The Long Game 163: Microplastics, Mold, Craft, AI Slop, Digital Hygiene, LLMs

🔙 We're back! NotebookLM, Dirt, Brain Predictions, The Score, Cousins, AI Progress, The FDE Model, and Much More!

Hi there, it’s Mehdi Yacoubi, co-founder at Mirage Metrics, and this is The Long Game Newsletter. To receive it in your inbox, subscribe here:

In this episode, we explore:

• Mold, microplastics and polyester

• Retardmaxxing

• A Missing Paradigm for LLM learning

• Craft vs slop

• The Forward Deployed model

---

🏗 Note:

It’s been a long time :)

I started a new company a few months ago, and did not have the time and mental bandwidth to keep writing this newsletter. However, I never wanted to stop it for two reasons: it helps me learn a lot, and it helps me meet great people and has led to countless positive things in my life, friends, hires, etc.

As for the new company, we’re building the AI-native Palantir focused on logistics, manufacturing, and mining. In short, we take industries that still run on manual processes and scattered systems, and give them an operating layer powered by AI agents and real-time data.

At the core is a unified data platform that pulls everything together, from sensors, ERPs, and documents to operations in the field, so decisions are made on a single, consistent source of truth. The result is fewer inefficiencies, faster decisions, and companies that can finally operate at full potential.

For the AI agents, think agents that fully automate freight forwarding paperwork and trucking operations, track and coordinate activity across industrial plants, handle order intake in large-scale manufacturing, optimize truck shifts in mining operations, or manage customs paperwork end-to-end.

We also build custom LLMs designed specifically for logistics and manufacturing teams, so operators can query and act on their data as naturally as having a conversation.

Lots of cool stuff.

If you have a company in those spaces or adjacent to them (or work at one) and suffer from similar challenges in your day-to-day, drop me a note 💌

Let’s dive in!

---

🥑 Health

🦠 Mold / Environmental Toxins / Plastics

Being honest, I’ve been less focused on health, tracking, etc., over the last 12+ months. I stopped wearing my Apple Watch, I stopped wearing my Oura, I stopped tracking everything. Even as I’m cutting right now, I don’t count calories. I’m going full vibes, leaning into auto-regulation and learning to understand the messages of the body.

I started paying attention to the messages coming from my body after my life-changing discovery of Dr Sarno (🐐) and the mindbody syndrome, which led me to understanding that overoptimisation, tracking and the metrics-oriented approach to health were fundamentally not adapted to how my brain and mind are wired.

I still think wearables and tracking have their place, especially for a few months, to learn what there is to learn, but after that, it can quickly become more noise than anything else. For now, auto-regulation and listening to the body work well for me.

Some interesting health topics that have been on my mind:

Mold:

This father shares his mold experience. Quite frightening because it can wreak your health, and if you’re not aware of the idea that it could come from mold, you could be running in circles for months/ years, not understanding what’s happening to you and your family.

So, the mold thing happened to us. And it’s been pretty bad. Been quite a process, and it’s amazing how commercial mold testing companies are totally useless. Take a moment to read this post for your family’s health. In Q4 2024 and Q1 this year, my youngest little boy was getting sick frequently. 105 fevers, respiratory infections. 7 times in about 4 months.

Polyester and fertility:

75% of female dogs wearing polyester couldn’t get pregnant because it tanked their progesterone. Polyester creates an electrostatic field that disrupts hormone production. Maybe time to switch the leggings for some cotton pants!

Microplastics in the brain:

Another alarming study on the detrimental effects of plastics on human health.

Finally, even greater accumulation of MNPs was observed in a cohort of decedent brains with documented dementia diagnosis, with notable deposition in cerebrovascular walls and immune cells. These results highlight a critical need to better understand the routes of exposure, uptake and clearance pathways and potential health consequences of plastics in human tissues, particularly in the brain.

Microplastics removal strikes me as one of the greatest opportunities in biotech over the next decade. Who’s working on this?

Pair with: Oasis App

It helps you analyze the water quality where you live and the top bottled water options.

---

🌱 Wellness

🕶 Retardmaxxing

I was coming to a similar understanding of life on my own (that you might have noticed if you’re a long-time reader of TLG), but it’s always better to put a name on it, and Elisha Long did it better than anyone else. Retardmaxxing.

