Griffin on Tech: ChatGPT-5 finally drops, RIP NCEA, and Elon’s share package

OpenAI today launched ChatGPT-5, its most advanced conversational AI model, following notable delays that fueled speculation that Sam Altman’s start-up was struggling to squeeze more performance out of its large language models (LLMs). 

Reports suggest the OpenAI team wrestled with achieving performance improvements significant enough to justify a new flagship model. Altman himself acknowledged the delays on X, writing of “capacity crunches” and suggesting a “choppy” period as new models and features rolled out. 

This new model is crucial for OpenAI. After losing ground to competitors like Google, Meta, Anthropic, and even DeepSeek. Altman has promised a “PhD-level” leap in intelligence with GPT-5, which is available in numerous variants, as well as speed and reliability improvements. GPT-5 has been touted as having powerful performance in areas like coding, reasoning, and autonomous “agentic” uses.

Simon Willison, a British software developer and co-creator of the Django web platform has been using ChatGPT-5 for two weeks under embargo and concluded this morning: “My verdict: it’s just good at stuff. It doesn’t feel like a dramatic leap ahead from other LLMs but it exudes competence. It rarely messes up, and frequently impresses me. I’ve found it to be a very sensible default for everything that I want to do. At no point have I found myself wanting to re-run a prompt against a different model to try and get a better result.”

I’ve been using ChatGPT-5 this morning, and it definitely seems faster in compiling results than its predecessors and has better subject matter understanding. Benchmarks show up to a 65% decrease in hallucination rates compared to GPT-3, and a 26% drop versus GPT-4o. In health-related queries and coding tasks, its scores significantly outperform older models. Making GPT-5 available to all ChatGPT users, not just premium subscribers, is a good move. Is the singularity a giant leap closer? Are we closing in on AGI? I doubt it.

“The way that most of us define AGI, we’re still missing something quite important, many things that are quite important, actually, but one big one is a model that continuously learns as it's deployed, and GPT-5 does not,” Altman said on a media call yesterday. So that’s a no. Interestingly, as Wired recently reported, once OpenAI does get to what it considers AGI, it can start charging Microsoft, its major investor, for access to the latest models - or cut the software giant off from accessing them entirely.

OpenAI maintains that scaling laws still apply, meaning continued investment in data and computational resources will yield further improvements. While sceptical voices point out that true artificial general intelligence (AGI) remains elusive and that existing approaches may encounter diminishing returns, GPT-5 looks like setting an impressive new benchmark for model performance and conversational AI. 

Bye, bye NCEA

I was done with the high school education system seven years before NCEA was introduced, and I’ve never been able to wrap my head around what replaced the system I grew up with. One thing I did like the idea of was not having everything rely on one set of terrifying three-hour exams at the end of the academic year. I saw the devastating consequences of that in friends who had a bad exam day.

Well, NCEA is about to be consigned to the bin too. So what will it mean for efforts to equip our kids to harness digital skills, and join our high-tech workforce? It’s hard to tell. Even the panel of experts the Science Media Centre lined up to comment on the new qualifications framework (Foundational Skills Award in Year 11, NZ Certificate of Education in Year 12, and NZ Advanced Certificate in Year 13) basically said it will be disruptive and we’ll have to wait and see how it works in practice.

What’s clear is that the changes represent the biggest secondary education shakeup in two decades, and it will have disruptive consequences for digital skills teaching and vocational pathways in high schools. The transition timeline forces schools to run two systems in parallel for several years, stretching teacher capacity and resourcing just as they try to keep pace with fast-evolving digital curricula. 

Consultation runs to 15 September 2025, with phased implementation from 2026 and first new qualifications landing 2028–2030, meaning cohorts will be split between old NCEA and the new certificates, and teachers will juggle different assessment models simultaneously. That administrative load alone diverts time from updating digital and tech-related programmes, which already suffer from declining enrolments and skills mismatches identified in prior sector reports.

Some argue that removing NCEA Level 1 eliminates a key early scaffold for digital and tech pathways that rely on staged progression and confidence-building at Year 11. Level 1 currently provides entry-level Digital Technologies standards and vocational tasters; swapping it for a narrow literacy/numeracy Foundational Skills Award risks deprioritising hands-on digital learning in a pivotal year when students choose senior subjects. 

The proposed shift to five compulsory subject slots in Years 12–13, with pass requirements across at least four, could squeeze out applied digital and vocational options in timetable-constrained schools. Where staffing is tight, high-stakes grading (A–E with marks out of 100) may incentivise “safer” academic offerings over industry-aligned, equipment-intensive courses like computing labs, mechatronics, or trades-integrated tech modules.  NCEA’s flexibility, for all its flaws, enabled schools to tailor digital and vocational offerings.

Finally, while the Education Minister touts AI marking and digital exams as an enabler of the transition, assessment technology does not substitute for the complex work of designing, staffing, and resourcing modern digital programmes. Marking efficiencies won’t resolve shortages of specialist teachers, or industry partnerships required for credible senior digital and vocational courses.

Even if the end-state aims are laudable, the phased abolition of NCEA will be massively disruptive to digital skills and vocational education in the next few years, just as New Zealand needs more, not fewer, stable pathways into tech and industry.

Musk’s mega-grant

Elon Musk just bagged a nearly US$29 billion stock award, 96 million new Tesla shares priced at a bargain $23.34 apiece, an “interim” gift to keep him parked in the C-suite for two years while courts chew over his voided 2018 megagrant. 

The Tesla board’s goal is to retain the world’s most irreplaceable founder amid an AI talent war, performance targets be damned. That’s at the same time as shareholders propose taking a class action lawsuit against Tesla for overhyping its new robotic taxi business.

It all explains the recent low-key Musk, a rare bout of volume control while the check clears and litigators lurk. The reality is that Tesla’s core business is hurting. It’s hurting. Tesla’s Q2 deliveries fell to about 384,000, down roughly 14% year over year, the steepest drop on record, marking a second straight quarterly decline and confirming the EV engine is running out of juice.

Tesla’s fix seems to be to double down on retention pay, and hope the stock keeps levitating on the promise of autonomy and AI tomorrow while the showroom today gets quieter.

Photo credit: Unsplash/Milan Csizmadia