Free skill check →Papers and reports from the Square 1 AI team on the systems we build and teach — AI safety, security, LLMs and agents, and applied machine learning.
22 papers6 topics~3 hours of reading
A framework for comparing how languages encode meaning — and what systematic cross-linguistic comparison reveals for language technology and language learning.
Research on systems that automate and augment organisational decision-making: architectures, governance, and the line between decision support and decision delegation.
Read the article3 min readEducation6 July 2026The 4D framework — delegation, description, discernment, diligence — plus a systematic review of ten studies on how teachers build AI fluency while keeping human judgment central to the classroom.
Read the article30 min readSecurity6 July 2026Securing cloud environments with AI: detection across sprawling attack surfaces, misconfiguration hunting, and the operational realities of trusting models with security decisions.
Read the article5 min readLLMs & Agents6 July 2026A tour of modern image recognition — convolutional and transformer-based approaches, benchmark progress, and the practical limits that still separate lab results from deployment.
Read the article3 min readSecurity6 July 2026Where machine learning genuinely strengthens defence — threat detection, anomaly spotting, response automation — and where attackers are using the same tools back.
Read the article7 min readIndustry & Society6 July 2026How a digital bridge between geological models and plant hardware — hosted in a digital twin — enables real-time ore sorting and dynamic filtration, cutting energy use and improving water circularity in mining.
Read the article3 min readHealthcare6 July 2026How machine learning models are pushing diagnosis earlier — the screening domains showing real clinical signal, and the validation hurdles between promising models and patient impact.
Read the article30 min readSecurity6 July 2026Applying AI to threat modeling: generating attack scenarios, prioritising risks, and keeping security analysis current as systems change faster than manual reviews can.
Read the article4 min readLLMs & Agents6 July 2026Evaluating a single model is hard; evaluating a system of interacting agents is harder. Methods and metrics for assessing multi-agent behaviour beyond per-agent benchmarks.
Read the article8 min readAI Safety & Ethics6 July 2026Can tamper-evident ledgers make AI systems auditable? An examination of blockchain-backed provenance, model audit trails, and the accountability gaps they can and cannot close.
Read the article2 min readAI Safety & Ethics6 July 2026AI systems are trained on some cultures more than others. This research examines what culturally responsive design means in practice and why it matters for global deployment.
Read the article8 min readSecurity6 July 2026Zero-trust assumes breach; AI makes the continuous verification it demands tractable. How adaptive models strengthen identity, access, and anomaly decisions in zero-trust architectures.
Read the article2 min readHealthcare6 July 2026Clinicians won't act on predictions they can't interrogate. This paper reviews explainability techniques for medical AI and the evidence on what actually earns clinical trust.
Read the article8 min readIndustry & Society6 July 2026How generative models are reshaping creative work — production pipelines, authorship and rights questions, and where human craft concentrates as generation gets cheap.
Read the article13 min readLLMs & Agents6 July 2026Research on making large language models cheaper to train and serve — quantisation, distillation, and architectural efficiency, and what they trade away for their savings.
Read the article3 min readIndustry & Society6 July 2026A report on how teams are actually structuring AI-assisted work in 2026 — the workflow patterns, tooling choices, and process changes that separate productive adoption from noise.
Read the article22 min readIndustry & Society6 July 2026A study of quantum computing readiness in Sri Lanka: the research capacity, industry appetite, and policy groundwork an emerging economy needs to participate early.
Read the article7 min readLLMs & Agents6 July 2026Two failure axes of deployed LLMs — wasted compute and confident falsehoods — and the techniques that push toward efficient resource use with grounded, verifiable output.
Read the article7 min readAI Safety & Ethics6 July 2026When multiple AI agents interact, new failure modes appear that single-agent safety work doesn't cover. This paper maps the risks and the emerging mitigations.
Read the article10 min readAI Safety & Ethics6 July 2026Why systems that feel empathetic and systems that explain themselves pull in different design directions — and how explainable-AI techniques can close the trust gap between the two.
Read the article4 min readAI Safety & Ethics6 July 2026What AI systems reflect back about the people and data that built them — an essay on bias, projection, and interpretation in human–AI interaction.
Read the article4 min read