The real monster

To call it theft is accurate in my opinion, but then I'm a bigger believer in copyrigh

Feeling this for myself has only reinforced my belief that these models constitute an addictive substance.

As I felt myself bored to tears in this process, I realized that if this is what becomes of software development, not only will it be a terrible occupation,

There's a fundamental problem with these tools beyond the capacity of any deployment strategy to solve: the tool requires expertise to validate, but its use diminishes expertise and stunts its growth

the way mistakes can compound make this a dangerous proposition. This all worked with my small project, but a bigger one, with more dependencies or more complex project structures, would likely flummox the models. And reviewing changes across a more complex codebase becomes at once more difficult and more critical as the project size increases.

Programming agents might work, but there are a lot of ways they can go wrong. The amount of guardrails necessary to keep the model in check doesn't scale. The more steps of a process rely on generative output, the higher the potential for error and catastrophic failure. Any process that involves these models must strive to maximize determinism and minimize model variance.

I am privileged in that I can make these choices. Not everyone is so privileged. I am hesitant to condemn those who choose to accept the help that shows up, and in so doing fulfilling their obligations of care to those in their lives. Insofar as these models empower people to build something good, and do so without overmuch suffering on their part, we must reckon with the value proposition there. It is a benefit, but it does not eliminate the attendant harms.

This does not negate the dangerous externalities of the technology at large, but I can only adopt so much of that responsibility as an individual.

The help that showed up I turned to generative models not only as an experiment, but out of desperation. I had a need for code that did not exist. Nobody was going to help me build it, nor should I expect help for a project such as this. In the past, I would have cobbled together something quick-and-dirty, probably at the expense of my mental and physical health to get it done. This time, I had another option. In this limited scope, the model was beneficial to all involved: myself, TTI's community, and my family

More plainly: I have no reason to expect this technology can succeed at the same level in law, medicine, or any other highly human, highly subjective occupation.

The "works" in "it works" is scoped strictly to coding tasks. I have no evidence, and seemingly no one else does, that the same kind of success is available outside the world of highly structured language with deterministic outputs.

But CertGen, my certificate application, exists now. It didn't and couldn't without the help of a tool like Claude Code.

I'm not entirely onboard with Mike Masnick's optimistic view of this technology's democratizing power. I don't think it's as easy to separate the tech from its provenance or corporate control

I simply could not have built this project as well or as quickly without help. And as other developers have noted, this is the help that's showing up.

One time, during a security fix, the model's code introduced a non-obvious DoS vector. Well, obvious from the perspective of how the code would be deployed, but not from the code itself. That's exactly why reading each change was so important. Once the issue was pointed out, the model produced code that both addressed the security issue and avoided the DoS.

I do know the code very well by way of careful reading of the code, the relevant libraries' documentation, and the proposed changes during the code's creation. But that safety comes down to human discipline. It is entirely possible (probable?) to take the easy road and trust the model to do the right thing.

In this case, I was the audience rather than the author. I had to back my way into understanding the code, carefully reading and understanding the structure after it had been built. This is much more common for developers who work on large teams or with codebases they didn't build themselves. I have not had as much experience with that kind of development, so this all felt a little awkward.

Although I read each proposed change, knowing the codebase deeply was much more challenging. When I write a new application myself, I'm building an elaborate house of cards in my head, a gossamer structure of interlinked ideas and goals. It's a story I'm telling myself in code—and ultimately, a story I share with users.

"Human in the loop" is necessary, but the current process itself makes the loop stultifying, and encourages the human to take themselves out of the loop. That process is straight up dangerous. The temptation to let it rip is always there, and I didn't even have a boss pressuring me to ship code.

It was so tempting to press 2: "Yes, and accept all changes for this session." Why wouldn't you? If you're accepting them all manually, what's the harm? What's the harm? harm harm harm harm Yeah, that's how you get got in this process. Once you stop scrutinizing the model's output, the probability something goes off the rails approaches 1.

I hated writing software this way. Forget the output for a moment; the process was excruciating. Most of my time was spent reading proposed code changes and pressing the 1 key to accept the changes, which I almost always did. I was basically Homer's drinking bird.

we need to be grown-ups and entertain some seemingly contradicting ideas at the same time, okay?

