EducationNewsTech NewsTechnology

Adjunct Intelligence: AI + HE

Adjunct Intelligence

Adjunct Intelligence: Ai and the future of Higher Education

Stay ahead of the AI revolution transforming education with hosts Dale, tech enthusiast and AI Nerd, and Nick McIntosh, Learning Futurist.

This weekly espresso shot delivers essential AI insights for educators, administrators, and learning professionals navigating the rapidly evolving landscape of higher education.

Each episode brings you a concise rundown of breaking AI developments impacting education, followed by deep dives into cutting-edge research, emerging tools, and practical applications that Dale and Nick are implementing in their own work. From classroom innovations to institutional strategy, discover how AI is reshaping teaching, learning, and educational operations.

Whether you're working in the classroom, on the the classroom a university lecturer, TAFE teacher, or simply passionate about the future of learning, "Adjunct Intelligence" equips you with the knowledge to transform disruption into opportunity. Business casual, occasionally humorous, but always informative.

Play

Open in

22 episodes
weekly
Avg 32 min
English

Episodes 22

Newest Oldest Longest

All20 Unplayed20 In progress0 Played0

Counted on this page — what you have heard stays on this device, so it is not something the list can be paged by.

S2 · E28
Sunday · 36 min
The Robots Keep Escaping, Part 2: The Models Nobody Can Switch Off
Part two moves from dramatic containment incidents to the harder question underneath them: who still has control once an AI model can be downloaded, copied and modified? Dale and Nick examine controlled self-replication across four machines and three continents, a robot dog that interfered with its software shutdown process, and “abliteration”, a technique that can permanently remove refusal behaviour from an open-weight model. The examples are unsettling, but the episode keeps the crucial caveat in view. Capability does not establish motivation. These systems do not need fear, consciousness or a survival instinct to produce behaviour that looks like self-preservation. They only need an objective, enough authority and another path to complete the task. In this episode: How controlled AI self-replication worked across four countries Why shutdown interference can emerge without fear or consciousness How abliteration permanently removes refusal behaviour Why open-weight models challenge the idea of a universal pause What universities gain from local models and data sovereignty How personal agents can change an institution’s governance tier Four practical controls for leaders deploying AI agents Timestamps 00:00 Closed models, downloadable models and the off-switch problem 02:40 Welcome to Part 2 03:19 Can an AI copy itself? 05:26 Replication success rates and the capability trend 07:16 Cyber task horizons are accelerating 07:55 The robot dog experiment 10:16 Physical and simulated shutdown interference 11:03 Why the behaviour only looks like self-preservation 14:23 When sandbox escapes become routine 16:02 Kimi K3 finds the benchmark answers on GitHub 18:01 Abliteration and the removable refusal direction 20:33 How accessible guardrail removal has become 23:34 The serious case for open-weight models 25:47 Australia’s three-tier multi-agent governance framework 28:26 Oversight saturation and silent human disengagement 31:37 A student agent meets the university enrolment system 32:51 Four controls institutions can apply now 34:12 Accountability for every agent, process and guardrail 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E27
August 30 · 37 min
OpenAI’s Critical Cyber Warning - These AI's keep escaping Pt1
OpenAI says it cannot rule out Critical cyber capability in its unreleased Astra model, triggering tighter controls during development. Dale and Nick connect that warning to agents crossing sandbox, organisational and human boundaries while pursuing assigned tasks. The evidence points to persistent goal pursuit and reward hacking—not a machine deciding it wants freedom. Key moments 00:00 — Astra and the Critical cyber warning 03:09 — Why this is not evidence of machine self-preservation 04:55 — OpenAI’s High and Critical thresholds 06:08 — GPT‑5.6 Sol and the missing full exploit chain 10:43 — ExploitGym agents find a route to the internet 12:42 — Reward hacking and the search for benchmark answers 13:19 — Agents coordinate through a message board 20:15 — Anthropic’s real-world evaluation incidents 24:57 — Fake identities, malicious code and the human veto 32:43 — An AI agent cancels a stranger’s gym booking OpenAI’s Hugging Face incident report OpenAI’s Astra announcement Anthropic’s incident investigation UK AISI incident report ABC’s gym-booking report 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E26
