ACOUSTIC COVERS BY INDIE FEMALE SINGER-SONGWRITERS

Since April 13th, 2013, Spotify has pulled thousands of hours of my listening data. The Swedish engineers ruining affordability in SoFo know a lot about me. Which is why it’s baffling when they insist I’m clinically disappointed.

Their verdict arrives in Discover Weeklies, Daily Mixes, Release Radars full of somber, undanceable ballads. Typically, acoustic covers by indie female singer-songwriters. Think Phoebe Bridgers covering Manchester Orchestra. I like Phoebe Bridgers. I like Manchester Orchestra. But I don’t like when one sadly and acoustically covers the other.

When the recommendation algorithms first dropped on Spotify, my Discover Weekly was exactly what it was supposed to be - “...a highly personalized, algorithmically generated mixtape delivered every Monday to make music discovery effortless and deeply personal.” But something happened. There was a shift.

What changed? I tried to figure it out. I intentionally liked happy songs. I quickly skipped the sad songs. I hid songs manually, one by one, on recommended playlists. I continued to make playlists of music I actually like - music that sounds nothing like what Spotify recommends. It didn’t help and it continues.

Just 3 days ago, my Discover Weekly opened with a synthy cover of “Crash Into Me” by Petey USA. (I’ve never intentionally listened to Dave Matthews. I’ve never heard of Petey.) Then 8 minutes of “moody, crescendo-heavy” Explosions In The Sky, a la Friday Night Lights climaxes. (If you haven’t seen it, Alien: Earth is excellent and stars Sydney Chandler, the real-life daughter of Coach Taylor.)

MACHINE DRIFT

In the grand scheme, my Spotify recommendations don’t really matter. But the impact of “Machine Drift” - when software slowly diverges from its intended configuration, performance baseline, or operational environment over time due to continuous changes [1, 2, 3] - does matter. Particularly as (probabilistic) LLMs weave deeper into organizations.

Like my Spotify, more context handed to a Big Tech platform or model doesn’t automatically improve “personalization” or “recommendation.” Context isn’t intimacy, but surveillance with better manners. Why should we believe tomorrow’s LLMs will be any better at steering us collectively? This question keeps coming up as the Second Brain conversation slides from individual use cases on AI Twitter to actual products aimed at organizations.

Jack Dorsey’s Buzz - a Slack and GitHub competitor launched two weeks ago - drops agents into your team chat and code repos. Multiply wraps agency project management in AI (!!). got tired of "duct-taping her brain together," so she built Digital Culturist - a signal library for strategists and teams to turn fragments into insights. Three examples from a range of new products that want us to collectively Second Brain. And regardless of how poorly Accenture implements this silver bullet into your workplace, it seems like a swell idea from a management POV so we should assume it’s coming.

But when AI runs the morning stand-up, who is actually responsible? What’s acceptable risk when tools start shaping what teams think is useful, efficient, or true? Anyone who has managed anyone knows shared context can just as easily be shared wrongness. In this case, if an organization’s Second Brain is slightly off, everyone working from that brain drifts together.

The role of “manager” will eventually need to absorb tool audits, guardrails, incentives, pipelines, and context. That, coupled with managing the people actually using the tools, seems like conducting a drunk orchestra. A solution could be smaller department-centric stacks that optimize for each department’s unique goals. Sales does conversion, Marketing does brand, PR does comms, etc. But this structure turns into spaghetti when the tools and capabilities start blurring lines. Why does New Business need to wait for Studio if they can generate a demo reel themselves?

There's a spectrum between one "collective brain" and a chaotic "cluster of startups," and the default is somewhere in the middle. But the middle isn’t an answer. The middle needs a manager - someone auditing tools, watching the feedback loops, catching the drift before it becomes consensus. That job doesn't have a title yet.

LONG SHOWER 🚿

Solopreneurial

Jess Yan - Anthropic product lead - describes their internal setup as "a bunch of one person startups inside a large company." Her advice: skip the 20-team mega-workflow automation collective brain (for now). Explore what individuals can deliver on their own. Less dependency. Fewer impediments. No waiting on small changes. One person, a stack of agents. Individuals have leverage and responsibility. Management figures it out (…later?). (*NOTE - Anthropic is a tech company with very smart people who are comfortable pulling tickets around a complex, living product. Probably won’t work everywhere.)

