Gemini 3.1 Flash-Lite for School Buyers
March 3, 2026
Small, fast AI models rarely get the same attention as flagship launches, yet they may matter more for schools trying to reduce workload without stretching already tight budgets. Using Gemini 3.1 Flash-Lite as a practical test case, this guide explores which routine teacher tasks a lightweight model can handle well, where it begins to struggle, and how leaders can judge whether paying for a full model is actually necessary. The aim is simple: help schools buy AI more carefully, with clearer expectations and lower risk.
Half-Term CPD: AI Safety Essentials
February 24, 2026
This half-term self-study pack helps educators turn fast-moving AI safety headlines into practical safeguarding action. It brings together four major developments, from youth-facing AI companion risks to prompt injection and wider system-level safety findings, and translates them into a clear CPD sequence that can be completed in under an hour. The article ends with a ready-to-run 20-minute staff scenario, plus debrief prompts for DSLs, senior leaders and classroom staff who want to strengthen school AI safeguarding without needing specialist technical knowledge.
Gemini 3.1 Pro Benchmarks Decoded
February 19, 2026
Benchmark headlines can make a new AI model sound either revolutionary or irrelevant, yet most school leaders and teachers are not given the context needed to judge those claims. This guide explains, in plain English, what Gemini 3.1 Pro’s reported results on ARC-AGI-2 and SWE-Bench actually mean, what those tests reward, and where their limits lie for real school use. It also offers practical ways to evaluate AI tools for teaching, planning, procurement, and governance without confusing lab performance with classroom value.
LGR22 Mathematics: A Five-Tool AI Playbook
February 16, 2026
Mathematics under LGR22 demands visible method, clear reasoning and manageable progression from E to A. This guide shows how Swedish teachers can build that work with five directly useful Automated Education tools: Word Problems, Quiz Generator, Excel Guru, Concept Explainer and Difficulty Adjuster. The emphasis is practical and product-led: better word problems, clearer method explanations, sharper misconception checks, stronger differentiation and faster data-handling routines you can use next week.
Gemini 3 Deep Think in Sixth-Form Science
February 12, 2026
Gemini 3 Deep Think looks promising for sixth-form science, but its real value appears when it is used as a thinking partner rather than an answer machine. This evaluation explores how it handles A-level and IB exam-style questions, as well as EPQ-style research prompts, with a focus on explanation quality, hypothesis building, source trails, and error-checking. The goal is practical: to identify where it saves time, where it misleads, and how teachers can help students use it without weakening independent thought.
Perplexity AI Model Council in the Classroom
February 6, 2026
Perplexity AI Model Council can become more than a novelty in school. Used well, it offers a structured way for pupils to compare how different models answer the same prompt and to discuss evidence, confidence, bias, omissions and persuasive style. This shifts classroom conversation away from “which AI is best?” and towards media literacy, judgement and source checking. In this article, you will find a practical lesson structure, a pupil-friendly comparison scorecard, subject-ready prompt ideas, discussion routines and sensible safeguards for teacher-led use.
Claude Opus 4.6 for half-term planning
February 5, 2026
Claude Opus 4.6 arrives with stronger workflow support, and that matters most when the test is practical rather than promotional. In this article, we put it to work on a demanding school task: building a half-term scheme of work through agent teams, then exporting the result into PowerPoint for real staff use. The focus is not on flashy outputs, but on teacher-ready quality: sequence, lesson coherence, curriculum fit, edit load and presentation readiness. The result is promising, but only in the right conditions.
AI Revision Strategies for Mock Season
January 28, 2026
Mock season often brings a rush of revision resources, but effective preparation needs more than quick AI-generated quizzes. This article outlines a practical workflow that uses AI to analyse past performance, identify knowledge gaps, build interleaved practice, and schedule spaced retrieval in ways pupils can actually sustain. It also shows how teachers can keep revision honest, safe, and evidence-led by checking outputs, protecting data, and preventing answer outsourcing or over-reliance on the tool.
Anthropic’s AI Constitution: School Lessons
January 21, 2026
Anthropic’s expanded 23,000-word AI Constitution offers schools a useful case study in how AI safety frameworks become more detailed over time. The value for school leaders is not in copying a vendor document, but in noticing what gets made explicit: new harm categories, clearer boundary rules, and sharper decisions about when a system should refuse, redirect, or escalate. This article explores what that evolution can teach schools about AI policy, acceptable-use clauses, and safeguarding guidance, and provides a practical way to turn broad vendor language into school-ready rules.
January INSET: Practical AI by Department
January 6, 2026
January is the right moment to move school AI training from broad awareness to practical, low-risk use. This workshop plan sets out a shared 90-minute INSET session that every department can join, using free-tier tools and clear safeguards. It includes parallel pathways for reluctant and confident staff, realistic use cases across teaching, pastoral and admin work, and a simple first-half-term implementation model. The goal is not novelty, but consistent adoption that saves time, improves quality and stays within sensible privacy rules.
LGR22 History with AI Source Criticism Workflows
December 16, 2025
History under LGR22 is about evidence, perspective, causation and justified conclusions, not just recall. This guide shows how Swedish teachers can use Meet the Character as the lead tool in a structured source criticism workflow, with concrete Gustav Vasa and Viking examples. It also shows how Concept Explainer, Quiz Generator and Lesson Planner can turn one strong AI activity into a coherent history sequence that is practical, inspection-ready and easy to repeat.
Claude Opus 4.5 School Briefing
December 12, 2025
Anthropic’s Infinite Chats feature changes how schools may use AI over time, especially for extended research, curriculum planning, and iterative drafting. The opportunity is clear: longer-running conversations can support richer projects and reduce repeated setup work. Yet schools also need stronger habits around evidence, privacy, safeguarding, and staff judgement. This briefing explains what changed, where the gains may be strongest, and how to build safe workflow patterns that do not drift into unsafe data sharing or over-reliance on one ever-growing chat thread.
Festive STEM Projects That Teach AI
December 10, 2025
Festive AI activities can easily become novelty lessons full of prompts, pictures and quick wins, but miss the core ideas pupils need to understand. This article offers three practical, paper-first projects that teach classifier rules, image recognition limits, and A/B testing through seasonal examples. Along the way, pupils explore features, training data, test data, evaluation, bias and false positives in ways that work across primary, secondary and mixed-access classrooms. The result is festive STEM work that feels engaging while still producing clear evidence of learning.
Gemini 3 Flash in Classrooms
December 3, 2025
Google’s move to Gemini 3 Flash as the default model will matter most in the small, repeated tasks that shape a teacher’s day. Faster replies can reduce friction when drafting emails, reformatting materials or generating quick classroom supports. Yet speed is not always the same as quality. In planning, differentiation and resource design, a faster model can produce tidy but thin outputs that need more checking. This guide explores where Gemini 3 Flash is likely to help, where caution is needed, and how schools can test the switch before it affects everyday workflow.
ChatGPT Turns 3: Education Impact Assessment
November 28, 2025
Three years after ChatGPT’s release, schools have enough experience to judge what has genuinely become settled practice and what remains volatile. This impact review offers a practical, evidence-pack approach for leaders: what to measure, what to stop measuring, and which artefacts to collect so decisions are defensible. It maps key shifts in policy, classroom routines and assessment integrity from 2022–2025, then sets out a realistic 2026 outlook without triggering tool sprawl. It finishes with a one-page SLT/governor briefing template for decisions in the next 30 days.