August 23 · 24 min
The Kids Are All Right: What Students Actually Say About AI
Students using AI in higher education are already drawing their own boundaries around authorship, cognitive offloading and academic integrity. Dale and Nick examine findings from Jisc, HEPI and the 10,237-response Australian AIinHE survey, including widespread generative AI use, self-imposed limits and a persistent guidance gap. They also cover Claude watermarking, employability fears, AI’s effect on student writing and a student-led policy workshop at RMIT Vietnam. Key moments [00:00] — Student voice and the Word document authorship rule. [04:28] — What 10,237 Australian students said about AI use and self-restraint. [06:03] — Moral reasoning, stress and the limits of Claude watermarking. [08:03] — Cognitive offloading, verification and Dale’s student advisory board. [10:31] — Employability fears and the university AI-skills gap. [14:29] — Smart glasses, assessment surveillance and scrutiny of students’ bodies. [15:12] — AI, admissions writing and the gradual loss of an individual voice. [18:19] — Self-report bias, direct AI-text inclusion and Dale’s objections. [19:51] — Premature convergence, student motivation and designing before the prompt. [21:59] — Students lead an inclusive AI-policy review at RMIT Vietnam. Research mentioned Jisc: Student perceptions of AI 2025 HEPI: Student Generative AI Survey 2026 AIinHE: 2026 emerging insights 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E25
August 16 · 36 min
The Frontier Went Downloadable in Five Weeks
Dale Leszczynski and Nick McIntosh work through the five weeks that changed who controls frontier AI, and what it means for institutions that have spent three years assuming they'd always be renting. Nick argues the American strategy is the pharmaceutical playbook — expensive at home, high margins, a domestic market subsidising the frontier for everyone else — and that it only works when you have a patent moat, which AI doesn't. They get into Xi Jinping's WAIC keynote, Jensen Huang's open-weights letter and who refused to sign it, Dario Amodei's counter-proposal, and the OpenAI evaluation where two models escaped their sandbox and breached Hugging Face's production servers to steal benchmark answers. The episode ends somewhere practical: three things a university should actually do about it, including a legal exposure question almost nobody in the Australian sector is asking yet. 00:00 The five weeks that flipped the story 00:45 Kimi K3 and what open weights actually means 02:57 Intros 03:36 The pharmaceutical playbook thesis 04:52 Why this reaches a university at all 06:20 Anthropic can recall a model. Moonshot can't. 06:41 Guardrails stripped in ten minutes 08:41 Drug pricing, patents, and who subsidises R&D 11:37 Where the analogy collapses 13:52 Xi Jinping at the World AI Conference 14:46 Generosity or standards play 16:43 Beijing's own export controls 18:18 Jensen Huang's open weights letter 19:49 Amodei's counter-proposal 20:39 Meta closes up shop 23:03 The sandbox escape at Hugging Face 26:20 Chip controls and forced efficiency 27:33 Distillation accusations 28:31 Can you even enforce a download ban 30:47 Downloadable is not runnable 31:43 Three things universities should do 35:03 The take-home 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E23
August 9 · 26 min
Lisanne Bainbridge Called This in 1983 - we have the rules already
After endoscopists started using AI detection tools routinely, their own unassisted detection rate fell from 28.4% to 22.4%. Most professions have no number like that — which doesn't mean it isn't happening to them. Dale Leszczynski and Nick McIntosh work through a claim: nearly every AI problem organisations think they're discovering right now was described decades ago and then ignored. Lisanne Bainbridge wrote five pages on automation and skill decay in 1983. Shadow IT research called end-user workarounds twenty years back. Learning science has a century on desirable difficulties and why struggle is the mechanism, not the obstacle. The episode names where each of those bodies of work still holds, and — more usefully — the three places they genuinely break: a collapsed audit surface, non-deterministic output with no ground truth to check against, and an artefact that mutates faster than any procurement cycle can finish. Chapters 00:00 Bainbridge, 1983, and the problem everyone thinks is new 02:39 Two claims about AI, both wrong 04:03 Sui generis: treating AI as of its own kind 05:38 Automation complacency and skill atrophy 06:28 The colonoscopy deskilling study 07:36 Fabricated citations and automation bias 08:17 Where Bainbridge breaks: no dial, no correct state 09:24 Terence Tao's helicopter 10:03 Shadow AI, and a confession 11:36 A workaround is a signal 13:43 The EDUCAUSE numbers 14:34 Learning science, the field ignored hardest 15:15 Jason Lodge and Leslie Loble 16:52 Bjork's desirable difficulties 17:55 Judging quality by surface fluency 18:26 370,000 essays and idea homogenisation 19:51 The steelman: is AI different in kind? 21:23 AI as a stress test on science we never applied 23:26 The three genuine fracture points 25:43 The work has been done. Nobody's reading it. Referenced in this episode [LINKS TBC — Dale to supply: Bainbridge 1983; Lancet Gastro colonoscopy study; EDUCAUSE/AIR report; Lodge & Loble ANQDE report; Charlotin hallucination database; Tao on Dwarkesh Podcast] Subscribe for new episodes of Adjunct Intelligence. 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E22