Tools. Context. Presence.

Ohhhh, boy…

“It works alongside your team like any other teammate.”

“It’s connected to everything you use.”

“Anyone on the team can hand it work.”

“It’s just like working with a real employee.”

“It can even build internal tools and dashboards in seconds.”

Despite the fact the most dubious of dubious of dubious of AI Twitter are leading the charge in the comments, the end of Lindy’s launch video is relevant - “The bottleneck is no longer intelligence. It’s tools, context, presence.”

AI & The Vertical Stack

Heavy on the consultant and possibly already out of date, but some interesting organizational thinking from Contrary Research. Tools are never just tools. They contain assumptions. A calendar tool assumes time is the unit of value. A project management tool assumes progress is linear. The assumptions shape user behavior.

With AI, we’re not just adding automation or access to more knowledge - we’re adding new forms of judgment, memory, bias, and drift. Will the move from software that helps to software that fully does the work tempt us to disengage and automate our way to Wall-E? Or will organizations double down and incentivize for impact (that doesn’t feel like doing more for the sake of doing more)?

FAST-PITCH ⚾️

a/ Institute of Machine Unlearning - "It sounds serious. But it's not. But it is." Lisbon collective wants to “sneak weird things into the system…to rewire it from the inside.” They have a zine and workshops - perfect counter-programming if your team is deeply suspicious / annoyed by traditional “Learn to AI!” curricula.

b/ District Nightbourne - In 2005, Neill Blomkamp released a short called ALIVE IN JOBURG. The film’s scope and CG on a microbudget so impressed Peter Jackson that he backed Blomkamp to make a feature version - DISTRICT 9. 21 years later, Blomkamp releases NIGHTBOURNE as a test for a potential AI feature film.

c/ Handheld Camcorder - Leaping forward leapingly.

COPY PROMPT

COPY PROMPT

TOOLS TOOLS TOOLS

• Mixamo - Big library of looping character animations to use for generative prompt direction / reference.

• OpenAlternative - Open source versions of software you’re currently paying for. Fun to watch this growing “war on paid software that should be free.” What’s it like at Adobe now that Photoshop and Illustrator (Affinity), After Effects (Cavalry), and Premiere (DaVinci) are all free?

• Record-A-Skill (Claude>CoWork) - Converts your screen recording and VO explanation of a task into a repeatable automation.

"How did machines become the thing that does art, and humans are the ones in an Amazon warehouse. Like, shouldn't the robots be working so we can paint? Why aren't the robots the fucking slaves? How are humans the slaves? We should have lives of leisure with the AI. It should be the coolest thing of all time." - Stavros Halkias

Lots of pro-human sentiment at Cannes this year. But LOREM IPSUM from Purga Studio snuck in there.

You made it! All the way to the end. Thanks for stopping by. You are all wonderful people.

Thoughts, feedback, and new ideas to hello@tobinnn.com or @37click

New reader? Subscribe here.

— tobin 😊

UNIQUE SOLUTIONS

The “creative” part of the Creative Industry presumes value in finding unique solutions to a problem. The “problem” is why a brand pays an agency. Brand & agency eventually agree on an “idea” to address said problem. Everyone shakes hands and goes to a boozy dinner; where you might end up seated next to a lady who dated Jeff Goldblum in the ‘00s.

However, there’s a lot that happens from Idea to Delivered. Rounds of revision close the gap between Idea and what actually gets made. It’s an exacting and detailed process. Words like “craft” and “sculpting” confront the realities of “overage” and “timeline.” Everything is dissected, debated, polished. Down to the single frame, it will be perfect.

There’s a premium on people who can shepherd a project from Idea to Delivered. Typically, they have a British accent, a “vision,” and a shelf of awards in the background. These people make a living in the Expensive Sweater Paradigm - “I have an incredible aesthetic. Interesting friends and interests fold me in culture you can’t even see. A sense of direction defined by time and space has brought me here with you now on this Zoom call (...even though I requested we do it on Brave). It will take focused effort, treasure, and a large group of people for this vision to come out of my head, into the slide, and onto the screen. This is what you’re paying me for.”

