Bookmarks

The biggest opportunities right now:

1. build for solving loneliness (the more AI floods everything, the more people crave real human connection, IRL and small social)

2. build for agents that need to spend money (they're getting virtual cards and budgets, someone builds the spend controls, fraud protection, receipts)

3. build for people drowning in AI output (everyone generates infinite drafts now, the bottleneck moved to reviewing and choosing, build the judgment layer)

4. build for the burnout economy (everyone is expected to always be on and always optimizing, and the backlash toward rest, slowness, and enough is building)

5. build for verifying humans (deepfakes broke trust, every dating app, marketplace, and video call needs proof-of-human within 2 years)

6. build for the physical world (the trades, hardware, robots that AI is finally reaching)

7. build for the agent that answers the phone (every local business misses calls after 5pm, a voice agent that books the job is worth thousands a month)

8. build for the aging (70M+ boomers who want to stay healthy, sharp, and connected)

9. build for the LLM-search land grab (being the cited answer is the new SEO)

10. build for the newly automated (the paralegal, the analyst, the marketer whose job just changed and needs to reskill fast)

11. build for the seat-pricing collapse (software repricing from $50/seat to per-outcome, whoever nails outcome billing wins a category)

12. build for AI enablement (95% of businesses use nothing beyond ChatGPT, someone has to onboard the other 95%)

13. build for the agency everyone resents (businesses pay $1k/mo to agencies they hate, an agent that does 80% of it undercuts the model)

14. build for verticals on 2011 software (dentists, HOAs, contractors, all overdue for an AI-native rebuild)

15. build for reviving dead software (thousands of abandoned apps with real users, agents can maintain what a team couldn't, buy and revive)

16. build for markets too small to matter before (500 lobster fishermen was never worth a team, now it's a weekend and a real business)

17. build for agents hiring agents (a shadow economy is forming, it needs escrow, reputation, and dispute resolution for machines)

18. build for the anti-AI premium (as everything gets generated, human-made and analog become status symbols people pay up for)

19. build for distribution-first (anyone can build the product now, so the audience is the moat, media company first, product second)

20. build for the reinvention of college (what does an MBA even mean anymore)

21. build for a world with more free time than it knows what to do with (if AI takes the busywork, the question becomes what people do with the hours, and that's a civilization sized market)

note: more trends/ideas @ideabrowser (free to sign up)

22. build for the return to the physical (screens fill with slop, people crave the real world, the hands-on, the local, the analog)

23. build for the caregiving wave. The population is aging fast, tens of millions are caring for parents, and the whole burden is landing on families with no support.

24. build for the longevity shift (people want to live to 100 healthy, and a whole industry is forming around actively managing your own biology)

25. build for the housing and rootlessness problem (people can't afford to settle down, and the whole idea of a stable home base is up for grabs)

26. build for spiritual hunger (as institutions hollow out, the need for meaning, ritual, and belonging is exploding into new forms)

KEEP BUILDING

The most influential mathematician of his generation stands in a German lecture hall and explains a foundation he is rebuilding from scratch. Almost nobody watches it.

This is Peter Scholze at Bielefeld University, November 2025, opening lecture of a new public series called Ars Mathematica.

Scholze became a full professor at 24 and won the Fields Medal at 30. His work on perfectoid spaces reorganized entire areas of arithmetic geometry inside a decade.

Condensed mathematics is his attempt to fix something deeper: the way analysis and algebra refuse to sit together properly. He is proposing new foundations for how mathematical objects carry topology.

Watch how he pitches it to a general audience. No prerequisites assumed, no dilution, a working mathematician showing why abstraction opens rooms you could not otherwise enter.

A researcher I know rewatched the middle section twice and said it was the first time condensed mathematics felt like an idea rather than a rumour.

Free on YouTube from a university channel, subtitles on.

Some people prove theorems. He is replacing the floor.

Der Montagmorgen danach: CSD Berlin, Abdul Ballout – und das internationale Netzwerk hinter dem „Stuhlkreis“

Nach dem mutmaßlichen Anschlag auf den Berliner CSD richtet sich der Blick auf das Violence Prevention Network (VPN).

Abdul Ballout sollte nach seiner Verurteilung an 26 Deradikalisierungssitzungen teilnehmen. Vor der Tat hatten erst zwei Gespräche stattgefunden.