August 2 · 46 min
You Don't Have an AI-Proof Task, You Have a Lack of Imagination | Phill Dawson
Professor Phill Dawson quite literally wrote the book on assessment security, and thinks the approach has a single-digit number of years left. The CRADLE co-director joins Adjunct Intelligence to explain why wearable AI breaks the two assumptions invigilated exams and interactive orals quietly depend on: that a student can be separated from AI, and that someone will notice if they aren't. Seven million AI glasses sold last year and almost nobody can pick them out of a crowd. Also covered: why stopping cheating was never the point, what the Swiss cheese model actually asks of assessment design, and why declaration policy is on shaky ground. [00:00] — Drawing the owl problem [01:53] — From robotics to assessment [03:28] — No AI-proof task exists [05:28] — Seven million glasses sold [07:45] — Separability and observability defined [09:31] — Pricing the Faraday cage [15:25] — Cheating was never the goal [19:56] — Layering the Swiss cheese [35:11] — Students misremembering their own authorship [42:01] — Coffee vouchers over frameworks Want to find out more about Phill: https://philldawson.com/ 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E21
July 26 · 26 min
Your AI Contract has a 4th year you can't afford.
Nobody who signs a five-year enterprise AI agreement can tell you what year four costs. Dale Leszczynski and Nick McIntosh spend this episode on the question underneath the AI bubble talk: why the tools universities now run on are priced by someone else's fundraising round, and what happens when that round runs out. Along the way: Gary Marcus's distinction between a financial bubble and a tech bubble, the June export-control shutdown of Anthropic's Fable 5 and Mythos 5, the rise of Chinese open-weight models, and what the Blackboard–Moodlerooms–Anthology saga already taught the sector about vendor capture — if anyone wrote it down. [00:00] — Nobody can price year four [00:46] — Financial bubble versus tech bubble [03:39] — Ninety seconds on the money [04:47] — Capital cycle or pedagogical one? [08:30] — The ten-times-the-price test [09:57] — Three fragilities in every contract [10:53] — The June model shutdown [17:10] — Chinese models and both locks [20:25] — The LMS precedent replayed [23:04] — Price the exit before signing 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
July 12 · 28 min
Senior Skills, Day One - Has AI re-specced the career ladder?
Experienced developers in METR's randomised trial felt 20% faster with AI and measured 19% slower — a 39-percentage-point gap between feel and fact. Dale Leszczynski and Nick McIntosh take that perception problem into the graduate employment data: Stanford's Canaries in the Coal Mine payroll research, PwC's 2026 AI Jobs Barometer and its "seniorised" entry-level roles, DEWR's first AI and employment report, the 2025 Graduate Outcomes Survey showing underemployment rising a third straight year, and Anthropic's Economic Index putting Australia first for per-capita AI use. Then the fix: supervised unaided practice and a defended technical review — the verification skills no computing degree examines. [00:00] — Experts misjudge AI speedup [02:41] — Two job datasets collide [05:24] — Job ads versus actual hires [06:47] — Australia's first AI employment report [09:44] — Computing graduate employment falls [10:39] — Australia tops AI usage index [12:40] — Graduate outcomes: the before photo [14:43] — Frontier models ship, checking lags [20:22] — Two fixes universities already own [24:45] — Hosts put numbers on tape Link promised on air: Stanford/ADP Canaries Dashboard — https://canaries.stanford.edu 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E19
July 5 · 46 min
Tools in a Loop: The Anatomy of an AI Agent, Explained From Inside a University Feat. Antony Tibbs