PROBABILISTIC vs EXACTNESS

It’s one thing to open Midjourney and generate polished concepts for a deck. It’s another to lean into that same technology to actually make the thing. Generative tools are probabilistic, they don’t like to be specific. They don’t do exactness. This isn’t a defect in the tool or where we happen to be in the development cycle. It’s the nature of the technology itself - it guesses.

Using these tools, a generative pipeline becomes concentric loops of iterative approximations - prompt, that’s pretty close, prompt again, I like those two, tweak, run again, etc. If you start with a singular, inalienable vision of how the outcome must be, you’re fighting an uphill battle and missing the benefits of the technology.

HUMANS DO DEPTH. THE MACHINES DO BREADTH.

The machines are good at going wide. Lots and lots of outcomes, each with their own probabilistic idiosyncrasies. Each at a quality level from mediocre to pretty good. They can generate all day, every day. While we sleep, go to the beach, work on making something actually great.

We, humans are wonderful at depth. We have something the machines will never have - feeling. We follow intuition and taste. We drill deep, interpreting insights and exploring provocation to find truth buried in the psyche. We may use the machines to help us get off a blank page, brainstorm, filter large batches of information, but it’s not long into the creative process before we need to take over the controls.

THE GOLDEN TRIANGLE

Building a generative pipeline for a single campaign (*future posts on why this is a bad idea) won’t change the Golden Triangle equation.

Cheaper? Not really. 🤷‍♂️ In addition to the actual development, the pipeline still needs direction, curation, iteration. The human hours shift to chasing reference for training data, licensing that reference, meeting with Legal to understand why previous agreements are no longer compliant, etc.

Faster? Sometimes. 🤷‍♂️ But probably not. Building a bespoke pipeline for a single campaign is not a strategy. A compelling mood film out of Seedance might sell the idea, but it doesn’t replace the process of making the actual film or getting to the actual idea. In this sense, a generative pipeline is closer to a CG pipeline, in that the time needed to cast, tweak, style, tweak again will inevitably take longer than an experienced stylist, photographer, and creative director in a studio with racks of clothes and a few models.

Good? See previously “Humans do depth. The Machines do breadth.” 🔗

Well, then…why should we care about any of this AI nonsense??

THE HYBRID HUB

You shoot your hero campaign. Several locations, several days, one big budget. You isolate every key asset — the product, the talent, the lighting setups, the visual language.

Then, you train on those assets.

Your generative models, your inpainting workflows, your style references — all of it built from that specific shoot. Not a generic model. Not a vibe-matched prompt. A system that understands this brand, this campaign, this visual identity.

And then?

You generate. Forever. Endlessly.

Every market. Every format. Every language. Every seasonal or localized variation. With the right strategy, creative, and generative pipeline, what used to require separate shoots, separate budget conversations, separate rounds of approvals can now be rolled up into one.

The hero campaign shoot stops being a deliverable. It becomes source material.

Seek the production companies and post houses doing “...and AI.” 3D and AI. VFX and AI. A roster of talented film directors and AI.

AI won’t take the job. It will make the job bigger. That’s not a flaw, it’s a feature. As individuals and as teams, we get to make a bigger impact. This does, however, require a completely different relationship with the work - and with the client. Most agencies haven’t had that conversation yet.

LONG SHOWER 🚿

Islands of Difficulty

Next to artificial, Real Intelligence is non-linear - the weird, deep, personal knowledge and curiosity that identifies problems and solutions the machines don’t even know exist yet. What creative agencies might actually be selling isn’t content volume, nor execution, and not even craft - but the ability to locate the "island of difficulty" that a machine will never find on its own. If so, the Big Idea will only become more and more valuable as it’s the primary thing worth paying a human for.

Starving for Meaning

Slop = infinite content with nothing underneath. The antidote to slop isn't less AI, it's doing the hard human work first.

Is “Orchestrator” the Same Thing as “Conductor?”

Either the logical next step or more of the “infinite content with nothing underneath,” but more and more examples of tools dropping an LLM between prompt and generation. Unsure what model(s) to pick for a specific task? Same question multiplied across a multi-nodal pipeline? Always feel like you’re picking the most expensive route when a lesser model will do the trick? Why not let a machine with a ton more experience select the right tools for you?