VPN-Geschäftsführer Thomas Mücke erklärte, sein Team habe Ballout als gefasst, vorsichtig, kontrolliert und freundlich erlebt – aber keinen wirklichen Zugang zu ihm gefunden.

Wer glaubt, dahinter stehe ein kleiner lokaler „Stuhlkreis“, dürfte überrascht sein:

Das Berliner Violence Prevention Network verfügte 2024 über ein Gesamtbudget von 11.063.726,56 Euro, beschäftigte 193 Personen und betrieb 44 Projekte. Ausgewiesen werden außerdem 135 Distanzierungsbegleitungen, 363 Gruppen- und Einzelmaßnahmen im Justizvollzug sowie insgesamt 9.284 erreichte Personen.

Von den Einnahmen stammten laut Jahresbericht rund 8,778 Millionen Euro aus öffentlichen Zuwendungen und weitere 1,524 Millionen Euro aus öffentlichen Aufträgen. Das sind zusammen etwa 93 Prozent des Gesamtbudgets aus öffentlicher Hand.

Größter Geldgeber war das Bundesfamilienministerium mit 4,903 Millionen Euro, gefolgt von Berlin, Hessen und Bayern.

Doch das Netzwerk reicht weit über deutsche Gerichte, Justizvollzugsanstalten und Ministerien hinaus.

VPN führt in seinem eigenen Jahresbericht eine Violence Prevention Network USA, Inc. in Pittsburgh, Pennsylvania als Auslandsstruktur. Die amerikanische Organisation wurde im Juli 2022 gegründet, nahm im März 2023 ihre Tätigkeit auf und ist als steuerbefreite 501(c)(3)-Organisation unter der EIN 88-3302794registriert.

Die US-Finanzunterlagen für 2023 nennen Einnahmen von insgesamt 575.369 US-Dollar. Davon kamen:

▪️ 473.000 US-Dollar von Rockefeller Philanthropy Advisors

und

▪️ 102.369 US-Dollar von der Berliner Violence Prevention Network gGmbH.

Die US-Organisation gab 2023 rund 197.875 US-Dollar aus und wies zum Jahresende ein Nettovermögen von 377.494 US-Dollar aus.

Damit nicht genug:

Das US Department of Homeland Security führt Violence Prevention Network USA für das Haushaltsjahr 2024 mit einer Förderung von

817.465 US-Dollar

im Programm des Center for Prevention Programs and Partnerships – CP3.

Auch die Berliner gGmbH erhielt direkt Geld aus Washington: Das US Department of State vergab 2024 einen Zuschuss von 12.200 US-Dollar für die Implementierungsphase des Projekts VIVA – Verschwörungsdenken individuell verstehen und auffangen, das sich mit Verschwörungserzählungen und Desinformation befasst.

Der internationale Fachbereich des VPN beschreibt außerdem:

– eine Unterstützung durch Public Safety Canada bei einem „Diversions Playbook“ als Grundlage für einen strategischen Zehnjahresplan, der Technologieplattformen in Online-Interventionen einbeziehen soll;

– eine ausdrückliche Projektpartnerschaft mit Meta;

– die Mitarbeit im Global Internet Forum to Counter Terrorism – GIFCT;

– sowie Projekte zur KI-gestützten Erkennung von Hate Speech und zur Schulung von Strafverfolgungsbehörden und zivilgesellschaftlichen Organisationen.

Jonathan Russell, Leiter des internationalen VPN-Fachbereichs, war 2024 stellvertretender Vorsitzender des unabhängigen GIFCT-Beratungsgremiums und wurde 2025 dessen Vorsitzender.

Das Gremium berät das globale Technologie- und Plattformnetzwerk strategisch bei der Bekämpfung terroristischer und extremistischer Internetnutzung.

Wichtig: Bislang gibt es keinen Beleg dafür, dass Rockefeller Philanthropy Advisors, Homeland Security, das US State Department, Meta, GIFCT oder VPN USA am konkreten Fall Abdul Ballout beteiligt waren.

Ebenso ist nicht belegt, dass Daten aus seinem Fall ins Ausland gelangten.

Aber genau deshalb stellen sich nach einem tödlichen Anschlag zwingende Fragen:

▪️ Welches konkrete VPN-Projekt betreute Ballout?

