Simulate realistic people,
living inside
a computer.

Find the risk before it finds you.
Verbll is a human simulation platform: realistic people, each with their own personality, emotions and memory. Their minds run on algorithms, not a language model, so every reaction comes with a reason.

  • 10,000unique personalities, built on official labour statistics
  • 118emotional states, blended from core emotions
  • 280personality traits in the emotion engine
  • 100simulated people testing a web app in parallel

The film

Watch it in 2:28

What Verbll is, why it is different, and why it matters.

What is Verbll about?

People you can test things on

Verbll is a human simulation platform. Each simulated person has their own personality, emotions, memory and goals. You place them in a real place or a web app and read how every one reacts.

Why is it different?

Minds built from algorithms

A common approach is to prompt a language model to play a person. Verbll's people run on algorithms for memory, emotion and decisions. The core has no language model. We also build the mind differently, as the next sections show.

Why Verbll?

Find the problem before people do

Change one thing, run again, and read a reason for each reaction. A mistake in a simulation costs nothing. A mistake in the real world costs people and money.

A milestone for us

The day the Superbrain's memory held together.

Our first memory test: one photograph, stored in a single pass. The first attempt came back as a coarse, blocky scene. After several rounds of tuning the memory's plasticity and recall, it came back sharp. We are proud of this one.

  1. Stored in one pass.The Superbrain took the photograph in once.
  2. First attempt.The whole scene came back from memory. The detail was coarse.
  3. Plasticity tuning.Several changes to how the memory strengthens and recalls, among them the resolution of each memory tile.
  4. Second attempt.The photograph came back sharp.
A person

Usually needs repeated exposure and rehearsal to hold a detailed scene. What they recall is a reconstruction, never a copy.

The Superbrain

Learns a scene the first time it sees it, and recalls it sharp. Its memory is built to get better with use.

Original photograph of a pink palace facade
Original
The image recalled from memoryWaiting for the result
Recalled from memory

The Superbrain

One mind per person, built from algorithms.

Every simulated person runs on a Superbrain: memory, attention, emotion and decisions, all as algorithms. This is what sets it apart.

  • Learns a scene the first time it sees it.It stored an 8 MB photograph in a single pass, in about a tenth of a second on our test machine. A person needs repetition for that.
  • Recalls in milliseconds.Recalling the stored photograph took a few milliseconds of pipeline time in our test run.
  • Memory that strengthens with use.What a person lives through is reinforced. They never start from zero, and every learned fact keeps its source.
  • No language model inside.Algorithms are fast and predictable, and there is no model bill for every person you add.
  • Reaches out only when it must.At a gap in what it knows, it searches the web or asks an outside AI model, then remembers the answer.

Inspired by neuroscience.

Training, done differently

Not datasets and model training. The methods people use to sharpen their own minds.

The usual route trains a model on data. We study what people already do to improve their own minds, try it on ourselves, and build what works into the Superbrain.

The usual route

Gather datasets. Clean them. Train models. Try to recreate the brain.

The Verbll route

Learn from what people already do to improve their own minds. Try it on ourselves. Build what works into the algorithm.

  1. 1
    Study how people train their minds

    Tutorials, memory tricks, brain teasers, illusions and mind-reading puzzles. Short videos like these draw millions of views each.

  2. 2
    Try it on ourselves

    We run each method on ourselves first, and keep the ones that work.

  3. 3
    Build it into the algorithm flow

    Each method that works becomes a step in the Superbrain's algorithm flow.

A page of short videos about brain teasers, memory tricks and optical illusions
Watch

The simulated world

Conflicts, weather, good news, bad news. The emotion engine reacts to all of it.

A situation can be an argument, a storm, a compliment or a lost job. Verbll turns it into a few numbers, and the Superbrain's emotion engine responds. Watch a situation push the emotions up and down.

Emotion does not sit beside thought. It steers it. The emotion engine feeds the Superbrain's attention, memory, risk appetite and voice, so every answer a simulated person generates is coloured by what they feel at that moment.

2,898unique situations in our library
26categories
29appraisal parameters
14variants of each
2,627distinct parameter sets

Emotion engine
Blended states
How it changes the thinking

Counts from the Verbll situation library.

Simulated person · A00001Taichi Oshima54 · Food processing worker · Fukuoka, Japan
Personality
Openness
Conscientiousness
Extraversion
Agreeableness
Neuroticism
Honesty
Emotional tendencies
Anxiety
Sadness
Anger
Shame
Reactivity
Calmness
Wants
To be the one people rely on
Fears
Being left out of the group
Contradiction
Wants to help everyone but resents being relied on
Blind spot
Remembers favours given more than favours received
Copes by
Reframing, with one close friend
Values
Learning, humility, care, wealth, duty
Formative memory

A friend group drifted apart after a quarrel, about 22 years ago. Lesson kept: quarrels need repair fast.

10,000 people

Ten thousand people. No two alike.

Each one is a full profile, grounded in real-world statistics we researched. A person is not an average. They have contradictions, blind spots and formative memories, so they react the way that particular person would.

10,000people
116attributes each
33countries
514occupations
80native languages
366regions

What we consider for every person

A reaction depends on who a person is, what happened to them and how they cope, not on one average profile. The card on the left is one of the 10,000.