A short description of what it is, and why it’s so important for you overthinkers, overoptimizers and wannabe overachievers:

High IQ often means overcomplicating everything. Always analyzing, always strategizing. I catch myself doing this all the time.

But the truth is, most high IQ people are miserable. When your brain is wired to constantly dissect, predict, and overanalyze, you see every flaw, every risk, every hidden angle.

It’s like carrying X-ray vision for reality. You can’t just enjoy the moment because you’re busy seeing how fragile everything really is. You know the odds, the statistics, the history. You can’t unsee it.

The more you know, the harder it is to laugh freely, to be lighthearted, to feel joy without caveats. That’s why so many brilliant people end up restless, cynical, even depressed.

Meanwhile, the simple ones look happier, lighter, free.

Moral of the story: You need to be retardmaxxing more often.

Pair with: Why We Prefer Complicated to Simple

---

🧠 Better Thinking

🪓 Craft Is the Antidote to Slop

I really loved this article by Will Manidis. A perfect response to Meta’s latest announcement of a dedicated AI video brainrot app.

From Genesis, man enters not a paradise without labor but a world of intentional creation. The LORD God places man in the Garden of Eden “to dress it and to keep it” (Genesis 2:15) establishing labor not as punishment but as sacred vocation. This original calling invites us to co-create the Kingdom, tending and developing the world with intention and care. Our fundamental purpose is not consumption but participation in the ongoing work of creation.

The serpent’s temptation represents the first shortcut in human history. “Ye shall be as gods” (Genesis 3:5) was not an invitation to deeper engagement with creation, but a way to get out of the work required to tend to it. The consequence wasn’t the introduction of work itself, but its corruption into burdensome toil: “In the sweat of thy face shalt thou eat bread” (Genesis 3:19). Humanity’s first sin was, in part, choosing the easy shortcut over the meaningful process - preferring effortless gain to the demanding but fulfilling work of tending the garden.

This first temptation remains alive today. Our post-enlightenment view that our world is purely material-that our lives are the outcome of physical processes devoid of feeling, craft, or meaning-is to discount the unique, historical, and stubbornly detailed nature of reality. This view misses what is evident: that the world we inhabit bears the marks of exacting, purposeful craftsmanship on both human and cosmic scales. Viewing the world as raw material is a cheap shortcut around the demanding, hermeneutic work of understanding the past and engaging deeply with the present.

The outputs of these shortcuts are “slop”— the dominant cultural output of the twenty-first century Slop emerges when we eliminate not just toil (the burdensome aspects of work) but labor itself (the meaningful human engagement with creation). Slop is production without history. Slop is detached from genuine human contribution. Slop born of effortless, replicable processes.

We see it in the overwhelming flood of Al-generated content on timelines. But slop is not just digital; it manifests physically in the uniform, history-denying aesthetic of new build transplant cities built from replicable templates, lacking any imprint of history or specific human interaction with place. It is also present in industrially farmed food that has seemingly never seen soil. This might explain why truly inspired creation, the antithesis of slop, often seems rooted in something beyond mere human effort or secular concern. Consider how no purely secular building has ever approached the grandeur of Chartres or any of the old cathedrals, and no secular music has reached the spiritual depth of Bach.

Slop, in all its forms, is the result of attempting to conjure life and meaning without labor or place in history.

Language models provide the means for industrialization of Slop. In their highest calling, these tools can eliminate genuine toil: tending to burdensome emails, litigating with your co-op board, or drafting a parking ticket appeal. But when we ask a model to write a poem, design a church, or compose a eulogy, we get something fundamentally different from human creation. The model has never lost a loved one, never stood in a holy space, never lived. We can and should automate toil, but we must preserve craft. The difference is simple: when I use a model as leverage to remove toil, I remain source; when I ask a model to remove my agency by replacing my labor, something ancient and unseen becomes source.

Demons thrive in these automated realms. Not metaphorical demons, but the actual spiritual forces that have always sought to separate humanity from meaningful engagement with creation as we see in Genesis 3. The demonic recognizes in our shortcuts the perfect opportunity: tempt humans away from the difficult labor of making, growing, and building with our own hands and minds. Instead, offer an endless stream of effortless consumption-images without artists, music without musicians, stories without storytellers. The devil’s oldest strategy is, of course, promising godlike creation without godlike effort.