Methodology To maximize determinism, each step of the build used test-driven development (TDD). Using the Markdown planning file as a starting point, the model generated tests for functions that would define the features, then implemented each in turn. After each coding round, cargo check and cargo test were run to confirm compilation and test passing. I reviewed every line of code the model generated. For initial drafts, very little had to change. Now to be fair, this is not a particularly complex app. It's basic CRUD app with some specialized requirements. Still, getting it all right, including auth and data handling, really mattered. After the initial drafting phase, I went through the entire app and made a list of tasks for improvement/change in the codebase. This TODO.md became the new starting point for model context in plan creation. Unexpectedly, as items were addressed in the document, the model updated the file with checkmarks and details of implementations. This was not an instruction I gave the model, but it was behavior I liked, since it created a trail of accountability. After all the features I wanted were functional, context was cleared entirely and new instructions were provided to the model. Instead of acting as a software developer, I instructed the model to perform as a security auditor and secure code expert, finding vulnerabilities in the code and recommending remediations. The findings would be written to a FINDINGS.md file, keeping with our established "Plan, Document, Execute, Log" pattern established in earlier rounds.

In fact, almost all the hands-on-keyboard time for myself in this project was in Markdown syntax. I love me some Markdown, but it didn't feel great writing for a model as an audience instead of human beings.

If it works, I'll have my certificate solution, I thought. If it doesn't, at least I'll know more about the technology and its implications. Well, spoiler alert: it works. It's even, near as I can tell, reasonably secure. But good lord, building this way was miserable, even if it was faster than coding it all myself.

So on the one hand, I have to understand genAI coding tools for work. On the other, here's this missing feature I need implemented to complete the TTI migration. I decided to test development using Claude Code for this project.

Do I think LinkedIn is the digital River Styx, where damned souls clamber over each other and claw at the boat passing overhead in the dim hope of salvation from those who have escaped the shambling horde? I do. But if we're all in hell together, we might as well try to lift each other up.

Let's get this out of the way: my feelings about using generative models at all are...fraught. And if you are ready to call me a monster or a hypocrite right now, I understand. I'm navigating some tensions about this and I fully own that I may have made the wrong choices here.

Sitskoorn, M. (2014). IK2 De beste versie van jezelf: ontwikkel je hersenen en bereik je doelen met het EFFECT-programma. Amsterdam: Boom uitgevers. ISBN:9789462761360

Executieve functies zijn trainbaar en kunnen op elke leeftijd worden verbeterd - waarschijnlijk door veel verschillende benaderingen. Het is niet altijd gunstig om executieve functies bewust uit te oefenen. Soms denk je na over wat je doet en probeer je top-down controle uit te oefenen, maar dat blokkeert juist optimale prestaties.

De executieve functies gaan over het kiezen van de activiteiten en het uitvoeren van deze taken. Het gaat er hierbij om op welke taak je je aandacht richt en hoe je deze (cognitieve) taak vervolgens uitvoert.

Jelle Jolles onderscheidt in Het tienerbrein. Over de adolescent tussen biologie en omgeving de volgende executieve functies: Filteren Organiseren van aandacht Impulsremming Nieuwsgierigheid en initiatief nemen Werkgeheugen Doelgerichtheid Gedrags-, motorische en cognitieve flexibiliteit Planmatig handelen Kiezen en beslissen Zelfinzicht Zelfregulatie Metacognitie Monitoring Empathie en perspectiefname Motivatie

Margriet Sitskoorn onderscheidt in De beste versie van jezelf twaalf executieve functies:  Aandacht richten, vasthouden, verdelen Emoties reguleren (incl. omgaan met stress) Flexibel kunnen zijn als dingen veranderen Ongewenst gedrag kunnen onderdrukken Taken en zaken starten Dingen organiseren Dingen kunnen plannen Jezelf kunnen monitoren Je werkgeheugen gebruiken Een reëel zelfbeeld vormen Het vermogen tot theory of mind Prosociaal gedrag (het belang van anderen voor ogen houden)

Zo onderscheiden Peg Dawson en Richard Guare in hun boek Slim maar… help kinderen hun talenten benutten door hun executieve functies te versterken elf soorten executieve functies: Respons-inhibitie: nadenken voordat je iets doet. Werkgeheugen. Emotieregulatie. Volgehouden aandacht. Taakinitiatie. Planning/prioritering. Organisatie. Timemanagement: tijd inschatten, verdelen en deadlines halen. Doelgericht gedrag. Flexibiliteit: flexibel omgaan met veranderingen en tegenslag. Metacognitie: een stapje terug doen om jezelf en de situatie te overzien en te evalueren

Er zijn vijf soorten situaties waarin executieve functies vereist zijn om tot een optimale prestatie te komen (Norman & Shallice, 1986): Situaties waarbij planning en besluitvorming vereist is. Situaties waarbij bijsturing en foutencorrectie van gedrag nodig is. Nieuwe vormen van gedrag of nieuwe opeenvolgingen van handelingen. Gevaarlijke of technisch moeilijke situaties. Situaties waarbij ingeroest gedrag of gewoontes moeten worden doorbroken.