Physics in the simulated world

Solid walls. Personal space. A goal for every person.

The world around each simulated person obeys physical boundaries. Walls, desks and pillars are solid, people keep their distance, and every person is heading somewhere. Physics is what turns a crowd of dots into a crowd that behaves.

Solid boundaries

Built into the engine

  • Collision volumes.Every wall, desk, pillar and person has a solid shape. Nothing passes through, and every contact is detected.
  • Sensors.Each person casts rays ahead, reads the free space and steers around what is in the way.
  • Personal space.People keep their distance, so crowds queue, spread and squeeze the way real crowds do.
  • Goals and rerouting.Every person has a destination. Close a way, drop a barrier, and the crowd finds another route.

Physics boundaries on the roadmap

Crowd pressure and crush forcesFire and smoke spreadDoors, stairs and liftsWind, rain and floodingLine of sight and noiseVehicles and trafficStructural limitsHeat, light and air quality

Maps to 3D

Any real place becomes a 3D town you can test.

Choose any real place and a radius. Verbll takes the real streets, blocks and buildings, builds them in 3D and fills them with people. You test the place before anything is built, closed or opened.

Test the real placeYour actual junction, venue or district, not a generic grid.
Change it before you build itClose a road, add a building or move a service, then run the same crowd again.
Find bottlenecks earlySee where queues form, where people get stuck and where they avoid.
Show it to anyoneA 3D town makes your case to a board, a council or a customer in seconds.
A real map

Computer use

It browses the real web, like a person does.

Simulated people use a real browser: they look at the page, move the mouse, click and type at human pace. Use it to test web apps or to automate your own repeated work.

shop.example/checkoutTeach mode: recordingRepeating on its ownSuperbrain on: reacts like a personScheduled: every Monday, 09:00
Checkout
NameAda LovelaceGrace HopperSam
Emailada@mail.comgrace@mail.com
Pay now
Where is the total? I am not sure I trust this page.
Next run: Monday 09:00
1 Teach once2 It repeats3 It reacts4 On a schedule
  • A real browser, with eyes and hands.It sees pixels and sends mouse and keyboard moves. Curved mouse paths, human typing, no page scripts.
  • Teach it on your own computer.Show it a task once. It repeats the task for you.
  • It reacts like a person.With the Superbrain on, it hesitates, gets confused and gives honest views, because it has a mind behind the mouse.
  • Runs on a schedule.Set a task to run by itself, every day or every week.

Use cases

Things you could not test before.

Select a card to read more.

Simulate citizens of every income, age and level of trust in institutions. Watch who complies, who objects, who is left out, and why, before the policy reaches a single real person.

You read: reactions by group, with a reason for each.

The AI revolution is changing the job market. Simulate a student on each path and explore how skills, stress and choices play out as work changes around them.

You read: how each path could unfold, with the stress and choices along the way.

Build the venue and its streets in 3D. Place the barriers, the gates and the police posts. Send in the crowd and watch entry, exit and waiting time.

You read: queues, bottlenecks and the safest layout.

Teach an agent a repeated task on your own computer, once. It repeats the task in a real browser and runs on a schedule you set.

You get: hours back, and a log of every run.

Many different people try your app. Some stay. Some leave. Each one tells you what confused them and whether they would use it.

You read: where they got stuck, and the honest view of each.

Rebuild a situation with the place and the people. Change one fact and run it again. Useful for lawyers and investigators working out how people could have reacted.

You get: the scene replayed with one fact changed, and how each person responds.

What is your use case? Join now →

Why Verbll

“We failed many startups, and every time the problem was the same. We had never checked whether anyone wanted the idea. So we tried interviews, but getting to the right people took so long that the time was gone before the answers came. Then we moved to AI agents and let different models brainstorm for us. It still did not work, because those agents say the same things over and over. They were trained once and then frozen, so there is nothing new inside them, and web search does not fix that. We needed something that could use the latest information and treat a problem like a real person would. So we asked why not place people that behave like real humans inside a computer simulation, and that is how Verbll was born. It runs a simulation as close to the real world as possible, so nobody gets hurt and nobody pays for a bad loan or a bad launch. Most simulations today use frozen AI agents. Run one again and you get the same result in slightly new words. Nothing changes and nothing grows. That is what we are solving. Every person in Verbll has their own brain, and it keeps learning as you keep running.”

Sumedh PatilFounder, Verbll

Built for people like you

You will recognise yourself.

Verbll was not made for a market. It was made for the person who has a decision to make and no safe place to try it first.

It is 1 a.m. and you are rewriting the pricing page again. You have no time to interview fifty strangers and no budget to be wrong. Verbll gives you fifty strangers before breakfast, and every one of them tells you why they would stay or leave.

You sign off on a plan that thousands of people will live inside, and the only test you get is the day it opens. Verbll walks the crowd through it first, so the problem shows up on your screen and not in tomorrow's headlines.

You have a theory about why people do what they do, and a study that takes eighteen months. Change one fact in Verbll, run it again before lunch, and see which people break first.

You are twenty-two, everyone has advice, and nobody knows what work will look like in five years. Walk each path in a safe world first and feel where it gets hard, before you commit a year of your life.

This is me. Join the waitlist →

Launching 2027

Join the waitlist.

Verbll launches in 2027. Join now to be first in line, and tell us what you would simulate.