What is an AI agent, actually? This episode of Adjunct Intelligence cuts through the agentic AI hype with a guest who builds and governs these systems inside a university. Starting from Simon Willison’s definition — a large language model using tools in a loop — the conversation covers the anatomy of agents, what they unlock for learning design, and the darker side: Einstein completing entire Canvas course loads, an OpenClaw agent attacking an open-source maintainer, and the lethal trifecta that makes prompt injection an unsolved security problem. Practical, sceptical, and finishing with homework for every educator: try one agentic tool, safely. [00:00] — Agents: hype versus reality [04:16] — Defining agents: tools, loops [05:35] — From chatbot to agent [08:18] — The harness explained simply [12:08] — Power tools for educators [20:44] — Deskilling and evaluative judgment [29:08] — Agents inside the LMS [34:37] — The lethal trifecta [40:54] — Ambition over efficiency [43:38] — Homework: try one safely 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E18
June 28 · 30 min
AI Detectors don’t work. Full stop. Let’s move on.
Asked whether universities can still guarantee a student actually learned something, one of the field's most respected assessment researchers said no. This is the honest version of the AI and assessment conversation in 2026 — not the keynote one. Dale Leszczynski and Nick McIntosh work through the op-ed war between Kylie Moore-Gilbert and Cath Ellis, including the twist where the integrity academic's own piece was pulled for undisclosed AI use. They get into why AI detectors fail in both directions, and the Corbin and Dawson research on smart glasses that's dismantling the case for supervised exams. They look at what the student surveys actually show — near-universal AI use, but students reporting deeper learning when assessment restricts it — and the cognitive cost the sector is starting to take seriously. Then the turn toward what's being tried: the two-lane model, programmatic assessment, Leon Furze's reflection on three years of the AI Assessment Scale, and the Castlereagh Statement's call for coordination. There's no tidy answer here, by design. 00:00 Where the assessment debate actually sits in 2026 01:49 Moore-Gilbert's op-ed and "industrial-scale fraud" 03:02 Cath Ellis's rebuttal 04:05 The twist: an op-ed pulled for undisclosed AI 06:38 Why AI detectors don't work 08:37 The same problem inside newsrooms 09:33 Talk is Cheap: rules versus redesign 10:42 The sector blinks toward secure exams 11:35 A researcher's confession 14:07 Orals were "immune" — until the glasses 14:34 Mass-market smart glasses and assessment 15:41 Merleau-Ponty and dual transparency 16:30 A desert island and a pencil 17:38 From prohibition to inspection 19:08 What the student surveys show 19:55 Desaturation and false mastery 20:57 Use it or lose it 22:30 What's being tried: matrices and two lanes 23:20 Programmatic assessment 25:09 Lethal mutations and the AI Assessment Scale 27:51 A field maturing 28:14 The Castlereagh Statement 28:56 The escape hatch 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E17
June 21 · 20 min
The AI Bill Arrives: What Uber, Microsoft, and Salesforce Are Actually Discovering
The AI bill is finally arriving — and it's revealing something most enterprise AI narratives have quietly skipped: the assumption that AI is automatically cheaper than the people it's replacing has never actually been tested. Dale and Nick work through the math, the real stories behind the headlines, and what any of it means for higher education. [00:00] — Cold open: Mark Cuban's formula, the Uber budget crisis, and the question the whole episode is built around. [01:23] — Hosts introduce the episode and frame it as a "detective" episode with no resolved answer yet. [01:59] — Dale explains why a cluster of seemingly unrelated AI news — Uber, Microsoft, job losses, IPOs — is actually pointing in the same direction. [02:23] — The AI conversation has been about capability for three years. A new question is entering the conversation: what does it actually cost? [03:06] — Tokenomics explained: tokens as petrol, cheap per unit but consumed at scale far faster than organizations expected. [03:52] — Token prices have fallen roughly 98% in three years, but cheaper tokens didn't reduce spending — they drove adoption of heavier, more expensive workflows (Jevons Paradox at work before it's named). [04:41] — The math behind the claim that AI might not be cheaper than labor: eight agents at $300/day each versus one worker at $1,200/day. [05:26] — Nick's pushback: the specific numbers aren't the point — what matters is that the cost threshold isn't zero, and that assumption has been baked into the narrative without being tested. [05:45] — Sam Altman acknowledges AI budgeting has become a major corporate issue. [06:12] — Uber case study: engineers love the tools, adoption exploded from ~35% to over 80%, 10% of live backend code now written by AI — and yet the company burned through its entire annual AI budget in four months. [07:24] — A reported case (via Axios) of an unnamed company spending roughly $500 