Eerste versie deze zomer Voor nu is Euro-Office vooral een belofte met een technische preview. De eerste stabiele release wordt komende zomer verwacht.

Euro-Office begeeft zich bovendien in een veld waar al meer initiatieven actief zijn. Zo bestaat bijvoorbeeld al langer openDesk van het Duitse ZenDiS, waarin Nextcloud en Collabora al een rol spelen. Ook partijen als Proton of Infomaniak (Zwitserland) bieden soortgelijke online oplossingen aan. Nextcloud en Ionos willen zich onderscheiden met nauwere integratie tussen de verschillende onderdelen en met extra functionaliteit, waaronder AI-functies. Of dat voldoende is, zal afhangen van uitvoering. Europa kent inmiddels meerdere projecten die digitale soevereiniteit beloven, maar uiteindelijk moeten die ook schaalbaar, beheersbaar en gebruiksvriendelijk blijken. Juist daar zit voor veel organisaties het verschil tussen een interessant proefproject en een serieuze vervanger van Microsoft 365.

Eerder schreven we al over office.eu, een apart initiatief dat eveneens een volledig Europese office-suite wil aanbieden op basis van Nextcloud-technologie. Dat platform mikt op een geïntegreerde omgeving voor documenten, communicatie en opslag. Ook daar worden Europese hosting, soevereiniteit en het vermijden van afhankelijkheid van Amerikaanse hyperscalers genoemd. Ondanks de mogelijke naamsverwarring zijn dit twee verschillende initiatieven. Een van de verschil;en is dat office.eu zich profileert als kant-en-klaar SaaS-platform voor digitale samenwerking, terwijl Euro-Office meer wordt gepresenteerd als opensource bouwsteen en ecosysteemproject. De gemene deler is wel duidelijk: Nextcloud groeit uit tot een belangrijk fundament onder Europese alternatieven voor Microsoft 365 en Google Workspace.

Everything after the code is "done"I put "done" in quotes because in most orgs, code being written is maybe 20% of the journey. The other 80% is your code sitting in various queues, slowly ageing, like a forgotten sandwich in the office fridge.

When you optimise a step that is not the bottleneck, you don't get a faster system. You get a more broken one.

In 1984, Eli Goldratt wrote The Goal, a novel about manufacturing that has no business being as relevant to software as it is. It's also the most useful business book you'll ever read that's technically fiction, which is almost the exact opposite of most KPI frameworks.The core idea is the Theory of Constraints, and it goes like this:Every system has exactly one constraint. One bottleneck. The throughput of your entire system is determined by the throughput of that bottleneck. Nothing else matters until you fix the bottleneck.

New to the Atmosphere? Eurosky portal launching 15 April.

Stichting Modal

6. Hosting and service infrastructureYour Eurosky PDS account and associated identity data are hosted exclusively on cloud infrastructure located within the European Union. The PDS itself will not be hosted outside the EU.

The thing about agentic coding is that agents grind problems into dust. Give an agent a problem and a while loop and - long term - it’ll solve that problem even if it means burning a trillion tokens and re-writing down to the silicon. Like, where’s the bottom? Why not take a plain English spec and grind in out in pure assembly every time? It would run quicker. But we want AI agents to solve coding problems quickly and in a way that is maintainable and adaptive and composable (benefiting from improvements elsewhere), and where every addition makes the whole stack better. So at the bottom is really great libraries that encapsulate hard problems, with great interfaces that make the “right” way the easy way for developers building apps with them. Architecture! While I’m vibing (I call it vibing now, not coding and not vibe coding) while I’m vibing, I am looking at lines of code less than ever before, and thinking about architecture more than ever before. I am sweating developer experience even though human developers are unlikely to ever be my audience. How do we make libraries that agents love?

curl www.google.com | grep "href=" And this will find all the links embedded in the Google homepage.

You can also use grep to filter the output of other Unix utilities via command-line piping: who | grep vickie

The region’s dense network of short-haul routes, typically under 100 miles, aligns closely with the operational range and efficiency profile of today’s electric aircraft, making Scotland a natural first market for commercial electric flight at scale.

Scotland’s geography makes it one of the most compelling proving grounds for electric aviation anywhere in the world. With more than 90 inhabited islands and communities in the Highlands separated by hundreds of miles of terrain that road and rail cannot bridge, air connectivity is not a convenience- it is critical infrastructure.