Yet, the biblical narrative does not end in the Garden lost to a shortcut. It culminates in Revelation with the vision of a city, the New Jerusalem (Revelation 21-22). This city is not a return to effortless paradise but a place whose jasper walls and river-lit avenues are inconceivable without millennia of labor and discovery - the accumulated craft and effort of humanity redeemed. The nations bring their

“glory and honour into it” (Revelation 21:24), a glory that is the fruit of human labor and culture offered back to God. The perfected creation still involves ongoing fruitfulness and tending (”the tree of life... which bare twelve manner of fruits, “ Revelation 22:2). The Kingdom of God is not merely a distant future state but is, in a sense, present alongside us, requiring our active participation in building and tending it through meaningful labor. It is no coincidence that Christ was a carpenter and Moses a shepherd.

In this vision of redemptive labor, we can glimpse a more hopeful future where technology serves its highest purpose by eliminating true toil while preserving the sacred space for human hands and minds to engage in genuine craft. When automation frees us we gain capacity to redirect our energies toward the kinds of deeply human creative acts that build the Kingdom: tending gardens, raising cathedrals, composing hymns, and nurturing communities. It is a participation in the ongoing work of building the Kingdom.

Pair with: We Are the Slop

They say my generation is wasting our lives watching mindless entertainment. But I think things are worse than that. We are now turning our lives into mindless entertainment. Not just consuming slop, but becoming it.

---

🤖 AI Updates

🔍 A Missing Paradigm for LLM learning

If you use LLMs a lot in your work, like I do, you might have encountered many situations where you feel the LLM does not respond the way you would like.

Part of the issue is explained here:

We’re missing (at least one) major paradigm for LLM learning. Not sure what to call it, possibly it has a name - system prompt learning?

Pretraining is for knowledge.

Finetuning (SL/RL) is for habitual behavior.

Both of these involve a change in parameters but a lot of human learning feels more like a change in system prompt. You encounter a problem, figure something out, then “remember” something in fairly explicit terms for the next time. E.g. “It seems when I encounter this and that kind of a problem, I should try this and that kind of an approach/solution”. It feels more like taking notes for yourself, i.e. something like the “Memory” feature but not to store per-user random facts, but general/global problem solving knowledge and strategies. LLMs are quite literally like the guy in Memento, except we haven’t given them their scratchpad yet. Note that this paradigm is also significantly more powerful and data efficient because a knowledge-guided “review” stage is a significantly higher dimensional feedback channel than a reward scaler.

Pair with: Matthew McConaughey talking about wanting a private LLM and (maybe?) the solution.

---

⚡️ Startup Stuff

💭 Reflections on Palantir

Over the last few months, I’ve been reading and re-reading this piece dozens of times. I’m learning everything I can about the early days of Palantir.

As we started taking on AI projects with companies, the reasons behind the FDE model (forward-deployed engineer) became obvious.

Forward deployed

When I joined, Palantir was divided up into two types of engineers:

1. Engineers who work with customers, sometimes known as FDEs, forward deployed engineers.

2. Engineers who work on the core product team (product development - PD), and rarely go visit customers.

FDEs were typically expected to ‘go onsite’ to the customer’s offices and work from there 3-4 days per week, which meant a ton of travel. This is, and was, highly unusual for a Silicon Valley company.

There’s a lot to unpack about this model, but the key idea is that you gain intricate knowledge of business processes in difficult industries (manufacturing, healthcare, intel, aerospace, etc.) and then use that knowledge to design software that actually solves the problem. The PD engineers then ‘productize’ what the FDEs build, and – more generally – build software that provides leverage for the FDEs to do their work better and faster.

This is how much of the Foundry product took initial shape: FDEs went to customer sites, had to do a bunch of cruft work manually, and PD engineers built tools that automated the cruft work. Need to bring in data from SAP or AWS? Here’s Magritte (a data ingestion tool). Need to visualize data? Here’s Contour (a point and click visualization tool). Need to spin up a quick web app? Here’s Workshop (a Retool-like UI for making webapps). Eventually, you had a damn good set of tools clustered around the loose theme of ‘integrate data and make it useful somehow’.