▪️ Wer bezahlte diese gerichtliche Maßnahme?

▪️ Welche Risiko- und Verlaufsberichte wurden erstellt?

▪️ Wem wurde gemeldet, dass die Berater nach zwei Gesprächen keinen Zugang zu ihm fanden?

▪️ Welche Informationen gingen an Gericht, Jugendgerichtshilfe, Bewährungshilfe oder Sicherheitsbehörden?

▪️ Welche Daten werden im Rahmen der „Sozialen Diagnostik“ erhoben, gespeichert und ausgewertet?

▪️ Und wie werden internationale Geldgeber, Technologieplattformen und ausländische Regierungsprogramme organisatorisch und datenschutzrechtlich von deutschen Einzelfällen getrennt?

Hinter den zwei Gesprächen mit Abdul Ballout steht jedenfalls kein kleiner Berliner Gesprächskreis, sondern eine überwiegend staatlich finanzierte, international vernetzte Organisation mit eigener US-Struktur, Rockefeller-Geld, Homeland-Security-Förderung, State-Department-Projekten und direktem Zugang zu Big-Tech-Gremien.

Transparenz über Finanzierung, Methoden, Datenflüsse, Berichtspflichten und Verantwortlichkeiten ist hier kein „Verschwörungsdenken“.

Sie ist nach diesem Behördenversagen das Mindeste.

Auszug aus der Website der NGO, bei der Abdul B. die gerichtlich auferlegten Beratungsspräche zur Deradikalisierung führen hätte sollen.
Rund 11 Millionen Euro an Fördergeldern erhielt die gemeinnützige GmbH im Jahr 2024.

"Violence Prevention Network gGmbH ist eine NGO, die seit über zwei Jahrzehnten mit Erfolg in der Extremismusprävention sowie der Deradikalisierung und Distanzierungsbegleitung tätig ist. Durch die Arbeit mit rechtsextremistisch und islamistisch gefährdeten sowie radikalisierten Menschen hat sich das Team von Violence Prevention Network über die Jahre eine international anerkannte Expertise im Bereich der Arbeit mit ideologisch motivierten Straftäter*innen erworben.

Das Ziel von Violence Prevention Network ist es, dass ideologisch gefährdete Menschen und extremistisch motivierte Gewalttäter*innen durch Deradikalisierungsarbeit ihr Verhalten ändern, ein eigenverantwortliches Leben führen und (wieder) Teil des demokratischen Gemeinwesens werden, um so Extremismus jeder Art vorzubeugen. Die gemeinnützige Organisation, mit Hauptsitz in Berlin, stellt Instrumente und Ressourcen zur Verfügung, damit radikalisierte Menschen ihre bisherigen Verhaltensmuster reflektieren und überwinden können. Sie sollen befähigt werden, ein Leben zu führen, in dessen Verlauf sie weder sich selbst noch andere schädigen. Zudem werden Angehörige und Fachkräfte im konstruktiven, präventionsfördernden Umgang mit radikalisierten Personen unterstützt und fortgebildet."

Ich sitze als Schöffe in Verfahren, in denen wir entscheiden müssen, ob z. B. eine demente 90-Jährige, die im Krankenhaus während eines Demenz-Schubes ihre Bettnachbarin angegriffen hat, im Maßregelvollzug neben Schwerstverbrechern bleiben muss – oder (mit Bewährungsauflagen!) in eine passende Senioreneinrichtung kann. Wenn sie in der Einrichtung aufgrund ihrer Demenz nochmal aggressiv werden würde, würde sie gegen die Auflagen verstoßen – und wieder einfahren.

Gleichzeitig lassen wir einen schon verurteilten Mann, der einen Anschlag geplant hat (plus schwere Körperverletzung und räuberische Erpressung in der Akte), nach 22 Monaten frei und schicken ihn zu 26 Therapiegesprächen. Zudem sammeln wir ihn NICHT ein, wenn er auf Social Media pro-terroristische Inhalte teilt. Ein Auto kann der 21-Jährige(!) trotz sporadischer Überwachung dennoch mieten und seine Tat begehen.