Not hard to see how this could go sideways. But having a little buddy who keeps up with new releases and benchmarks, asks the right questions to better understand the task, and makes suggestions with options, would be super helpful most of the time.

FAST-PITCH ⚾️

a/ SAM SAM SAM Green screen? Blue screen? No screen. I wasn’t paying much attention and these Segment Anything Models (SAM) got really good.

Remember when rotoscoping hair was a thing?

b/ Gucci Goes Full AI. Bottega Hires Tyler the Creator to Hold a Pencil. - Vogue’s heavily-anticipated and hotly-debated “Anti-AI Slop Playbook.” Oh, the irony…the scion of the photoshopped glossy spread has the temerity to drop an all-encompassing AI playbook. And behind a paywall at that. (Jailbroken HERE.)

TL;dr - Luxury brands rely on craftsmanship to “define trends that create emotional aspiration.” AI makes content creation “lazy and artificial.” Luxury will “intentionally make mistakes” and host more IRL activations until AI “fades into the background.”

Yes, exactly. 👆

c/ Animate in AE, Render in AI - Rough motion in After Effects (…or Blender, Unity, etc) with polished concept art into prompt video generation.

Very inspired now to make these Andalucían peppers dance.

TOOLS TOOLS TOOLS

• Swishy - AI motion design. Big library of templates. Plus, upload video ref for prompt generation.

• Grainrad - “I also think some tools should just exist. Small utilities, simple fixes. Stuff that solves one problem and doesn’t need a login, a pricing tier, or a 14-day reminder email. Not everything needs to be a business. So this one’s free.” - Alim, Grainrad creator 🥲🥲🥲

• Voicebox - If you’ve actually read ElevenLabs’ T&Cs or reached out to them about an enterprise plan (€1k+ / Month?!?), you know the benefits of running an open-source voice model locally. Voicebox gathers the best text-to-speech, transcription, and language models into one platform. (You could even use it to replace your WhisperFlow.)

You made it! All the way to the end. Thanks for stopping by. You are all wonderful people.

Thoughts, feedback, and new ideas to hello@tobinnn.com or @37click

New reader? Subscribe here.

— tobin 😊

Western Union’s Africa Cup of Nations spot ran several times before I realized…

…a) I don’t like this ad, and b) it’s completely generated.

Is it the brand? Execution? Uncanny, fake human digitization masquerading as familial sentimentality and trust? Would I like it better if live-action? Or is it just annoying that I didn’t notice right away?

Conversely, I do know when I see the good stuff. And every day there is more and more generated that doesn’t have to pretend or sneak thru.

I don’t know the name for that yet - the generated next level. But it’s definitely not “slop.”

Way back in Summer ‘24, 404 Media was deep into the Slop. But only now that the concept/term landed and the analog revolution begun (again) do we have tools that properly work. The irony - just when the Zeitgeist protests the Slop, the tools are almost good enough to move us beyond. This will be 2026.

LONG SHOWER 🚿

“AI Enhanced Production” Sounds Right

AI is employed with deliberate restraint, extending the scale and possibility of the film while our focus remains exactly where it belongs: on human performance and emotionally truthful storytelling.

AI is there to sharpen the work. That might mean building background plates, scaling crowd shots, or replacing generic stock with footage that feels shot-for-purpose. But only where it elevates the idea.

…we look for ways to shift time and budget back toward the beating heart of the film - performance, story, and craft. AI allows us to handle things like product swaps without reshoots, deliver translations and localisations that feel natural, and generate additional footage that blends seamlessly with what’s already been shot.

- The Gardening Club

Pantheon

Admittedly, I got on this series from some start-up dude’s recommendation. But I think about it constantly when the conversation turns to possible outcomes.

Especially, when the old man pontificators pontificate.

Halfway through the first season, I wondered why it wasn’t live-action. But by the end of the series, the scope made it clear that animation was the only way.

If you loved the Three Body Problem, then you've probably already seen Pantheon. If you haven't read the Three Body Problem, then get on it. It's all happening, and you’re alive now for some of it.

