Independent AI product studio
Intelligence, grounded.
We design and build AI products for the physical world: , , and .
Software people can act on. 45 areas moved against the season. Each gets a level and its evidence, and a person checks the top of the list.
Contains modified Copernicus Sentinel data 2026. Demonstration on open data, not a client project.
Climate Intelligence
Know when a site stops behaving the way its risk model expects.
Climate Intelligence keeps a digital twin of each site: satellite, radar, hazard, permit and ground sensor data on one map, updated as new data arrives. When the site drifts from what its risk model expects, the twin raises a level, shows the evidence, and tells the person responsible. Landfill sites come first.
Every new pass lands on the same grid.
Each Sentinel-2 pass is cloud-masked and aligned to the twin, so two dates compare pixel for pixel. Here, the same fields three months apart.
- Sentinel-2 optical10 m, every 5 daysShown here
- Sentinel-1 radarGround motion, mm per yearIn the product
- Hazard mapsFlood, heat, droughtIn the product
- PermitsFootprint, cells, capacityIn the product
- Ground sensorsMethane, leachate, weatherIn the product
Change is measured against the season.
The twin compares each pixel with how the season should move it, not with a fixed threshold. Most of the change here is the rice crop. A small share went the other way, and that is what needs explaining.
- of the land greened with the season
- 59%
- lost vegetation against it
- 1.8%
- areas over 2 ha to explain
- 22
A risk level, with its reasons.
Each area gets a level, the evidence behind it and how sure the twin is. In the product, the level also weighs hazard exposure and the permit. A person decides what happens next.
- High11.6 haModerate vegetation to bare ground or water
- Elevated11.4 haDense to sparse vegetation
- Elevated11.2 haModerate to sparse vegetation
Confidence: Medium. Two clear dates, optical only.
Ebro Delta, Spain. Contains modified Copernicus Sentinel data 2026. Open-data demonstration on farmland, not a client site or a risk assessment.
Tenet
A private memory for the life you share.
Tenet is a private graph of the moments, places and people in your life, built by voice. Mention a dinner, a shop or a promise in passing and it connects them. Ask later in plain words, and share a thread only with the few people closest to you.
Ask Tenet
Pick a question. Tenet answers from your own graph, not from a public ranking.
Fictional people and places. An interactive sketch of the idea, not the product.
Company Brain
A team memory that does the follow-up.
Company Brain reads what the team produces each day: email, call notes, documents and meetings. One graph sits behind the CRM, projects and tasks, the knowledge base and wiki, and the funding pipeline. It prepares a brief before every meeting and flags what needs attention. Agents propose changes. A person approves them.
Read today38 emails4 call notes11 documents2 meetings
- Marta Oliveras confirmed a January pilot on the last call.
- The data agreement is still unsigned.
- Their last email has waited six days for a reply.
Proposed by an agent
Marta confirmed the pilot can start in January once the data agreement is signed.
- PersonMarta Oliveras, Head of Operations
- RelationMarta Oliveras works at Ribera Logistics
- OpportunityRibera pilot
IntroProposal - CitationQuote linked to the call notes
Nothing is written until you approve.
- Deadline in 19 days. Co-funding up to half of eligible costs.
- Eligibility notes drafted from the project file. Not yet sent.
- Send the pilot scope to Ribera: you, Friday.
- Draft the data agreement annex: Júlia, Wednesday.
- Book the January kick-off: Pau, next week.
- Retention changed from 12 to 6 months. The previous version is kept with its source.
Published research, tested in the field.
We contributed to published research on portable methane sensing and machine-learning calibration, tested in the laboratory and at landfill sites.
Bertin, L.; Mentasti, M.; Pittorino, F.; Villa, V.; Zanni, E.; Viscardi, G.; Ponzani, Y.; Massara, A.; Roveri, M.; Dellaca’, R.; Capelli, L. Sensors 2026, 26(13), 4321. Published 7 July 2026.
Andrea Massara is credited for software and resources. The paper reports work within ESCAPE through Integraciones Digitales Gold SL (Indigo). It does not validate Climate Intelligence satellite screening.
Read the paperQuestions we are usually asked first.
What does Adaptivecity do?
Adaptivecity is an independent AI product studio in Barcelona. We design and build AI products for the physical world, from sensing and geospatial analysis through to the interface where a person makes the decision.
What does AI for the physical world mean?
Software that reads real conditions from sensors, satellites and places, and turns them into a judgement someone can act on. Our scope is sensing, spatial context, machine learning and the software people use. It does not cover robot control or autonomous machines.
What is Climate Intelligence?
A digital twin for sites exposed to climate risk, starting with landfills. It combines satellite, radar, hazard, permit and ground sensor data, compares each site with what its risk model expects, and returns a level with its evidence for a person to review. It is in development with Vuesensing.
What is Tenet?
A private, voice-first graph of the moments, places and people in someone’s life. You add to it by speaking, ask it questions in plain words, and share threads only with your inner circle. It is a product exploration, not a released app.
What is Company Brain?
An active memory for a team. It reads email, call notes, documents and meetings, and keeps one graph behind the CRM, projects and tasks, the wiki and the funding pipeline. It prepares meeting briefs and proposes changes that a person approves. It is in development with Vuesensing.
Who do you work with?
Founders and product teams building environmental intelligence, geospatial products, connected hardware and software for field operations, and teams building location-aware consumer products.
How does an engagement start?
Usually with a short paid discovery. We map the user, the decision, the evidence available and a measurable acceptance test, then plan the build against it. A well-scoped build can start directly.
How do I get in touch?
Email andrea@adaptivecity.com with what you are building, who it is for, and what has to work. We are at Tech Barcelona, Pier 1, Plaça de Pau Vila, Barcelona, and work remotely with teams elsewhere.
What should your product understand about the world?
Tell us what you are building, who it is for, and what has to work.