We use emerging technology to make business less complicated.
yellow3 lab researches, experiments with and builds platforms that remove complexity, lower barriers and help businesses operate better.
Technology should increase what people can accomplish. Humans remain responsible for the outcome.
Research first. Build when it helps.
We do not begin with a product category. We begin with a problem worth understanding, establish what is true, test what technology can change and build only when the outcome is useful enough to repeat.
Understand before we build.
Markets, regulation, technology and evidence come first. Research is how we separate signal from noise.
Test what technology can actually change.
We prototype, connect technologies and challenge ideas. Some experiments become products. Many should not.
Turn useful outcomes into platforms.
When a problem deserves a repeatable solution, we build something businesses can operate without us.
Three problems. Three different outcomes.
Our products are different because the problems are different. The common thread is removing work, uncertainty or barriers that should not be there.
From DPP uncertainty to a decision a buyer can defend.
The DPP Buyer Platform structures the buyer's work from strategy and readiness through evidence-based provider selection and procurement. yellow3 works for the buyer. The selected provider operates the passports.
Buyers do not need another supplier list. They need a governed way to turn regulation, internal facts, provider claims and implementation evidence into a programme they can explain and defend.
Establish scope, ownership, systems, data and what the organisation can already prove.
Turn the buyer's situation into explicit requirements before comparing providers.
Separate demonstrated evidence, provider statements and what remains not established.
Carry unresolved conditions and decision evidence into the buying process.
Where AI demand is actually flowing.
Model Adoption is yellow3's live weekly instrument for routed AI usage. It measures where demand is flowing by where the models were built, tracks the most-routed models and grades yellow3's own weekly call in public.
Launch announcements and benchmark headlines tell you what vendors say. Model Adoption follows observed routed usage over time, so businesses, journalists and researchers can see how the market is actually moving.
Asia, the US, Europe and other origins measured as model origin, not user location.
A ranked weekly view of which models are actually receiving traffic.
Movement, streaks and changes preserved over time instead of disappearing into a news cycle.
This week's call and last week's call, graded in public.
DPP supplier evidence
190 organisations mapped in an independent public register, with a machine-readable CSV and JSON dataset generated from the same evidence base.
Explore DPP suppliers →EU AI Act record
A living instrument for what changed, what applies and what evidence sits behind the obligation status.
Explore EU AI Act →Research that can be checked
Sources, methodology and dated records remain visible so conclusions can be inspected rather than merely trusted.
Explore yellow3 research →Machines can do more of the work. Humans should retain responsibility.
AI will automate work and change organisations. We should not pretend otherwise. The opportunity is to give people more leverage without giving away responsibility for the outcome.
People define what matters.
Objectives, constraints, priorities and material judgments remain explicit human responsibilities.
Systems do more execution.
Technology should remove repetitive coordination, complexity and unnecessary operating work where it can do so reliably.
Evidence stays inspectable.
Useful automation should make decisions easier to understand, not harder to question.
Technology should lower barriers, not create new ones.
Capabilities that once required specialist teams, consultants and significant capital should become available to smaller businesses too.
yellow3 builds from the perspective of people who actually have to start companies, run operations and make decisions with limited time and resources. Useful technology is the goal, not impressive technology.
Remove friction without hiding the decisions that still matter.
Product transparency, circularity and better information are practical places digital infrastructure can help.
European rules and standards should shape systems as they evolve, not be bolted on after launch.
Technology, markets and regulation move. Products should be designed to move with them.
Language models have already changed what software can do. They may also be one stage in a much longer transition. That uncertainty is a reason to research and experiment, not a reason to pretend certainty.
Some of our experiments will never ship. That is useful too. Learning by building often creates knowledge that improves a completely different product.
Evidence, analysis and what changed.
We publish when there is something worth checking, measuring or understanding, especially across AI, Digital Product Passports, European technology and regulation.
Research what matters. Build what survives the evidence.
Explore the platforms we are building now, or bring us a problem worth understanding.