Skip to content
MinisodasEXPERIMENTATION AT ITS CORE
Menu
HomeSolutionsCase StudiesLabsMoreBOOK·a quick scoping call

In this brave new world, make rapid experimentation with AI your moat.

Minisodas turns emerging AI capabilities into measured experiments for real business workflows.

Our hypothesisExplore LabsNEW

TEST

the latest
AI stack

MEASURE

against a
baseline

DEPLOY

only what
performs

Intelligence is abundant

Thanks to the latest AI models and the speed at which they're getting even better.

A new world is being shaped right before our eyes.

Speed is an advantage

Things are now changing within days instead of months.

Keeping the testing → deployment loop short is a way to stay ahead of the curve.

Outcomes over "it works!"

Making it work is the easy part.

Real outcomes are achieved by measuring AI workflows against baselines and beating them.

Humans stay in the loop

Human approval and auditability matters.

Intelligence inside a black-box cannot be harnessed efficiently.

SELECTED CASE STUDIES

Case study carousel placeholder

Explore all Case Studies

OUR HYPOTHESIS

We believe AI will reshape human work around two roles. But before we get to these roles, a fair question:

Will humans even be needed?
Won't AI replace us?

“Imagination is more important than knowledge”
—This em dash was not generated by AI :) Albert Einstein

And AI is not the best at it.

EINSTEIN'S “JUMP”

In a 1952 letter, Einstein drew discovery as a loop.

What we observe cannot lead us straight to a new idea. Someone has to imagine it, that is, they have to make the “jump”. Then logic and experiments test it.

Einstein's discovery loopWhat we observe leads through an imaginative jump to a new idea. Logic produces predictions, and experiments test those predictions against what we observe.A NEW IDEAWHAT IT PREDICTSWHAT WE OBSERVEJUMPLOGICEXPERIMENTS
A simplified version of “Einstein’s 1952 sketch”.

“LLMs CAN’T JUMP”

A position paper by Tom Zahavy, a Research Scientist at

Google DeepMindRead the full paper

This position paper argues that current LLMs can easily find patterns and reason from ideas they are given.

What they still struggle with is forming a good-quality new idea itself, which requires the “jump”.

That's why, we believe smart human jumps are important for
discovering
new ways of harnessing the latest
capabilities of AI.

OURBET

AI-powered work will increasingly
depend on two human roles
working together.

The rapid experimenter

A curious builder who finds new ways to use the latest AI tools and tests them quickly.

The evaluator

A domain expert who checks the work and decides what is good enough to use.

One explores what is possible.The other decides what is valuable.Businesses will need both.

MinisodasLabs is being built to turn this way of working into something businesses can
discover and use.

Minisodas Labs

A place to discover top-notch AI experiments for real business workflows, and the people behind them.

The landscape of AI integration ideas for businesses

What you find on the internet
/ LLMs suggest you to do

What's worth evaluating

What actually delivers
valuable outcomes

This is what makes it to the
 Labs page

How do we ensure this?

By keeping the bar high.

That's it.

  1. Shortlist

    The top experiments are shortlisted

  2. Evaluate

    A human domain expert evaluates them

  3. Publish

    Only the best performing ones make it to the Labs page

One place. Standard format.
Measured outcomes. Real people behind them.

GOING LIVE SOON.

Join the waitlist