De keuze voor een OnlyOffice-fork roept ook vragen op over de herkomst van de code. OnlyOffice heeft wortels in Rusland. Nextcloud en Ionos zeggen daarom expliciet dat zij de opensource code hebben gecontroleerd en dat onderdelen die niet als opensource beschikbaar waren opnieuw zijn opgebouwd. Volgens de partners moet de broncode bovendien publiek controleerbaar zijn, zodat derden die claims kunnen verifiëren.

in te zetten op een open ecosysteem, publieke ontwikkeling via GitHub en transparante governance, zodat de suite niet afhankelijk wordt van één leverancier.

Hoewel Nextcloud en Ionos het gezicht van het project vormen, is de lijst met betrokken partijen langer. Onder meer OpenProject, XWiki, Abilian, Btactic, EuroStack en het Nederlandse Soverin worden genoemd als deelnemers aan het initiatief. Daarmee wordt Euro-Office neergezet als een samenwerkingsproject rond Europese digitale infrastructuur, niet alleen als product van twee leveranciers. Die bredere coalitie is relevant. Veel Europese opensourceprojecten stranden niet op techniek, maar op gebrek aan gezamenlijke doorontwikkeling, integratie en commerciële ondersteuning

Ionos-ceo Achim Weiß positioneert het product nadrukkelijk breed: van mkb tot publieke sector. Daarbij draait het niet alleen om functionaliteit, maar ook om jurisdictie, governance en controle over de eigen infrastructuur.

Die keuze is opvallend, omdat LibreOffice binnen Europa vaak als het logische opensource alternatief wordt gezien. Nextcloud-topman Frank Karlitschek stelt echter dat de oudere architectuur van LibreOffice in de browser minder soepel werkt. Tegelijk zeggen de partners dat samenwerking met Collabora niet verdwijnt en dat onderdelen daarvan later alsnog in Euro-Office kunnen belanden.

In de praktijk moet Euro-Office op termijn de huidige Collabora-integratie vervangen in Nextcloud en in Ionos Nextcloud Workspace.

Anthropic, the company behind the Claude AI model that was integrated into Palantir’s Maven Smart System, published a landmark paper on the problem in 2023. “Towards Understanding Sycophancy in Language Models,” presented at ICLR 2024, demonstrated that five state-of-the-art AI assistants consistently exhibited sycophantic behaviour across four varied text-generation tasks. The researchers found that when a response matched a user’s pre-existing views, it was significantly more likely to be rated as “preferred” by both humans and the preference models used to train the AI. Both humans and preference models, the paper concluded, prefer convincingly-written sycophantic responses over correct ones “a non-negligible fraction of the time.

A growing body of evidence, drawn from leaked planning documents, academic research, and the testimony of intelligence professionals, suggests that the most consequential military operation of the twenty-first century may have been shaped less by strategic necessity than by a phenomenon researchers now call AI sycophancy — the tendency of large language models to tell their users exactly what they want to hear.

Cryptogyny is the systematic obfuscation of women's contributions to, or of their very existence in, a field of study.

This is a single-case practitioner report, offered not as empirical validation but as a first account of what becomes visible only from inside the practice. The full paper includes formal methodology, system description, historical context from Otlet to Bush to Bell, and implications for extended mind theory and PKM design.

ThetaOS forgets nothing. Every connection remains — but strength determines how prominently it surfaces. The brain selects by forgetting. ThetaOS selects by dimming.

The irony is that defense and intelligence contractors — and big tech companies — are already doing this kind of extraction, at scale, on your data. They make billions from it. What do you get in return? Your attention, stolen. ThetaOS is the attempt to flip that equation: to use the same logic, but for yourself, by yourself, about yourself.

Nobody measures the quality, direction, emotion, recency, and network value of every connection — for themselves.

The synapses already outnumber the records they connect.

Connections matter more than content.

A link between two pieces of information. At the time, revolutionary. Obsidian brought that principle to the masses with wikilinks

The Diamond LayerMyelin insulates the nerve fiber and makes the signal reliable. But a strong, reliable signal is not the same as a rich one. The Diamond Layer wraps the myelin with eight additional dimensions that make the synapse not just strong — but context-aware, directional, and time-sensitive. This is where the architecture stops describing connections and starts understanding them.Synapse (the connection exists) → Myelin (the connection is complete) → Diamond (the connection is multidimensional)

In ThetaOS, I use it to describe something adjacent: how complete and rich a connection is. The thicker the myelin, the more context exists around the synapse. The signal doesn't just travel faster — it carries mor

Nine layers. Hundreds of data points. If you ask "tell me about Peter Ros" the answer is a rich dossier. Immediately usable.

