should is exactly the kind of question that can't be answered from inside the industry. It needs the people who weren't in the room.

How I think it should move forward

When AI research happens primarily inside AI companies, shaped by those companies' visions of what the future should look like, you don't get progress. You get a feedback loop. A small group decides what AI should be able to do. They build toward those goals. They measure themselves against them. They declare the results progress — and use that progress to justify building more. At no point does anyone outside that loop get asked whether the goals were the right ones in the first place.

The questions the AI industry is trying to answer aren't new.

What is intelligence? How do people actually learn and decide? What does it mean to understand something? How do humans actually reason? What does it mean to know something?

These are some of the oldest questions humans have, and entire fields — psychology, linguistics, philosophy, neuroscience, sociology, and more — have spent generations, sometimes centuries, working on them carefully. Real progress on questions this big has almost always come from collaboration across these fields, not from any one of them alone.

And yet right now, we're blindly following a narrow path — a small group of AI researchers and the companies they work for — and hoping it turns out to be the right one. It would take a miracle for this to be the right path, let alone the most efficient one. And we are betting an enormous amount on that miracle.

On building an AI platform while saying this

I know how this might read. Why do I still build an AI platform if I'm opposed to the industry's practices?

I don't deny AI can be beneficial to all of us. If I believed AI shouldn't exist, I wouldn't have built it. But I also think there has to be a more responsible way of developing and using AI — one that actually benefits humanity as a whole — and right now we are not moving in that direction.

So I try to build the version I want to exist. Agora defaults to smaller, more efficient models because most questions don't need a frontier one. And for the questions worth thinking carefully about, Agora's panel mode is built to be a thinking partner, not a black box. You see what each AI said, where they agree, where they disagree, and how the synthesis was actually arrived at. You're not handed a verdict to trust. You're shown the reasoning behind it, so you can think with it, not just trust it. Pricing is meant to be accessible and transparent: you know what you're getting, what it costs, and you're never pushed toward a tier you don't need. The premise that you don't always need the most powerful model isn't just a marketing line. It's a commitment to not selling people more than they need, and to using only what the question actually calls for.

I know this isn't what a "Why Agora" page usually looks like.

Saying all of this out loud — including the parts that don't flatter the industry I'm part of — is the only version of this I can live with.

What I actually believe

I'm one person building one tool. Agora doesn't fix any of this on its own. But a lot of people reading this probably already feel what I'm describing — the sense that something is moving very fast in a direction that isn't ideal. And a lot of us have quietly concluded that there's nothing to be done about it.

I don't accept that. If most of us want technology that serves people instead of replacing them, that distributes power instead of concentrating it, that respects the land and the water and the labour it depends on, that's built in conversation with every field that has something to contribute — we are not the underdogs in that fight. We are the overwhelming majority. The only reason it feels otherwise is that a very small group has spent a lot of money making sure it does.

I want a future where AI helps people think, not one where it thinks instead of them. I want a future where everyone's kids inherit tools that expanded human possibility rather than narrowed it. I want a future where the question should we build this? is asked as seriously as can we build this? — and where the people answering aren't only the ones who'll profit either way.

Before you close this page, I want you to sit with a few questions:

• Is this a technology we need to build at any cost?
• Do we want a technology built on exploiting people and resources, or one that lets everyone flourish?
• Are you comfortable with this much power concentrated in so few hands and voices?
• What kind of future do you want to live in — and to leave behind?

I said all of this not to convince you to use Agora. I said it because I want more people to understand what's happening behind the scenes of this glorious-sounding development. You don't have to take my word for any of it. In fact, I'd urge you to look into it yourself, and to keep practicing the skill of distinguishing fact from opinion, because I think that's one of the most important things any of us can do right now, in any conversation, not just this one. Whether you end up agreeing or disagreeing with me doesn't really matter. What matters is that we keep having this conversation, and that we keep it grounded in evidence. That's how we make sure we're the ones deciding what this technology becomes — not just the ones living with what it became.

— The Agora founder