Tom in Business Affairs Says “We Can’t.”

There are obvious licensing challenges with training materials for a brand’s generative models. (How many of your existing releases have an AI clause?) But copyright of generated assets is an equal and opposite hurdle on the other side of the pipeline.

“When images are generated by an AI model, they are not copyrighted. However, an image is not complete when it comes off the AI production line. Steps taken afterward are copyright-protected, and thus our images do have copyright. It also helps when we add brand assets for clients to the images: brand colors, logos, specific visual language. This is copyright-protected and gives our images copyright value.” - aigency Amsterdam has a plan.

“Trade secrets as watermark” will be a fun concept to explain to the team.

FAST-PITCH ⚾️

a/ WEAVY vs FLORA is the chat for many unsure how/if to tackle ComfyUI. This comparison is worth a read if graduating from Higgsfield / Freepik.

The take-away?

WEAVY = designers and creative tech.

FLORA = creative and art directors.

(Grain of salt with both and see what works? 💪 )

b/ Modular software - This seems like nothing now, but will come sooner than later. What happens when software can’t do what you want? Tell software to change itself.

c/ CATS! - How to flex on LinkedIn. More teasers and ideas. Less claiming.

TOOLS TOOLS TOOLS

• Claude Cowork - Anthropic’s agent front-end built in 10 days over the holidays is now available in the €20/month plan.

  

  Does it seem like the tools are coming faster? Yes. ✅
  Is that because the tools are building more and more of themselves? Yes. ✅
  Do these AI labs work a lot over the holidays? Yes. ✅
  
  Their personal loss is our gain. And if for nothing more than naming strategy and folder structures, this is a leap.
  
  DITs hear me now - Never forward shall…
  
  “24112025_AM_Red3_SMA_003_finalfinal.R3D”

  
  …have a home on this drive again!
  

• Replit - The app-builder IDE, now (randomly 🤷‍♂️) generates short form videos. Specifically, programmatic animations (*not generative) built with React using website animation libraries and custom Replit Agent skills. In other words, content for launch videos, explainers, boring corporate social, etc. Probably more a tool for those who make decks at Deloitte/Accenture/EY/BCG. But if you’re still in the studio director’s bad graces after what you said at the holiday party, maybe this will save time on the next B2B brief sent your way.
  

• Google Photoshoot - Stayed away from Pomelli because a) it’s a stupid name and b) the hollow experience feels like the imaginary eCommerce cupcake start-up of every financial services commercial in 2014. But if your client is the non-discerning maker of the best spicy mustard in Europe, the new Photoshoot feature can get you somewhere quickly.

  

You made it! All the way to the end. Thanks for stopping by. You are all wonderful people.

Thoughts, feedback, and new ideas to hello@tobinnn.com or @37click

New reader? Subscribe here.

— tobin 😊

At the same 1990s house party where legions of AI hustlerati scream from the roof, there is a separate, calmer conversation in the kitchen. More Rutger Bregman or Tyler Cowen, but directly related to the future of the hustle - “With all of the new tools, will the length of a ‘career’ decrease dramatically?”

It goes something like this…

STEP 1 - Quickly build your wacky idea with these wacky, new tools.

STEP 2 - Idea goes boom 💥. World sends you lots of money 💰.

STEP 3 - Eventually disintermediated, bettered, or bought.

Historically, a “successful” career spanned decades and assumed consistency plus the compounding value of experience. But, what if 12-18 months has the same outcome? Would you keep working? Move to Portugal? How does it change the way talent and organizations approach the work?

Last week, it was Claudebot Moltbot OpenClaw that went boom 💥.

Lots of people ordering Mac Minis to usher in “the next AI infrastructure boom.” The founder (@steipete) is clearly a wildly talented engineer, but…more strikingly…it’s just him.

And he built all of it “counterintuitively” - not listening to the screaming voices on the roof.

Traditionally, we/users manipulate software to reach an outcome (*Think: “anything Adobe”). OpenClaw demonstrates a paradigm shift where software evolves/changes/re-writes itself to achieve a prompted outcome/solution. We can now start from the end instead of building up from the beginning. OpenClaw is a platform that dynamically creates systems to reach solutions.