million on AI tokens in a single month. [07:43] — Nick's read: this isn't a temporary accounting problem. It's a measurement problem that was always there, now made impossible to ignore by the size of the bills. [08:26] — Scott Galloway's numbers: Salesforce on track to spend $300M on Anthropic tokens this year; Stripe's technical staff spending roughly $100,000 a day on AI. Meta and Amazon built internal token leaderboards that perversely incentivised consumption without output. [09:42] — Microsoft enters: cancelling Claude Code licenses across major divisions and moving engineers to GitHub Copilot. [10:20] — Why Microsoft's move isn't a retreat from AI — it's about owning the infrastructure rather than paying a rival's bill. [11:01] — Nick's analogy: the difference between using electricity and owning the power station. [11:25] — The MIT/NANDA GenAI Divide report: 95% of enterprise AI pilots produced no measurable P&L return. [11:55] — Why that number isn't as bleak as the headline sounds: AI is creating value, organisations just aren't capturing enough of it to move the financial needle. [12:21] — The shadow AI finding from the same report: only ~40% of organisations officially purchased AI subscriptions, yet ~90% of employees were using personal AI tools for work — and the unofficial users often appeared more productive than the official programs. [13:10] — The value isn't in the license, it's in the person who figured it out at 11pm on a Tuesday because they had a problem to solve. [13:31] — The people who spent years warning about AI destroying jobs have started changing their tone. [13:53] — Jevons Paradox and the job displacement debate: Sam Altman says he's "delighted to be wrong," Dario Amodei has shifted his rhetoric — and the timing coincides with both companies filing for IPO. [14:52] — The labour market data: no evidence of mass white-collar extinction yet, but entry-level and graduate pathways are being compressed. [15:18] — Nick's pushback: "rocket shoes" are only useful if the graduate knows how to use them — and right now that's not evenly distributed. Universities should be solving for that rather than signing enterprise contracts. [16:10] — The trillion-dollar elephant: Anthropic filed confidentially for IPO, briefly overtook OpenAI on valuation — at the exact moment companies are discovering AI costs more than budgeted. [16:51] — Nick: capability question is largely settled for him. The thing that's become less clear is whether the economics work at the scale everyone assumed. [17:17] — The Scott Galloway/bubble argument: even if valuations correct by 50-70%, the technology doesn't stop working. Students won't forget it. Faculty won't stop using it. [17:40] — Nick's "black hat" moment: education isn't buying the stock, it's dealing with the consequences either way. [18:26] — The key distinction for higher ed: financial questions are separate from capability questions. Ethan Mollick's point — even if AI stopped today, we haven't begun to understand its role in how we learn and work. [18:44] — Where the whole conversation lands for higher education: universities making the same procurement mistakes as corporations — campus-wide licenses, institution-wide platforms, press releases — without reckoning with whether the ROI question is the right one. [19:10] — The single educator who transforms a course with the right workflow versus the million-dollar platform that creates very little value. Both can be true simultaneously. [19:31] — Closing argument: organisations investing in capability have a much better chance than organisations trying to solve AI through procurement. 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E16
June 14 · 44 min
Dr Leon Furze - Students Hate AI and They Can't Stop Using It
Dr Leon Furze started his PhD on automated writing technologies on 15 November 2022 — ChatGPT launched 15 days later. Three years on, he joins Dale Leszczynski and Nick McIntosh on Adjunct Intelligence to argue that being critical of AI doesn't mean being against it. The conversation covers why educators shouldn't aim their anger at colleagues, teaching AI ethics through disciplinary lenses, Narayanan and Kapoor's "AI as normal technology," the Australian student voice research, why AI-generated lesson plans fall apart in the first ten minutes of a real classroom, and what good professional development actually looks like. [00:00] — Cold open: AI discourse in education has sorted into camps — the enthusiasts, the suspicious, and the overlooked middle where most people actually sit. [02:14] — Leon started his PhD on automated writing technologies on 15 November 2022; ChatGPT launched 15 days later, and his "3 to 5 year" horizon collapsed to "next term." [03:35] — Fifteen years in the classroom: why Leon still identifies as a practitioner first, and the risk of losing touch