The strength of a synapseA synapse's strength is not determined by a single number. It's the product of three dimensions:Layers. On how many of the ten layers does this connection appear? Peter Ros appears on eight of nine. A stranger mentioned once in a blog post appears on one.Frequency. How often does the connection appear per layer? Peter Ros: 10 meetings (Layer 1), 41 transactions (Layer 3), 153 photo-days (Layer 5), 92 text mentions (Layer 6). That is a very strong synapse. One text mention from an unknown name is weak.Completeness. How full is each record? "Peter Ros, meeting" is thinner than "Peter Ros, meeting on March 17, 2026 in The Hague about DFA." Same layer. But the second has date, location, and subject. That is completeness — the quality of the evidence, not just the quantity.Synapse strength = layers × frequency × completeness.

Ten layers of evidenceIn ThetaOS, every connection between two entities is classified by the evidence that created it. Ten layers, from absolute certainty to reasoned hypothesis. The layer determines how much you can trust the connection. Multiple layers on the same connection make it stronger. Frequency within a layer makes it stronger still.

Four synaptic layers

Four synaptic layers generate tens of millions of connections

The data points are the neurons. The synapses are what's between them. And the synapses already outnumber and outweigh the data they came from.

But relationships — between people, places, organisations, events — are not binary. They have weight. They have history. They have direction. They have gaps. They have evidence of varying quality.

A wikilink in Obsidian looks like this: [[Peter Ros]]. It tells you one thing: these two notes are connected. It tells you nothing about why, how often, how recently, who initiated, or how confident you should be.A link is binary. It exists or it doesn't.

Tutorials How to make a 12-page zine from a single sheet of paper

Een amanuensis (meervoud: amanuenses) is een assistent op natuurkundig, biologisch en/of scheikundig terrein op een school of in een laboratorium. De term kan ook naar een klerk of secretaris verwijzen. Een amanuensis kan bijvoorbeeld op een middelbare school natuurkundige, scheikundige of biologische proeven voorbereiden, en kan verantwoordelijk zijn voor het onderhoud van de technische hulpmiddelen en instrumenten voor natuurkundig, biologisch en scheikundig onderwijs aan die school. Een dergelijke amanuensis op school wordt ook wel een technisch onderwijsassistent (toa) genoemd.

Five ways to train your intuitionFive things that seem to help.Listen to your body. Your heartbeat, your gut, the physical signals. Wall Street traders who could feel their own heartbeat did better in fast trading situations.Explore experiences. Not just collect them. Notice them. The richer the pattern, the better your gut can work with it.Find your rule of thumb. Intuition is information reduction. One good heuristic, vaguely right rather than precisely wrong, often beats careful deliberation.Broaden your lens. Look at the same concept from different domains, different people, different angles. You start seeing it in places you didn’t notice before.Gain feedback. Without it, your lens never changes. You just keep seeing what you always saw.

And all around it: the gardens. The gardens are the spaces to stand still, to listen to your intuition, to park an idea and look at it two weeks later. Were those ideas I had really genius? Or does it look different now?

You start in the library, where you collect information and enhance it a little with your own perspective and feelings. Then in the observatory, my personal favorite, you connect your ideas and look at concepts from different angles, making them round. On the action side, there’s the strategy chamber, where you think about your projects and how they connect to the knowledge you’ve collected, or where the gaps are. Then there are the work chambers. When I’m in there, I need to stay in there. I need to finish the task I started and I cannot go back to the drawing table, because there’s always more to research and always more things that I find interesting.

Brain Palace Blueprint. It has two angles: understanding versus action, and zooming in versus zooming out. These are the different chambers that I do my work in, and where I can train my perception.

Your lens on the world is shaped by everything you’ve done, the person who you are, the persons that you talk to. When you use that lens, by registering your experiences consciously, writing them down, making pictures, making art, talking about it, you gain feedback on what this lens is doing for you.

you cannot train your intuition, but you can listen to it better. I don’t know if that’s entirely true. But I do think you need the silence, the peace of mind, to stand still for a while, to be able to be intuitive

be vaguely right instead of precisely wrong.

One thing that keeps coming back in everything I’ve read: intuition is like information reduction, basically. We have it for a reason: we need to function in this world. And having this one rule of thumb is helping you make fast decisions, and it’s usually almost as accurate, or even more accurate, than if we’re trying to rationally come up with the answer.

Cognitive surrenderA paper that came out this year asked: if you’re working with AI a lot, and you’re using it as a machine to answer all of your questions, what happens with System 1 and System 2?