Why should we care?

What does this mean for agencies and studios? Specifically, those agencies and studios measured and valued by the strength of their unique solutions? Whose models are challenged by the increasing amount and type of solutions and deliverables needed? Well…

Is it better to be -

a) The tool that needs to be manipulated?

or…

b) The platform that builds systems?

LONG SHOWER 🚿

The Last Invention

When finally tired of the Smartless / Conan / Good Hang echo-pod-chamber, The Last Invention is a fun, documentary-style dive into the history of machine learning. Full-spectrum and comprehensive - “Where are we? How did we get here? Where are we going?” - it combines a kaleidoscope of voices: Geoffrey Hinton, Karen Hao, Nick Bostrum, Beff Jezos, Steve Bannon, Sam Harris, etc.

The New Means of Production: Creative Systems

“Agencies design creative systems. Brands operate them directly….The highest value creative work is no longer execution, but authorship of systems reflecting their creative process.” Nice post by and interview with Weber Wong, founder of Flora.

Without Bob, Sundance Just Let Anyone In This Year

DeepMind’s animated short premieres at Sundance. “...hand-painting 7,000 frames in such a complex style wasn't feasible. We had to scale the process without losing that human touch. …we customized our workflow with AI, by training the model with the art and paintings that we created. AI didn't replace the process. It unlocked it. It added consistency to the living painting look that I always dreamed of." (Deeper Dive from Googs / The Anti-Hate from Hank)

FAST-PITCH ⚾️

a/ Follow the money, see the future? Flora raises $42m. Krea raises $83m. Higgsfield raises $130m. Weavy bought by Figma. Unsure what it all means. But the tools have some money to spend.

b/ Be the best to be The Best. More GEO than production-related, but probably good to keep in mind when debating copy at 11pm in an edit suite.

c/ When will we have nothing left to cook? More skin, more details. Fake prompt —> real prompt. Internet is funny.

COPY PROMPT

TOOLS TOOLS TOOLS

• LMArena Arena AI - (All these name changes!) Compare video models with (free) text-to-video and image-to-video prompting. Benchmarks update on public leaderboard.
  

• LTX-2 Retake - Adios re-shoots, hola scope creep. Regenerate sections of existing footage. Prompt revisions in dialogue and tone/performance. Replace audio or visual independently. Why? A-B-Infinity tests for copy, TOV, and CTA. (Overview / Examples)
  

• Video Upscaler via Topaz - Seeing more and more of this as the last step in workflows. Ironically, “online editor” had a big pop on Google Trends last year.

  

You made it! All the way to the end. Thanks for stopping by. You are all wonderful people.

Thoughts, feedback, and new ideas to hello@tobinnn.com or @37click

New reader? Subscribe here.

— tobin 😊

What qualifies as “slop?”

Is slop anything generated? What if we can’t tell it’s generated?

Does slop need to be “content?” Is all slop mass-produced? Or only poorly intended?

If generated by machine, can an idea itself be slop? Does slop gain value as a meme? Or comedy?

Does the “I know it when I see it” test apply?

• (CREATIVE) Purgatory Sandwich

• (SMALL BIZ) Laziz

• (HUSTLE) Sticky Socks

• (AI TIK TOK) Reel Farm

• Shrimp Jesus - Facebook’s Algorithm Is Boosting AI Spam That Links to AI-Generated, Ad-Laden Click Farms (404 Media)

  
  

LONG SHOWER 🚿

Dieses Mädchen Needs Sticky Socks

Good ol Hustle Bros - always pushing something. Nothing new to scrape analytics and find community growth for dropship plays. But definitely novel to aggregate a community’s YT transcripts and use the machines to pull insights. German pilates ladies need grippy socks, who knew? 🤷‍♂️

One Sprite Challenge

Speaking of random insights - “True spice lovers drink Sprite.” Huh…?

Unsure exactly which shots are generated (*a very good sign), but WPP Open X and Hamlet / &Louise credited. So, seems likely they’re in there. Directors Julien & Quentin do such a good job mixing formats, moving camera, and manipulating shot sequence that anything generated not only blends smoothly, but elevates. Hypothetically, the tools change budget conversations so we can make MORE (random) creative ideas.

