Bringing the course modern relevance
Reading List
Articles, papers, and essays connected to every topic the course teaches — drawn from the knowledge library, not re-authored per cohort.
What Is AI?
Anthropic
How is the world actually using AI?
A live dashboard tracking AI's effects on the economy — which tasks people automate versus augment with Claude, broken down across occupations and task categories. Students mine the real distribution of AI use, then survey their own class to compare local habits against the global picture.
What Is AI?JobsData.ai
Which parts of your future job will AI do first?
An interactive tool that decomposes a job into its bundle of tasks and shows which face AI cost-pressure first, which stay human, and how the economics shift. Students look up their intended career, audit its exposure task by task, and compare across the class.
What Is AI?
Models & Cost
Jean Marie Bonthous (Medium)
$200 to $1,000 for 15 minutes, broken down
A working filmmaker itemises the true cost of a 15-minute AI film, from $200 to $1,000, with where the money goes.
Models & CostMindStudio
What an AI film actually costs to make in 2026
A line-by-line breakdown of the real spend behind an AI-assisted short — model credits, tools, and time — the reference figures for the cost-per-prompt reasoning.
Models & CostMindStudio
A full AI short-film production workflow for under $200
An end-to-end account of producing a short film with Claude for scripting, Seedance, and Luma Canvas — the concrete pipeline behind a microbudget AI production.
Models & Cost
Critical Media
Ars Technica · Jun 2026
Your doorbell knows your face — and a lawsuit says Ring owes you for it
Ars Technica (surfaced on Hacker News) reports a lawsuit arguing that Amazon-owned Ring should pay Americans for scanning their faces — a biometric-privacy / facial-recognition claim against a consumer surveillance product. Pairs with the surveillance and deepfake threads on the ai-seoul critical-media spine. (Article body not retrieved — see curator_note.)
Critical MediaSignal · Jun 2026
Signal to the UK: a snitch in every device is not safety
Signal's statement opposing the UK's latest privacy proposal, which (per the linked statement and discussion) would push device-level / client-side scanning and age-verification — real-time, on-device detection of content before it is encrypted — effectively building surveillance into every operating system, camera and app. Signal's argument: surveillance is not safety; mandated client-side scanning breaks the security and privacy guarantees it claims to protect. A strong critical-media case on the encryption-vs-child-safety framing. (Statement is a PDF; body not clipped — summary draws on the title/description and the HN thread.)
Critical Media
Six Lenses
New Scientist · Jun 2026
The kill-decision left the loop: autonomous drones reportedly killed soldiers
New Scientist reports — attributed to a senior figure in the Ukrainian defence industry — that fully autonomous, AI-controlled drones have killed human soldiers for the first time, i.e. with no human in the targeting-and-fire loop. A textbook autonomous-weapons case for the ethics tribunal: who is accountable when the kill decision is delegated to a machine? (Article body not retrieved — see curator_note.)
Six LensesNature · Edmond Awad et al. · Oct 2018
What 40 million people decided a self-driving car should do
Awad et al. crowdsource trolley-style verdicts for autonomous vehicles across 233 countries — and find ethics is not universal.
Six LensesThe Yale Law Journal · Judith Jarvis Thomson · May 1985
Why pulling the lever feels different from pushing the man
Judith Jarvis Thomson sharpens Foot’s case into the lever-vs-footbridge contrast that still divides every room.
Six LensesOxford Review · Philippa Foot · Jan 1967
The essay that invented the trolley problem
Philippa Foot introduces the runaway-tram case to probe the difference between intending harm and merely foreseeing it.
Six Lenses
Ethics Tribunal
Car and Driver · reported · Oct 2016
A carmaker says the quiet part out loud
A 2016 Mercedes-Benz statement that its autonomous cars would protect the occupant first — and the public backlash that followed.
Ethics TribunalAnthropic
What rules should an AI live by?
Anthropic's framework for how Claude should behave — four core principles in priority order (broadly safe, broadly ethical, compliant with guidelines, genuinely helpful) plus value categories of helpfulness, honesty, and harm-avoidance. It favours contextual good judgement over rigid rules, which makes its trade-offs genuinely debatable.
Ethics Tribunal
Knowledge & Retention
Image to Motion
Lushbinary
Head-to-head on the three leading AI video models
A practical comparison of Sora 2, Veo 3.1, and Kling 3.0 (plus Seedance) across motion, fidelity, and control — the model-selection map for the image-to-motion stage.
Image to MotionRizzGen
Runway, Kling, Veo, Sora, LTX, Wan, Seedance side by side
A broad survey of seven AI video generators on the same prompts — useful for seeing how much model choice changes the output.
Image to MotionRoborhythms
Two flagships tested on actual creator jobs
A workflow-grounded comparison of Sora 2 and Runway Gen-4 — not benchmarks, but how each behaves on real shot briefs.
Image to Motion
Score & Sound
AudioCipher
A hands-on AI scoring pipeline inside a real DAW
A practical walkthrough of scoring to picture using MusicGen and GPT-4 alongside a digital audio workstation.
Score & SoundMargabagus
Vocals, editors, stems, pricing and the legal question
A current comparison of the two leading AI music tools across vocals, stem export, pricing, and the copyright situation — the brief for choosing a scoring tool.
Score & SoundSoundverse
Using AI music as a creative partner for filmmakers
How AI music tools fit dramatic scoring — where they help and where a human ear is still required.
Score & Sound
Trailer Edit
The Drunk Projectionist
The editor behind Scorsese on what a cut is for
A study of Thelma Schoonmaker’s editing — timing, rhythm, and why the cut is an emotional decision, not a technical one.
Trailer EditNo Film School
Lessons from cinema’s most decorated editor
No Film School distils Schoonmaker’s craft into transferable editing principles for new editors.
Trailer EditAI Tools for Content Creators
Which AI editor for cutting the trailer
A 2026 comparison of OpusClip and Descript for AI-assisted editing — the tool decision for the trailer-edit stage.
Trailer Edit
Finish & Screen
Frame.io Insider
What changed in the delivery + review pipeline
The latest Frame.io release notes — relevant to how finished work is delivered, reviewed, and version-tracked.
Finish & ScreenGoogle
Google ships Content Credentials across its products
Google’s account of adding C2PA Content Credentials to its generative tools — a major-platform anchor for the provenance discussion.
Finish & ScreenNumonic
How AI-content provenance metadata actually works
A primer on the C2PA 2.1 and IPTC 2025 standards for tagging AI-generated and AI-edited media — the technical backbone of the disclose-your-AI requirement.
Finish & Screen