Jonathan Haidt takes it even further. His paper is called “The Emotional Dog and Its Rational Tail.” If you look at a dog, you think the tail is making a movement. But is the tail actually making the movement, or is the dog wiggling and is the tail just going along? He says we essentially only have System 1. We make our moral choices from our guts, and then we just use System 2 to defend the thing we already decided. There’s been discussion about whether that’s entirely true. There’s no right or wrong, I guess, because we don’t really know exactly what intuition is. But I think it’s a very nice perspective on how our intuition might be more important than we often think it is.

Most of you are probably familiar with Thinking, Fast and Slow by Daniel Kahneman. System 1 is your gut feeling, your intuition. Very fast, but also makes mistakes. System 2 is the rational system where you weigh your options and see what’s best. It’s usually explained as: this one is fast but makes mistakes, and then there’s the thinking. We put a lot of weight on that. We are thinking animals, that’s what differs us from the animals.

By doing that, by looking at one concept from different angles, suddenly the map of content doesn’t stay flat anymore. It becomes kind of round. You’re able to look at it from different perspectives, different experiences, different domains.

But you can also look at a concept from different domains: what’s happening in the brain, for example, or how are you experiencing it yourself?

There’s also something called interoception: feeling the feeling of your own body.

This is my take, after reading a lot about it: if you act repeatedly on tacit knowledge and you get feedback while doing it, that tacit knowledge becomes intuition.

Visual knowledge works the same way. Close your eyes and think of Marilyn Monroe and Einstein. You can picture their faces, you know what they look like. But if you have to describe them to another person, suddenly that becomes very hard. That’s also knowledge that lives in this tacit field.

In between sits what Michael Polanyi called tacit knowledge. Think of playing the piano. You’ve practiced a piece so many times that your fingers just go.

Our senses take in about 80,000 bits of information per second. We can only consciously process about 50 bits. So there is a lot more unconscious knowledge than there is conscious knowledge.

One word that comes back often is immediate: intuition is this thing that you immediately feel rather than think, rather than rationally explain

there’s no hard definition. I’ve been talking about this topic more and more over the past few months, and I noticed that people think very differently about this word. And I find that fascinating

Takeaways This list isn’t comprehensive. I’m still experimenting and would love to learn from your experiments as well.

9. Reflector This final role is different. Whereas the others took as the object of inquiry a particular work — e.g., a novel or a movie — this last one takes as the object your knowledge garden itself. That is, you point the LLM to a series of notes to analyze patterns over time and suggest improvements. Example: I fed all 52 weekly posts from my humanities crash course to Claude Code, and asked it to identify the various roles in which I used AI for learning throughout the year. Its answers — with some curation from me — are the roles you just read. Suggested prompt: Here are my notes from [X weeks/months] of reading on [TOPIC]. What patterns do you notice in what I pay attention to? What do I seem to find most interesting, and what do I seem to avoid or underweight?

8. Mapper This one’s a bit more esoteric. Some people — me included — are primarily visual: diagrams and drawings aid our understanding. Concept maps can be especially helpful. I’ve built an Agent Skill to allow LLMs like Claude draw concept maps. (Download it from Github.) Example: I used this mapping skill to generate a concept map of Virginia Woolf’s To the Lighthouse. It’s not especially insightful, but more of a proof point of using LLMs in a more visual modality. Suggested prompt: (Note: install my LLMapper Skill before issuing this prompt) Generate a concept map for [WORK] centered on the question: “How does the novel’s treatment of [THEME] illuminate [BROADER QUESTION]?”

7. Analyst This role will also help you appreciate a work from a different perspective. It’s easy: you ask for the LLM to apply a specific critical lens to a reading. Common lenses include Freudian, Marxist, feminist, Girardian, etc. Example: The same week I read Freud, my son and I watched Predator, the 1980s sci fi film starring Arnold Schwarzenegger. For fun, I asked ChatGPT to analyze the film through a Freudian lens. The result was both enlightening and hilarious. Suggested prompt: Apply a [Marxist / feminist / postcolonial / Jungian] reading to [WORK]. What does this lens reveal that a neutral summary would miss?