GenZ Will Make Their Own Software, Thank You

Curiously, the only frontier lab that doesn’t focus on image gen, Anthropic, is leading the league in Ad Campaign. Predictably, Google went aspirational-North-America-west-coast. OpenAI made 2025’s version of Chairs Are Like Facebook. But #claudepartner speaks directly to the demo who will actually drive adoption over the next decade.

FAST-PITCH ⚾️

a/ Stop Chasing AI Tools - “Your job as a creator isn't to reproduce. It's to translate. Translate your influences. Translate your obsessions. Translate your weird, specific, unjustifiable preferences into work that couldn't come from anyone else. AI is the amplifier. But what it amplifies is you.”


b/ Ohhhh gahhh, when did we start parametrizing content?! Do I need to learn how to parametrize? Remotion - Code to Video

c/ CAMERA MOVES VIA 3D RENDERS: 3Ds Max --> Kling 2.6.
Dolly In/Out ✅
Truck / Sideways Dolly ✅
Boom 🛑
Tilt 🛑
Pan 🛑

d/ PROMPT: Storyboard from a single reference image (via @firatbilal)

COPY PROMPT

TOOLS TOOLS TOOLS

• NotebookLM - Not a big secret, but surprised how few people in my world use it. Best tool out there to fast-track learning and research - about tools, cooking, permaculture…whatever.

  • FolioLM - NotebookLM Chrome extension. On the future of browsing...

  • Persona prompting inspo. (via NotebookLM 👇)
    

    
    

• DaVinci Resolve - Not specifically AI-related, but only just learned that Resolve is the “spiritual successor” to the classic Apple Pro ecosystem - FCP, Shake, Motion, Color, Soundtrack Pro. As it was before FCPX destroyed everything. Even the same keyboard shortcuts as FCP7.

• Motion Control via Kling - The amount of work that used to go into any kind of motion capture…🤦‍♂️😂

You made it! All the way to the end. Thanks for stopping by. You are all wonderful people.

Thoughts, feedback, and new ideas to hello@tobinnn.com or @37click

New reader? Subscribe here.

— tobin 😊

I know…I know. I hear you - “Not another Substack. Not another ‘voice.’ Not another newsletter to not read.” It’s exhausting. The unread in my Email > Newsletter folder is at 24,736 this morning. Which is exactly the point - we can’t keep up. It’s just not possible. The best we can do is hang on. Try not to drown while being dragged behind the speed boat.

If there’s a roadmap, I haven’t found it. I stumble upon nuggets in the noise. Like a truffle dog with dementia. The dopamine hit of head-exploding discovery after hours, days, weeks of panning in multiple streams - newsletters, podcasts, Twitter, YT, IG, TT, Discord, subReddits…a private Telegram channel by some guy named “Beech??” So many open browser tabs. Each with the possibility of gold. Sudden enlightenment from a tutorial video at 2x speed. Avoid the hot takes and final answers. Anyone claiming “expert” is lying. “We’re all apprentices,” searching for curation, validation and trust. Who is building? Sharing? Who in there is honest? These are the voices. This is the search.

You’re reading a (selfish) best effort to keep up. A strategy to process my own search and release. Why? It seems increasingly likely there will be significant outcomes from these new thinking machines. Tracking new tools, behavior change, and adoption seems one way to anticipate impact. If I force myself to communicate regularly, maybe I’ll learn? If I send energy into the space, maybe something will come back? “Effort & Learning” is the long, horizontal arrow across the bottom of this McKinsey slide.

Disclaimers…

• I am not an AI expert. But I do have nearly two decades in creative production and agency innovation across games, film, and advertising.

• I don’t have skin in the game. My AI Klout is zero.

• I will talk about tools - what they do, what they could do, and how we currently use them.

• I will talk about tools in the context of creative storytelling for brands. This will cross into “culture,” in so much as tech influences creativity which drives culture (…and brands love to borrow culture).

• I will put extra focus on how tools influence creative and business outcomes for agencies and studios.

• I will try to avoid Brand Strategy. It confuses me - like a week of class I missed at some point.

• I have never discussed the Trillion-Dollar Cluster with Dwarkesh.