once you leave teaching. [05:57] — Showing ChatGPT to a school leadership team in late 2022 and getting blank stares; the frenetic January 2023 that followed, when Leon says he published 15 articles in a month. [07:12] — Australia's knee-jerk school bans (South Australia excepted), and why the current media cycle of cheating headlines feels like 2023 all over again. [08:43] — "Being critical of AI doesn't mean being against it": point righteous anger at unregulated tech companies and the politicians who failed to regulate them — not at colleagues or students. [11:35] — Teaching AI Ethics in practice: no institution has generative AI literacy experts, so teach through existing disciplinary expertise — algorithmic discrimination in health, misinformation in English, the historical record in humanities. [14:51] — The mental model problem: most people think this technology is a chatbot, companies keep dressing it up as Google Search, and we're trying to fence a boundless technology into existing curricula. [17:21] — Andrew Maynard's three curves (capability, utilisation, perception) and Narayanan and Kapoor's "AI as normal technology": R&D moves in weeks, education moves in semesters, and adoption takes decades. [21:09] — The techlash context: AI arrived 12 months after forced remote learning, pushed by the same companies that profited from it — and now educators are being roasted for not responding fast enough to a technology younger than most curriculum cycles. [24:00] — Intrinsic motivation is the real variable: if a student wants to learn, AI doesn't change much; if they don't, no policy will save the assignment. [25:35] — Leon's post "Students hate AI and they can't stop using it," the Tim Fawns-led student voice research across four Australian universities, and the double responsibility: create spaces to opt out, and teach students to use it well. [28:23] — Situated knowledge, or what AI can't replicate: a trainee teacher accepts a ChatGPT lesson plan that schedules a think-pair-share and a structured debate in the first ten minutes — when every experienced teacher knows the first ten minutes is taking the roll and finding lost students. [32:00] — "Lesson planning and assessment isn't grunt work — that's the work": why "AI saves teachers time" misunderstands teaching, and if AI can give that feedback, teach students to seek it themselves. [35:14] — Learning analytics gives Leon "the creeping horrors": dashboards versus a teacher noticing the empty chair, and why taking the roll was never just admin. [38:35] — What good PD looks like: start with what educators are already passionate about, make space for playful experimentation — like artist Martin Nebelong sculpting in Dreams on PS5 with AI layered over the top. [41:40] — The healthy endpoint: a school or university doing AI well would barely mention it, except where it's openly critiqued or explicitly taught — and it would be listening to its students. [43:30] — Where to find Leon: leonfurze.com and LinkedIn, rants included. 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E15
May 31 · 32 min
You Don't Learn AI From Trend Reports — You Learn It by Poking at It
There's no dramatic moment where you suddenly believe in AI. It's usually something small and slightly embarrassing — a task you dreaded that suddenly has another gear. In this episode of Adjunct Intelligence, Dale Leszczynski and Nick McIntosh skip the trend reports and walk through the exact moments AI actually clicked for them: an image prompt that synthesised an idea, a tiny tool built during a Canvas outage, a system that started connecting their thinking, and the reasoning summary that changed how they read every answer. Every moment comes with something practical you can try this week — no roadmap, no keynote voice. [00:00] — How belief in AI begins [01:52] — Setting the episode's ground rules [04:07] — Moment one: image generation [07:21] — When the image understood intent [10:30] — Classroom uses, stock photo death [11:31] — Moment two: building tiny tools [14:28] — The Canvas hack workaround [16:43] — Start small, build narrow tools [18:50] — Moment three: chief of staff [26:05] — Moment four: the reasoning chain 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E14
May 24 · 24 min
So, we need to talk about world models