At the time, it was seen as a radical step to give customers access to these tools — they weren’t in a state for that — but now this drives 50%+ of the company’s revenue, and it’s called Foundry. Viewed this way, Palantir pulled off a rare services company → product company pivot: in 2016, descriptions of it as a Silicon Valley services company were not totally off the mark, but in 2024 they are deeply off the mark, because the company successfully built an enterprise data platform using the lessons from those early years, and it shows in the gross margins - 80% gross margins in 2023. These are software margins. Compare to Accenture: 32%.

Tyler Cowen has a wonderful saying, ‘context is that which is scarce’, and you could say it’s the foundational insight of this model. Going onsite to your customers – the startup guru Steve Blank calls this “getting out of the building” – means you capture the tacit knowledge of how they work, not just the flattened ‘list of requirements’ model that enterprise software typically relies on. The company believed this to a hilarious degree: it was routine to get a call from someone and have to book a first-thing-next-morning flight to somewhere extremely random; “get on a plane first, ask questions later” was the cultural bias. This resulted in out of control travel spend for a long time — many of us ended up getting United 1K or similar — but it also meant an intense decade-long learning cycle which eventually paid off.

My first real customer engagement was with Airbus, the airplane manufacturer based in France, and I moved out to Toulouse for a year and worked in the factory alongside the manufacturing people four days a week to help build the version of our software there.

My first month in Toulouse, I couldn’t fly out of the city because the air traffic controllers were on strike every weekend. Welcome to France. (I jest - France is great. Also, Airbus planes are magnificent. It’s a truly engineering-centric company. The CEO is always a trained aeronautical engineer, not some MBA. Unlike… anyway.)

The CEO told us his biggest problem was scaling up A350 manufacturing. So we ended up building software to directly tackle that problem. I sometimes describe it as “Asana, but for building planes”. You took disparate sources of data — work orders, missing parts, quality issues (“non-conformities”) — and put them in a nice interface, with the ability to check off work and see what other teams are doing, where the parts are, what the schedule is, and so on. Allow them the ability to search (including fuzzy/semantic search) previous quality issues and see how they were addressed. These are all sort of basic software things, but you’ve seen how crappy enterprise software can be - just deploying these ‘best practice’ UIs to the real world is insanely powerful. This ended up helping to drive the A350 manufacturing surge and successfully 4x’ing the pace of manufacturing while keeping Airbus’s high standards of quality.

This made the software hard to describe concisely - it wasn’t just a database or a spreadsheet, it was an end-to-end solution to that specific problem, and to hell with generalizability. Your job was to solve the problem, and not worry about overfitting; PD’s job was to take whatever you’d built and generalize it, with the goal of selling it elsewhere.

Pair with: How Palantir Built the Ultimate Founder Factory and The FDE Playbook for AI Startups, and for my Spanish-speaking friends, El auge del Forward Deployed Engineer.

---

📚 What I Read

🧹 Digital hygiene

Almost everything you need to fix your digital hygiene (which is getting more and more important with the new type of AI-powered scams).

Every now and then I get reminded about the vast fraud apparatus of the internet, re-invigorating my pursuit of basic digital hygiene around privacy/security of day to day computing.

The sketchiness starts with major tech companies who are incentivized to build comprehensive profiles of you, to monetize it directly for advertising, or sell it off to professional data broker companies who further enrich, de-anonymize, cross-reference and resell it further.

Inevitable and regular data breaches eventually runoff and collect your information into dark web archives, feeding into a whole underground spammer / scammer industry of hacks, phishing, ransomware, credit card fraud, identity theft, etc. This guide is a collection of the most basic digital hygiene tips, starting with the most basic to a bit more niche.

📉 The Great Cousin Decline

A sad development of people having fewer kids.

Perhaps you’ve heard: Americans are having fewer children, on average, than they used to, and that has some people concerned. In the future, the elderly could outnumber the young, leaving not enough workers to pay taxes and fill jobs. Kids already have fewer siblings to grow up with, and parents have fewer kids to care for them as they age.

Oh, and people also have fewer cousins. But who’s talking about that?

