CurricuLLM LogoCurricuLLM
SchoolsAI tutorFeaturesPricingTraining HubDevelopers
An age-appropriate AI tutor, and when to switch the noise off
1 August 2026

An age-appropriate AI tutor, and when to switch the noise off

Dan Hart

Dan Hart

CEO, Co-Founder, CurricuLLM

An AI tutor for a six-year-old and an AI tutor for a Year 12 student are not the same product. I spent twenty minutes reading AI news this morning and came away with whiplash. Stocks are surging and collapsing at the same time. Every job is about to disappear except none of them will. The future is either a techno-utopia or a surveillance state, depending on which tab you have open. I work in this field every day and even I find the noise exhausting. So this is a week of trying to hold onto the few things that actually help you decide what an AI tutor should do.

What we mean by an AI tutor is set out on the AI tutor for schools page.


Switch off

A regular reminder, from someone who genuinely believes in this technology. Switch off. Go outside. Talk to a person. Read something printed on paper. AI matters, but it is not the entire world. I took a photo this week at Cowan Creek, zero mobile reception, bliss. I am leaving it here as the argument.


One AI tutor does not fit every year group

Jennifer Weber's brief for the Manhattan Institute proposes no student-facing generative AI in K-2, teacher-directed critique only in grades 3 to 5, mandatory AI literacy plus tightly constrained tool use in 6 to 8, and expanded but governed use in high school.

I like the shape of this more than I like most school AI policies, because it starts from development rather than from panic or from boosterism. Younger children do not need a chatbot. Older students need to learn how to use one without handing over the thinking. The error most systems make is applying a single rule to every year group and then wondering why it satisfies nobody.


Data centres, energy, and jurisdictions that will not sign

Tuesday's Energy and Climate Change Ministerial Council communique commits the Commonwealth to national AI standards setting minimum requirements on energy, water and location for data-centre investment. Large-scale data centres would have to offset their demand by funding new renewable generation in the jurisdiction they sit in, verified through a guarantee of origin scheme, with no cost passed to households. National Cabinet considers the standards in August, National Electricity Rule changes come back in September, and legislation is expected in early 2027. AEMO has data centres reaching around 10 per cent of NEM demand by 2050.

Both Queensland and the Northern Territory opposed every data-centre resolution in it. David Janetzki has said Queensland wants costs, benefits and risks spelled out before signing up to what he called "underdeveloped national proposals to advance other policy objectives", and David Crisafulli has been explicit that renewable requirements would make the state less attractive to investors. The Territory is backing a gas-powered data-centre industry off the Beetaloo Basin, including a 2 GW hyperscale precinct near Darwin.

The whole framework needs every jurisdiction to agree. That is a high bar, and the politics of it will decide the policy more than the modelling will.


Faster is not differentiation

I have been thinking about service this week: the role IT systems play in it, and whether service can genuinely differentiate a business. Yes, I think it can. In a world where everything is being automated, human connection and relationships matter more than ever.

The ticketing platforms that organise work for service teams are IT systems, and they are commodities now. It is hard to see how a commodity delivers a differentiating experience. I saw demos this week of AI in the loop that were genuinely cool, but they were demos of using AI to speed up the human version of supplying service, not reimagining what service looks like in an AI world. Faster is not differentiation. It might feel novel for a while, then it will be everywhere.

The more interesting use of the time you save is tracing root causes, people-based rather than tech-based, and doing the underlying work that eradicates the need for the ticket in the first place.


A warehouse of books

Laws drive unintended action sometimes. A warehouse of books with their spines cut off by a hydraulic blade, pages fed through a scanner, the remains pulped.

Anthropic ran the programme from early 2024 under the internal name Project Panama. Judge Alsup ruled that buying a print book, scanning it and destroying the original counted as fair use, partly because the digital copy replaced the physical one rather than adding to it. The reporting on the unsealed documents is grim in a very ordinary way.

Although it seems most of the books were not irreplaceable, there must be a better way.

Back to all posts
CurricuLLM Logo
CurricuLLM

AI for schools

Product

FeaturesPricingUse CasesSchoolsAI tutorDevelopersFAQ

Resources

Training hubSupportBlogResearchEvidenceEventsPress

Company

About usTrust & SafetyPrivacy policyTerms of useStatusContact