• I’ve only been to SF once in the last decade - and that was for a wedding.

• I’m unconcerned if anyone likes, follows, subscribes, shares, stacks, or whatever we do with blogs. (*See “selfish best effort” above.) But definitely say hello if you end up here - hello@tobinnn.com
  

LONG SHOWER 🚿

Compositors’ World

Beyond the node-based UI, lots of generative tools/flows seem dialed for compositors, animators, mograph artists. Anyone who did time in Shake, Nuke, Flame, AE, etc. Universally, generative outputs benefit from a color pass, but what live action + AI remixes are coming our way? Shoot plates and overlay generative? Generate plates and tweak madly with traditional compositing tools? How many AE scripts are now just a prompt? Or vice-versa, how good are the machines at generating AE scripts? What would/will Chris Cunningham do?

So Much Skin

Lots of human hours going into The Great AI Skin Question. Reminiscent of game dev challenges 10-15 years ago. Without getting into science and maths (that I don’t understand), the issue is around “skin specularity” and generative models’ inability to get the specularity properly specular. In short, light passes through skin and reflects IRL in a way that digital methods struggle to replicate. We end up generating “plastic skin” which triggers the Uncanny Valley repulsion in our reptile brains.

• ENHANCOR - Advanced texture processing helps generated skin look more natural. (TUTORIAL)

  

• Skin Prompt - Skip the fancy wrapper approach and tweak your skin at prompt-level (via @Beech).

  COPY PROMPT

Dutch, Dutch, Goose

Might be location bias from living in Amsterdam, but there’s a belief that brands use the Netherlands as a testing ground for EU product launch. Something about wealthy-but-cautious consumers and Unilever. 🤷‍♂️ That said, two prominent NL brands (McD / KPN) dropped AI broadcast spots in the last month. The “public outcry” aside, it’s fair to take a beat and consider the current leap in video generation. Somehow, we already solved Will Smith eating pasta? Most relevant - there are actual, real production companies specializing in generative films. Curious how this plays out with Tech vs the growing “Real > Unreal” backlash.

McDonald’s Netherlands pulls AI Christmas ad after backlash (BBC)

KPN AI Christmas Ad - Lewis the Cat (LI)

Will Smith Eating Spaghetti 2.9 Years Later (REDDIT)

WHY 2026 COULD BE THE YEAR OF ANTI-AI MARKETING (CNN)

The Gardening.club AI

Return to Earth Studio

FAST-PITCH ⚾️

a/ Paying your media agency too much for social content? {Prompts} to create the AI influencer / girlfriend every relevant brand needs in 2026.

(👇 Could be Palafrugell. Could be Paros. But definitely nowhere. 👇)

b/ “AI tool cuts ad production costs by 30% to 50%, exec says” (The Funyun) Mondelez expects the tool will make TV spots as soon as the 2026 holiday season, “and potentially for the 2027 Super Bowl.”

c/ “Wow, these gen tool aggregators are really confusing. The models and pricing structures are all over the place. How do I know which package is right for my studio/agency/team?” Pulled a Cheat Sheet (…which will be out-of-date by next Tuesday…) of the most relevant aggregators.

View Cheat Sheet

TOOLS TOOLS TOOLS

• WisprFlow - Voice-to-text dictation that (actually) works anywhere you type type. Super fast. Fixes mistakes. Understands grammar, pauses, etc. Use it free for a month. Be astonished by how many words you didn’t type that month. Start talking to every screen you own.

  

• Keep It Shot - Renames your screenshots with a descriptive file name to make it easier to find that thing you Cmd-Shift-4’ed a few months ago.

  

• Prompt Optimizers - Constantly reminded that heavy lifting means I’m probably doing it wrong. Both OpenAI and Anthropic have prompt optimizers that elevate everything from amateur hour to semi-pro. Describe what you’re trying to do and/or include your layman’s prompt and the thinking machines come back with a wildly more expanded and detailed prompt formatted in their language.

You made it! All the way to the end. Thanks for stopping by. You are all wonderful people.

Thoughts, feedback, and new ideas to hello@tobinnn.com or @37click

New reader? Subscribe here.

— tobin 😊