Tech & AI

Researchers have created the first formal model to simultaneously manage carbon emissions, energy costs, and power grid stability across distributed cloud and edge computing systems. The work matters because data centers and edge devices now consume massive electricity—and companies face mounting pressure from regulators and customers to cut emissions while maintaining service quality.EN

Kraft­massatillverkning utan tryckkärl blir möjlig genom att ersätta vatten med glyce­rol — en förändring som kan minska investeringskostnaderna drastiskt för pappersbruken. KTH:s forskargrupp demonstrerar att processen fungerar vid atmosfäriskt tryck, men glyce­rols högre viskositet försämrar impregnering­en av trächipsen. Genom systematiska experiment på både löv- och nålträ identifierade forskarna tre kritiska parametrar: mindre chipstorlek och högre temperatur accelererar lignin­upplösningen, medan minskad kemikaliemängd har motsatt effekt. Resultaten visar att temperatur­ökningar är särskilt effektiva för att öka ligninfrisättningen. För leverantörer av pappersteknik öppnas en marknadsmöjlighet i form av lågkostnadsudgåvor för mindre bruks. Regelefterlevnaden förenklas utan högt tryck. Kommersialisering kräver dock vidare optimering av glyce­rolåtervinningen för att processen blir ekonomiskt konkurrenskraftig i full skala.

# MiLLy: Undervisningsagenter integrerade i uppgifter KTH har utvecklat MiLLy, ett system som bäddar in pedagogiska agenter direkt i studentuppgifter för att ge realtidsstöd under problemlösning. Systemet kombinerar instruktionsdesign med agentbaserad feedback, vilket minskar behovet av manuell rättning samtidigt som det ökar lärsutbytet. Forskarna testade MiLLy i praktiska kurssammanhang och visade hur inbäddade agenter guidar studenter genom uppgiftsstegen utan att avslöja fullständiga lösningar. Detta reducerar administrativa kostnader för undervisningsinstitutioner och möjliggör skalbar utbildning. För teknikledare och AI-produktchefer är detta relevant för utveckling av adaptiv learningtech och personaliserade lärplattformar. KTH:s arbete demonstrerar en viabel väg att automatisera handledning genom agenter, vilket förbättrar marginalerna på digital utbildning och skapar konkurrensfördelar i edtech-marknaden.

Utan tillgängligt abstract kan jag inte skriva en faktagrundad sammanfattning enligt de krav som ställts. Jag skulle behöva antingen: - Abstract eller fulltext från artikeln - Konkreta resultat, siffror eller fynd från studien - Metod- och deltagaruppgifter För att uppfylla kraven om att "HALLUCINERA INTE" och att "varje konkret påstående måste finnas i abstract eller metadata" kan jag endast konstatera: **Vad jag vet:** KTH Royal Institute of Technology har publicerat en studie om karriärprogressioner inom teknisk- och datalogiutbildning (författare: Camilla Björn, juli 2026). Studien är inte tillgänglig som öppen källkod. **Vad jag saknar:** Forskningsdesign, resultat, slutsatser och affärsimplikationer. Kan du tillhandahålla abstract, sammanfattning eller nyckelresultat från artikeln?

Versionskontroll är infrastruktur i mjukvaruutveckling, men utvecklare följer inte alltid bästa praxis för commit-meddelanden, branching och dokumentation. KTH:s forskning undersöker hur nudging—subtila designgrepp—kan förbättra beteendet i versionskontrollsystem utan att påtvinga hårda regler. Studien fokuserar på att identifiera vilka interventioner som verkligen påverkar utvecklarnas arbetsflöde positivt. Genom att designa användarupplevelsen kring versionshantering kan organisationer höja kodkvaliteten och minska onödiga fel vid integration och deployment. För teknikledare är detta relevant eftersom svag versionskontrollhygien ökar kostnader för kodgranskning, debugging och onboarding. Nudging erbjuder en låg-friktion väg att standardisera arbetssätt utan att implementera betungande tvångsregler. Detta påverkar både utvecklarproduktivitet och långsiktig underhållsbarhet i CI/CD-pipelines.

Researchers have identified the competencies teachers need to effectively guide students through digital learning environments. The findings could reshape teacher training programs and help schools close gaps in how educators leverage technology—a growing challenge as institutions invest heavily in digital tools.EN

Grafteori-resultatet etablerar optimala täthetsgränser för att garantera vertex-disjunkta trianglar i tripartita grafer — en fundamental fråga inom kombinatorisk optimering med tillämpningar inom nätverksdesign och distribuerade algoritmer. Forskarna härledde exakta vilkor baserade på kanttäthet för olika konfigurationer av grafen. För specifika parameterkombinationer presenteras både nödvändiga och tillräckliga täthetskriterier för existensen av k vertex-disjunkta trianglar samt för triangel-faktorer. Studien utgår från Zhengzhou University och Umeå University. Resultaten releverar för utvecklare av nätverksalgoritmer och systemarkitekter som optimerar resurstilldelning i distribuerade system, där triangelstrukturer modellerar tredimensionella beroenden. Fynden möjliggör effektivare verifikation av nätverksegenskaper utan uttömmande sökning, vilket sparar beräkningskraft vid validering av storskaliga system.