Within many families—and I’m sorry to have to say this—cousins occupy a weird place. Some people are deeply close to theirs, but others see them as strangers. Some cousins live on the same block; some live on opposite sides of the world. That can all be true about any family relationship, but when it comes to this one, the spectrum stretches especially far. Despite being related by blood and commonly in the same generation, cousins can end up with completely different upbringings, class backgrounds, values, and interests. And yet, they share something rare and invaluable: They know what it’s like to be part of the same particular family.

Pair with: The parental dead end of consent morality

Humans are deeply mimetic creatures. It’s imperative that we celebrate what’s good, true, and beautiful, such that these ideals become collective markers for morality. Such that they guide behavior.

I don’t think we’ve done a good job at doing that with parenthood in the last thirty-plus years. In fact, I’d argue we’ve done just about everything to undermine the cultural appeal of the simple yet divine satisfaction of child rearing (and by extension maligned the square family unit with mom, dad, and a few kids).

Partly out of a coordinated campaign against the family unit as some sort of trad (possibly fascist!) identity marker in a long-waged culture war, but perhaps just as much out of the banal denigration of how boring and limiting it must be to carry such simple burdens as being a father or a mother in modern society.

👨‍👩‍👧‍👦 The ultimate status symbol? A big family

I generally disagree with explaining failing birthrates with the rising cost of living (just watch countries like the Nordics have similar failing rates with immense financial help from the state), but this is an interesting angle around large families as a status symbol.

What is the ultimate luxury status symbol? Once upon a time it may have been flaunting a rare handbag, sports car or flashy watch.

But with soaring costs of living and shrinking household sizes, an uncertain global future and a noisy pronatalist rhetoric, perhaps the most serious flex of wealth in developed economies in 2025 is something once considered a natural part of human life. Having kids — specifically, lots and lots of them.

“For most working parents, and particularly those living in cities, even to have one child comfortably is a major economic calculation that requires considerable financial stability,” says Eliza Filby, author of Inheritocracy: It’s Time to Talk About the Bank of Mum and Dad. While that was also true historically — when having an enormous brood was bolstered by a need for labour, religious imperatives and preserving family legacies in the face of high infant mortality — children today aren’t a source of positive household cash flow. They are a drain on it.

Pair with: The Birthgap

🚧 The #1 Bottleneck to AI Progress is Humans

I often think about this exchange; it sums up very well some of what is wrong with Europe.

Dwarkesh Patel 00:51:51

That makes you wonder, when you go around the world—because I know you go outside the Bay Area and the East Coast as well—and you talk about progress studies related ideas, what’s the biggest difference in how they’re received versus the audience here?

Tyler Cowen 00:52:02

Well, the audience here is so different. You’re the outlier place of America. And then where I normally am, outside of Washington, D.C., that’s the other outlier place. And in a way, we’re opposite outliers.

I think that’s healthy for me, both where I live and that I come here a lot and that I travel a lot. But you all are so out there in what you believe. I’m not sure where to start.

You come pretty close to thinking in terms of infinities, on the creative side and the destructive side. And no one in Washington thinks in terms of infinities; they think at the margin. Overall, I think they’re much wiser than the people here.

But I also know if everyone, or even more people, thought like the D.C. people, our world would end. We wouldn’t have growth. They’re terrible.

People in the EU are super wise. You have a meal with some sort of French person who works in Brussels—it’s very impressive. They’re cultured, they have wonderful taste, they understand all these different countries, they know something about Chinese porcelain. And if you lived in a world ruled by them, the growth rate would be negative 1%.

So there’s some way in which all these things have to balance. I think the US has done a marvelous job at that, and we need to preserve that.

What I see happening—the UK used to do a great job at it. UK, somehow the balance is out of whack, and you have too many non-growth-oriented people in the cultural mix.

🇪🇺 I try to remain a European optimist because I love to live here, but so many things have to change, fast.

🏆 The Score Takes Care of Itself, by Bill Walsh

This is one of the books I frequently come back to. I like to always have it in mind all the time because of how important the lessons I took from it are.