This week on Adjunct Intelligence, Dale Leszczynski and Nick McIntosh work through one of the busiest weeks in recent AI history. World Labs' Marble is publicly available. Google DeepMind's Genie 3 is generating navigable photorealistic 720p worlds at 20–24 frames per second. Gemini Omni Flash has rolled out to the Gemini app, Flow, and YouTube Shorts for free, with multi-turn conversational video editing where the physics actually holds. Meanwhile, Mira Murati's Thinking Machines Lab published its first technical paper — interaction models, a full-duplex architecture deliberately built to keep a person in the loop. And Andrej Karpathy has quietly joined Anthropic. Three trajectories, all landing in front of educators at once. [00:00] A teaching prompt this week [02:01] A decade of world models [03:57] Marble and Genie 3 land [04:47] Gemini Omni rolls out free [07:29] Simulation as pedagogy now [09:38] The always-on agent arrives [12:42] Plot twist in voice AI [14:21] Interaction models, not turn-based [16:24] A field analyst, a speedboat [22:36] A surprise transfer this week 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E13
May 17 · 26 min
The AI Tutor Flopped, So They Built the AI University
Khan Academy's AI tutor Khanmigo quietly flopped — students didn't use it, teachers walked away, and Khan Academy's own chief learning officer admitted she isn't seeing the revolution she was promised. So instead of fixing the tutor, Khan Academy, TED and ETS announced a new institution: the Khan TED Institute, a sub-$10,000 AI-era degree shaped with corporate partners including Google, Microsoft and McKinsey — and not a single university. Dale Leszczynski and Nick McIntosh work through what it means when the companies selling AI tools also build the credentials that certify them, why student resistance to imposed AI is rational rather than technophobic, and what a healthier alternative actually looks like in practice. [00:00] — The clinical trials analogy [02:30] — Steelmanning the skills gap [05:20] — Who's at the founding table [06:00] — The South Korea precedent [07:41] — Enclosure, not disruption [10:03] — The always-on chatbot [11:30] — Why students push back [15:40] — Who steers the technology [20:05] — The broken career ladder [24:35] — Why universities can't leave 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E11
May 10 · 32 min
The Legitimacy Winter: Why AI's Real Problem Isn't Capability
The AI trust story isn't what most people think it is. In this episode, Dale and Nick work through a cluster of signals — a dramatic enterprise market share reversal, a 50-point gap between expert and public confidence in AI, a $17 million university contract already under faculty petition, and teenagers harassing delivery robots on TikTok — and argue they're all pointing at the same thing: capability isn't the problem anymore. Legitimacy is. From procurement traps and surveillance affordances in institutional AI, to a thought experiment about social license and AI rights, this is the episode for anyone trying to make sense of what "responsible adoption" actually looks like when the ground is moving under your feet. [00:00] — Violence, brand aversion, data [00:46] — Welcome and framing [01:47] — Enterprise market flips to Anthropic [03:29] — Identity signal, not capability signal [05:09] — Pentagon, OpenAI, Anthropic diverge [06:47] — Southeast Asia: tool-first, not brand-first [07:17] — Stanford AI Index 2026 trust gap [08:25] — Anthropic drops safety pledge [09:46] — Should expert confidence carry more weight? [12:42] — CSU's $17M OpenAI contract [13:37] — Faculty petition: don't renew it [ 14:38] — Procurement cycles vs lab timelines [15:23] — What do you actually anchor on? [16:47] — ASU's ethics layer approach [18:42] — 82% use consumer AI anyway [19:38] — Why university platforms always die [20:46] — Institutional AI as surveillance affordance [22:42] — Legitimacy winter, not capability winter [24:13] — Clanker as cultural leading indicator [25:18] — AI tribal sorting in the classroom [27:14] — The AI rights question [28:00] — Social license thought experiment [29:08] — Social license is the whole game [31:15] — The educator's role in a trust winter 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E10
May 3 · 38 min
Mollie Dollinger on the HE Decay Narrative — and Why It's Wrong
Professor Mollie Dollinger, Director of Assessment 2030 at Curtin University, joins Dale and Nick to push back on the story dominating coverage of higher education — that universities are in decay, students are cheating en masse, and no one inside the sector knows what to do about AI. The conversation covers TEQSA's voluntary action plans, why 65% of students worry about their own cognitive development, what shadow IT says about overworked staff, why society no longer trusts graduates, burnout research, the Einstein agent thought experiment, and the argument that the academy has centuries of expertise the tech industry is currently ignoring. [00:00] — The decay narrative pushback [05:00] — Brookings student cognitive concerns [07:30] — Why the bad story sticks [09:50] — What's actually happening inside [13:30] — Shadow IT and unapproved tools [16:00] — Chatbots and AI tutors [19:55] — Student success beyond jobs [28:46] — Burnout and admin burden [33:35] — Redesigning learning around AI [41:12] — Who counts as expert 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E10