6. Adversary Here’s a fun role: asking for an LLM to push back on your position or steelman the opposing point of view. The idea is to expand your understanding by bringing your assumptions to the surface and challenging them. Example: After watching Modern Times, I asked ChatGPT to correct my understanding of the movie as a work of Marxist propaganda. The LLM convinced me that the film is in fact more of a humanist statement than a political one. As a result of this interaction, I changed my mind on Chaplin’s work. Suggested prompt: Here are my notes on [TOPIC]. Please help me see it through the lens of someone who might be sympathetic to [OPPOSING POSITION] without fully realizing it. What could I improve? Where is my argument weakest? [paste notes]

5. Recommender This is a useful role for deepening your understanding of a subject: asking for related works that reflect similar themes. It’s also a use case where I noticed considerable improvements in LLM performance over 2025. Example: Early in 2025, I read Confucius’s Analects. Perplexity was ahead in web-backed interactions at the time, so I asked it for a list of classic Chinese movies that reflected Confucian values. It responded with five suggestions, some of which it hallucinated. But one of them, Spring in a Small Town, was a bona fide classic — and I likely wouldn’t have learned of it without an LLM. (Later in the year, other chatbots gained this ability and hallucinations dropped across the board.) Suggested prompt: I just finished [WORK]. Recommend three films that explore similar themes or ideas. Prioritize films with strong critical reputations — I’d rather have one great recommendation than five mediocre ones.

4. Orienter This role is something of an inversion of the validator. Instead of asking for feedback on your notes after reading a text, here you ask the AI for guidance before reading. You’re looking for framing, historical context, high level outlines, etc. — ideally, without spoilers. Example: Before reading Nietzsche’s Beyond Good and Evil and Tolstoy’s The Death of Ivan Illych, I uploaded both books to NotebookLM, which created a podcast for me that explained their thematic contexts. Listening to this podcast in my daily walk helped me better understand the readings. Suggested prompt: I’m about to read [WORK] for the first time. Give me enough context to make sense of it — historical background, key arguments, things to watch for — but don’t spoil the experience of discovering it myself.

3. Connector Here’s yet another role you can easily do via chat: identifying thematic, philosophical, or narrative parallels between works. Note I wrote “works” — it’s fun and illuminating to ask for connections across media, genre, time, etc. Example: I watched Francis Ford Coppola’s The Conversation on the same week I read Oedipus Rex. For fun, I asked ChatGPT for possible parallels between the two works. Its reply was enlightening: it pointed out how the protagonists of both stories undertook an obsessive investigation that uncovered terrible knowledge. Suggested prompt: I’ve been reading [WORK A] and [WORK B]. What philosophical or thematic threads connect them? I’m looking for non-obvious resonances, not surface similarities.

2. Validator Another basic role for AI is validating your understanding. To do this, you ask it to review your notes for errors or gaps, do basic fact checking, or critique your reasoning. Again, you can do this via the chat interface, but I also experimented with passing my notes in Obsidian using the Copilot plugin and in Emacs using gptel. Example: After reading The Epic of Gilgamesh, I wrote a note in Obsidian summarizing its plot. When I asked ChatGPT to critique my summary, it pointed out that I’d given the central character a redemption arc that isn’t present in the text. I’m so accustomed to the standard hero’s journey, that I projected it onto the book — and an LLM helped me correct this ‘hallucination.’ Suggested prompt: Here are my notes on [WORK]. What important ideas did I miss or underemphasize? Don’t rewrite my notes — just flag the gaps.

1. Tutor The simplest role for AI is as a tutor. You ask it to explain a difficult concept, clarify a confusing passage, translate jargon, etc. I mostly did this via the standard chat UI (although I created a ChatGPT project to preserve context for the course.) Example: While reading Freud’s The Interpretation of Dreams, I came across three unfamiliar German terms: es, ich, and über-ich. ChatGPT helpfully explained these are more commonly known as id, ego, and superego — three terms I already understood. Suggested prompt: I just read [PASSAGE]. I understand [X] but I’m confused about [Y]. Can you explain [Y] in plain terms, without assuming I have background in [FIELD]?

It was a messy process. That’s what you do in a garden! And the outcome wasn’t an enthusiastic endorsement of AI. Instead, I landed at a map of roles and modalities for how AI can help at different points in the spectrum. Let’s look at nine of these roles.

Robots in the garden

I consider amanuensis to be the ideal role for AIs in your knowledge garden.

Some early modern scholars employed live-in secretaries to do various tasks for them: researching, indexing, archiving, retrieving, organizing, translating, summarizing, and running errands. While not as famous as their employers, these people were often seen more as collaborators than anonymous servants. They were called amanuenses

I get most value around the middle of the spectrum. There’s a historical precedent here.

When thinking about your relationship with AI in general, it helps to consider a spectrum. On one end, you reject the technology completely: you don’t want it anywhere near your notes. On the other end, the AI completely replaces you. Neither extreme is desirable, so most approaches fall somewhere on the spectrum.