Researchers have developed a method for processing constantly-changing network data—like transaction flows or sensor readings—while keeping multiple search queries running simultaneously. The technique could help companies analyze massive, fast-moving datasets more efficiently, potentially reducing computational costs for financial systems, logistics networks, and real-time analytics platforms.EN

A new study shows that immersive VR lets doctors and nurses spot design flaws during building reviews that traditional 2D drawings miss. The finding matters because poorly designed healthcare facilities can reduce efficiency and compromise safety for decades—and VR could prevent costly mistakes before construction begins.EN

A new Swedish study tracking antibiotic resistance patterns across Nordic hospitals from 2010 to 2018 found persistent rates of dangerous drug-resistant bacteria despite similar consumption levels. The findings highlight a critical gap: while resistance rates stabilized, hospitals adopted wildly different treatment protocols, raising questions about care quality and infection control costs.EN

A new thesis reveals that open science initiatives—designed to rebuild public trust and tackle replication crises—often fail because institutions lack the infrastructure to sustain them. The finding matters to policymakers and research organizations betting billions on open data and transparency mandates without ensuring the systems exist to support them.EN

Researchers have published EC5, a process framework that addresses a critical gap: AI systems generate plausible-sounding but dangerously flawed analysis when organizational processes aren't designed to catch them. The framework, which extends quality-control methods from manufacturing, gives companies concrete steps to verify AI outputs before they drive strategy or operations.EN

A new study reveals that predictive analytics—touted as a fix for managing complex power grids—stalls when utilities and regulators don't share a common vision of its value. The research shows adoption hinges less on technology prowess than on aligning stakeholder expectations, with major implications for energy companies racing to modernize aging infrastructure.EN

A new mathematical framework has identified a minimal set of two operators that can generate every standard elementary function through finite composition. The discovery, verified across 60 physical systems with extreme precision, could streamline how engineers and scientists build computational models—potentially reducing complexity in everything from physics simulations to machine learning algorithms.EN

Researchers have developed neural rendering technology that generates photorealistic simulations of real-world driving scenes, allowing autonomous vehicle makers to test edge cases without physical risk. The technique cuts testing costs and accelerates development cycles by enabling companies to modify scenarios digitally rather than conducting expensive real-world trials.EN

Researchers have created the first systematic method to validate whether computer simulations of car crashes are realistic enough to prove autonomous vehicles are safe. The work addresses a critical gap: carmakers currently test self-driving systems in virtual environments, but nobody has reliable tools to confirm those simulations match real-world conditions—a problem that affects insurance, regulation, and deployment timelines.EN

Researchers have significantly improved the success rate of a quantum algorithm that cracks certain types of encryption, pushing it toward practical viability. The finding accelerates debate over when quantum computers will pose a realistic threat to today's security infrastructure—a concern already shaping investment and regulatory planning in cryptography and cybersecurity.EN

Researchers have solved a 30-year physics puzzle to build a quantum key distribution system that remains secure even if eavesdroppers have perfect equipment and unlimited computing power. The breakthrough uses individual photons to generate unbreakable encryption keys, potentially transforming how governments and enterprises protect their most sensitive data.EN

Researchers have developed a machine-learning approach that reconstructs images transmitted across multiple unreliable network hops with significantly better visual quality than existing methods. The breakthrough could improve wireless imaging in remote sensing, autonomous vehicles, and secure communications where traditional compression fails under poor signal conditions.EN

Researchers have developed a machine learning approach that measures how accurately 3D cameras align spatial data—a critical bottleneck for autonomous vehicles, drones, and robotics. The method replaces binary pass/fail checks with continuous error scoring, enabling systems to fix misalignments before they cascade into navigation failures and costly re-mapping cycles.EN

A comprehensive review of 317 papers reveals the field lacks consensus on what 'care' means in technology—creating confusion for companies and designers building care-focused products. The analysis identifies six distinct frameworks that often contradict each other, suggesting the industry needs clearer standards before scaling care-centered applications.EN

Researchers have shown that machine learning can assess user satisfaction with voice assistants by analyzing how people actually speak — not what they say. The finding suggests companies could soon monitor customer frustration in real-time and fix problems automatically, shifting voice interface design from reactive surveys to predictive intervention.EN

As artificial intelligence systems begin autonomously designing, running, and learning from biological experiments, researchers are grappling with a fundamental problem: who's responsible when outcomes emerge from humans, algorithms, and living organisms working together? The stakes span product liability, regulatory compliance, and whether today's governance frameworks can even handle this new kind of innovation.EN

A new panel framework identifies how designers can create sustainable body-transformation technologies—from smart textiles to VR experiences—without forcing permanent or predictable changes on users. As these technologies move into everyday consumer products, the research addresses a critical gap: how to balance innovation with ethical responsibility and user autonomy.EN

Researchers have identified three core values—universality, objectivity, and efficiency—embedded in the foundational theories behind widely-used visualization tools. The discovery matters because these embedded assumptions may be filtering out important perspectives, potentially causing organizations to miss critical insights and making their data tools less inclusive by design.EN