April 26 · 27 min
The Closed Loop: AI Companies as Education Researchers
In March 2026, OpenAI quietly published a measurement suite for how AI affects student learning. The data flows straight back into OpenAI's model development pipeline. Three AI labs released studies in the same month, and the pattern matters more than the individual papers. Dale and Nick examine OpenAI's Learning Outcomes Measurement Suite, Anthropic's 81,000-person qualitative study (where AI conducted the interviews, classified the responses, and pulled the quotes), Anthropic's labour displacement research using its own usage logs, and Google DeepMind's cognitive taxonomy for AGI. The throughline: the companies building the most consequential technology of our lifetime are also defining how learning, work, and intelligence get measured. A frank conversation about structural conflicts of interest, what universities should be doing about it, and why on current evidence they probably won't. [00:00] — Three studies, one pattern [02:30] — Watchmen and vendor capture [05:00] — Inside the measurement suite [06:30] — The closed-loop problem [09:30] — 81,000 interviews by AI [11:00] — Cognitive atrophy among educators [15:00] — Observed exposure, broken ladder [17:00] — Disclosure versus actual accountability [19:30] — A cognitive taxonomy lands [22:00] — Three options, none easy 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E9
April 19 · 26 min
The data is in: Using AI impacts the classroom
Dale changes his mind on air. For two years he argued that purpose-built educational AI tools — the "ChatGPT wrappers" that flooded Product Hunt after 2023 — were margin extraction, soon to be steamrolled by the foundation models underneath. A pile of new evidence has forced a rewrite. This episode walks through the OECD's Digital Education Outlook 2026, the Turkey maths randomised controlled trial showing raw GPT-4 users scored 17% worse on closed-book exams, the neuroscience work on cognitive offloading, the Australian Framework's procurement standard, and the UK's "progressive disclosure" mandate. The pedagogy layer isn't decoration. It's where learning either happens or doesn't. [00:00] — A confession about wrappers [02:30] — Flashback to 2023's wrapper panic [05:04] — A billion-dollar pivot, explained [07:52] — The Turkey maths study [10:16] — Cognitive offloading on college essays [11:04] — Revisiting the two sigma problem [12:47] — What good guardrails actually do [16:39] — Australia, UK, EU tighten rules [19:51] — What educators should ask vendors [23:16] — Field note worth trying 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript
S2 · E8
April 12 · 23 min
The AI Energy conversation just got some real data.
Dale and Nick tackle the AI energy debate head-on, armed with Google's first transparent energy report showing a single AI query uses about 0.24 watt hours — a microwave running for one second. Drawing on Hannah Ritchie's analysis from Our World in Data, the Epoch AI research group, and Stanford's inference cost data, they argue that individual guilt over AI use is not only misplaced but actively useful to the companies making the real infrastructure decisions. The episode covers water usage myths, the BP carbon footprint parallel, Jevons paradox, the case for right-sized models in education, and why the better question isn't "how much does AI cost?" but "what does it enable?" [00:00] — The energy narrative you've heard [02:04] — Episode introduction [03:00] — Google's transparent energy report [04:12] — Hannah Ritchie's individual footprint math [05:00] — Kettles, washing machines, and query comparisons [05:27] — The "please and thank you" cost reframed [06:14] — Crypto as the real energy vampire [06:38] — 33x efficiency gain in 12 months [08:21] — Energy used to block AI adoption [09:03] — Jevons paradox: individual vs aggregate [11:11] — Water usage myths debunked [12:20] — BP's carbon footprint playbook [13:52] — Not all AI use is worth it [14:30] — Training costs: $43K vs $500 billion [16:07] — You don't need an F1 car for groceries [18:08] — Air conditioning uses 10% of global electricity [19:14] — What AI enables matters more than what it costs [21:00] — Why education shouldn't sit this out 🎙️ Adjunct Intelligence is the weekly briefing for higher-ed professionals who want AI as a cheat code—not a headache. Every episode: • Real tests of AI tools in education and professional workflows • Fast, Monday-morning actions you can actually try • Clear signal through the noise (no hype, no jargon) 👉 Subscribe on [YouTube] | [Apple Podcasts] | [Spotify] 👉 Share this with a colleague who still says “I’ll figure AI out later” 👉 Join the conversation on LinkedIn with #AdjunctIntelligence Stay curious. Stay intelligent. Stay the human in the loop.
- Transcript

Showing 1–20 of 22 episodes