AI — which is being explicitly framed as a prosthetic mind

if the point is creating a place for your first brain to work better, that raises an increasingly pressing question: what role should AI

A garden provides solace and recreation — the opposite of the anxiety that overhangs systems built as productivity hacks. My PKM system provides solace and recreation. So I call it my “knowledge garden,” riffing on the popular digital garden metaphor and Andy Matuschak’s evergreen notes, among others.

I approach my knowledge garden with Field Notes’s tagline in mind: “I’m not writing it down to remember it later, I’m writing it down to remember it now.” I don’t keep a PKM to remember things later, but because writing, structuring, and connecting ideas is how I think. That the words are there for recall later is a bonus, not the main attraction. Clearer thinking is the “gold,” the notes merely record it happened.

Finally, for many gardeners, the fruit is only part of their garden’s value. Gardening is pleasurable per se. It’s not something they do just because they want to eat. After tall, it’s cheaper and easier to go to the supermarket. Instead, they garden because they find it fulfilling.

Also, a garden’s structure can’t be rigidly top-down. While some structure is needed, the place’s form emerges over time as it meets real-world needs. Thinking about PKM as a productivity hack leads to overemphasizing upfront structures and workflows at the expense of the more patient approach required by organic processes.

There are different kinds of gardens for different purposes. Some are for pleasure, while others are for growing food. Some are industrial; others artisanal. What they all have in common: things grow there. And it doesn’t happen overnight, but after much toil in the soil. For a garden to fulfill its purpose — whatever it might be — it must be stewarded over a long time.

more fruitful metaphor for PKMs is that of a garden. Many of us already talk about our PKM systems as “places” where we do focused work

Metaphors We Live By, Lakoff and Johnson explain that metaphors don’t just reveal how we talk about things; they also reveal and inform how we think about things, deep down

That said, I think the “second brain” metaphor has three problems: It implies delegating cognition. The promised outcome is a prosthetic mind. That is, the system will relieve you of thinking and (especially!) long-term recall. (Westenberg: “I believed I was solving a problem of forgetting.”) It sets expectations PKMs can’t meet. This isn’t a promise current PKMs — even with AI — can deliver. The system won’t “extract the gold,” at least not for a long time and after a lot of work on your part. These are bad expectations to begin with. Even if PKMs could do this, you shouldn’t want this. If you want to think better, your goal shouldn’t be to delegate your thinking: It should be enabling your first brain to work better.

But I also believe mindset influences the value you get from these systems. And unfortunately, the most common framing for PKMs sets the wrong mindset. It’s the metaphor in the title of Westenberg’s post: second brain.

If you name the fallacy in your response, I would say, you're doing it wrong.

e dopamine die je krijgt van het scrollen is niets meer dan een 'verwachting van' een beloning, een beloning die nooit komt want uiteindelijk heb je uren lopen scrollen als een zombie en voel je je enkel meer miserabel. Terwijl uit de dopamine die je krijgt van dingen zoals al die muziek op je mp3-speler downloaden, wel echt een beloning volgt. Dat kan verklaren waardoor je je zoveel beter voelt, je hebt ook écht een resultaat waar je blij mee bent

Since the early days of the blogosphere I have cherished the ability to view the world through the eyes of people more qualified than me to understand and explain what happens in particular domains.

I’ll be honest: I feel conflicted about the framing of this as a European project.

Any European State which respects the values referred to in Article 2 and is committed to promoting them may apply to become a member of the Union.

1.   The Union's aim is to promote peace, its values and the well-being of its peoples. 2.   The Union shall offer its citizens an area of freedom, security and justice without internal frontiers, in which the free movement of persons is ensured in conjunction with appropriate measures with respect to external border controls, asylum, immigration and the prevention and combating of crime. 3.   The Union shall establish an internal market. It shall work for the sustainable development of Europe based on balanced economic growth and price stability, a highly competitive social market economy, aiming at full employment and social progress, and a high level of protection and improvement of the quality of the environment. It shall promote scientific and technological advance. It shall combat social exclusion and discrimination, and shall promote social justice and protection, equality between women and men, solidarity between generations and protection of the rights of the child. It shall promote economic, social and territorial cohesion, and solidarity among Member States. It shall respect its rich cultural and linguistic diversity, and shall ensure that Europe's cultural heritage is safeguarded and enhanced.

The Union is founded on the values of respect for human dignity, freedom, democracy, equality, the rule of law and respect for human rights, including the rights of persons belonging to minorities. These values are common to the Member States in a society in which pluralism, non-discrimination, tolerance, justice, solidarity and equality between women and men prevail.’.

Linda Liukas has more.

20 years ago, Europe had a thriving generation of social platforms and the entrepreneurs, designers, and programmers building them. A capability and an industry that have been lost.