A short part I like:

The Gladiator Mentality: Get Your Mind Right

The gladiator mentality is common in sports, especially football at all levels. Although it’s played out differently in business, I think there is a similar phenomenon—that is, the effort to get your mind right, totally focused—before a significant event, whether it’s a major sales presentation or something else. Among other things, it involves the preparation, the “ceremony” before the main event. Top performers utilize this opportunity to get ready for battle.

There is a ritual, sort of a crescendo, that takes you to the very peak of preparation and readiness. The gladiator is thinking, mentally narrowing his focus, as he goes through the ritual before the game. It draws him upward smoothly into the increasing intensity and pressure of the event like a high-performance car going from zero to sixty, the gears shifting seamlessly and without notice.

In addition to our pregame discussions, I had my own ritual as a coach before each kickoff and did it almost unconsciously. I always went to my locker first and then walked through the locker room, taking exactly the same route each time. I would sit in my office and watch another NFL game on television for five minutes or so—not really paying much attention to it, just distracting myself. Then I would leave my office, and just before going out to the field I would shake hands with every single player on our team. If I got done and had missed one of them, I somehow knew it and would search him out and shake hands.

It was that ritual that helped me to create the mindset I wanted before each game. It helped me to focus on what I was about to do, allowed me to methodically narrow my concentration to the point where I could block out everything but the game plan and its execution. The routine was part of the grounding process in which I sought to eradicate worry, excitement, stress, distractions, hopes, fears, and all personal issues. It was like walking into a completely different room mentally, like being on a different planet. And it didn’t end when I left the locker room.

Pair with: The Upside of Stress

---

🍭 Brain Food

🧠 Your brain’s next 5 seconds, predicted by AI

This crazy paper shows that your brain states can be predicted using transformers.

The bottom line of the paper: Transformer predicts brain activity patterns 5 seconds into the future using just 21 seconds of fMRI data, and it achieves 0.997 correlation using a modified time-series Transformer architecture. Crazy!

Abstract: The human brain is a complex and highly dynamic system, and our current knowledge of its functional mechanism is still very limited.

Fortunately, with functional magnetic resonance imaging (fMRI), we can observe blood oxygen level-dependent (BOLD) changes, reflecting neural activity, to infer brain states and dynamics.

In this paper, we ask the question of whether the brain states represented by the regional brain fMRI can be predicted. Due to the success of self-attention and the transformer architecture in sequential auto-regression problems (e.g., language modelling or music generation), we explore the possibility of the use of transformers to predict human brain resting states based on the largescale high-quality fMRI data from the human connectome project (HCP). Current results have shown that our model can accurately predict the brain states up to 5.04s with the previous 21.6s.

Furthermore, even though the prediction error accumulates for the prediction of a longer time period, the generated fMRI brain states reflect the architecture of functional connectome.

These promising initial results demonstrate the possibility of developing generative models for fMRI data using self-attention that learns the functional organization of the human brain.

Pair with: A streaming brain-to-voice neuroprosthesis to restore naturalistic communication.

A woman who cannot speak now speaks through her brain, in real time, with her own voice. No typing, no delay, or sounds made. Just neural intent to streaming speech this isn’t prediction. It’s embodiment

---

🎥 What I’m Watching

DIRT Series

I love this YouTube series. The guy makes me think a little bit of Anthony Bourdain.

🎙 Herbert Allen Interview

A great interview. A few ideas from it:

Do the work you actually enjoy, not the work that looks lucrative. In the U.S., if you are competent, your needs get covered and often more. Treat losses as tuition because they teach faster than wins. Build and back things where people share real risk, take personal responsibility, and keep the business simple enough to understand on one page.

Pair with: The Consigliere

---

🔧 The Tool of the Week

📜 NotebookLM

I find a lot of people still don’t know about NoteBookLM. I find it to be the most underrated AI tool. Particularly useful if you want to quickly master a topic and have a few PDF documents on the topic.

This is also a great way to read a book or find answers to some of your challenges, using the learnings of your favorite books.

---

🪐 Quote I’m Pondering

I think of each hour spent on fitness as one day less that I’Il spend in a hospital.

— Ed Thorp

Barcelona 🫶 - hit me up for a coffee if you’re in Barcelona ☕

---

👋 EndNote

Thanks for reading,

If you like The Long Game, forward this email to someone who might enjoy it!

Until next time,

Mehdi Yacoubi