Wie soll man das noch irgendwie erklären? Ich bin in einer Partei engagiert, ich bin Schöffe, ich hab Ehrenämter, ich mache alles gerne mit, damit das Land funktioniert. Ich glaube im Grundsatz an unser System. Aber es muss sich was ändern.

met a guy making $51,000/month by scraping reddit for the phrase "anyone got recommendations for" and emailing the posters within 2 hours

not joking

he opens reddit every morning
runs a search on 8 subreddits in his niche
finds people who just publicly asked for exactly what he sells
and emails them while the post is still on the front page

reply rate: 23%

average cold email: 0.3%

his is 76x higher

because these people literally just raised their hand in public saying "please take my money"

heres the exact phrases he searches every morning:

"anyone got recommendations for [service]"
"alternative to [competitor]"
"looking for a good [service]"
"got burned by [company], where do i go now"
"has anyone used [company] — are they worth it"

every one of these is a public declaration of buying intent with a timestamp on it

someone posted this 4 hours ago in r/[niche]:

"I need [service] urgently. Budget is $15K. Prefer someone who actually knows what theyre doing and isnt sketchy"

he saw the post at 9am
pulled the username
cross referenced linkedin
found the email
sent this at 10am:

"hey [name], saw your post in r/[sub] about needing [service] urgently. we do exactly that — last client we booked 34 qualified calls in a month. free 15 min call?"

reply at 10:11am
call at 2pm
signed $15K contract by 4pm

7 hours from public reddit post to signed deal

from reading reddit in the morning

this is the process:

step 1: find the 5-10 subreddits where your buyers hang out

step 2: run searches for "recommendation" "alternative to" "looking for" "anyone use" in those subreddits every morning

step 3: filter for posts under 4 hours old with clear buying intent

step 4: pull the username → cross reference linkedin → find their email

step 5: send a 2 line email referencing their EXACT post

step 6: warm call every positive reply within 30 minutes

thats it

last month:

• 187 reddit posts scraped
• 164 emails found (90% find rate)
• 38 positive replies
• 24 booked calls
• 11 closes
• $4,600 average deal
• $50,600 in revenue

from reading reddit while eating breakfast

heres why this layers perfectly on top of a normal cold email system:

your main campaigns are emailing 4,500 cold leads a day building pipeline at scale

reddit is a surgical strike on top of that

the guy on reddit literally JUST declared their problem hours ago

the declaration is the qualifier

theres no discovering pain points

they posted the problem
you showed up with the solution
the sales conversation is already 90% done before the call starts

its the highest intent cold email you can send because it might not even feel cold

they had the problem at 9am
you showed up at 10am
thats not a cold email thats a well timed introduction

and heres the part that should make every cold emailer sick:

every single person on this app is scraping apollo fighting over the same 50 million contacts

meanwhile reddit has 500 million monthly users posting their exact buying requests in public

timestamped
organised by niche
completely uncontested
free

nobody in cold email is scraping reddit

because nobody thinks of reddit as a lead source

they think of it as a place to post memes

while angry buyers are literally typing their credit card out loud every single day

one of the most profitable lead source on the internet has been sitting there for 15 years

organising itself
timestamping itself
qualifying itself
telling you their budget
telling you their pain
telling you what they just got burned by

and agency owners are still writing "hey firstname hope this finds you well" to marketing directors on apollo

go open reddit

search one of those phrases in a subreddit in your niche

youll find 10 hand raisers before your coffee gets cold

simples

p.s. if youre an agency owner with a proven offer and want us to setup a cold email system that books you 10-30 calls per month - DM me "EMAIL"

(you ONLY pay for qualified calls actually booked onto your calendar)

GeoLibre v2.3.0 is here!

GeoLibre is a free and open-source, lightweight, cloud-native GIS platform for visualizing, exploring, and analyzing geospatial data. It runs everywhere you do, in the web browser, on the desktop, on mobile, and inside Jupyter notebooks, all while keeping your data local and private.

This release brings a legend that writes itself from your symbology, a new GeoLens catalog browser, and 200+ GeoLibre Rust geoprocessing tools running entirely in the browser.

What's new in v2.3.0

• Automatic on-map Legend: the legend builds itself from your visible layers, with class rows for graduated, categorized, rule-based, and expression styling, gradient bars for heatmaps and raster colormaps, and land-cover labels from a Raster Attribute Table. Rename, hide, reorder, or add your own entries, and it saves with the project.
• Symbology swatches in the Layers panel: every row shows a dot, line, square, or image glyph in the layer's own color, so a tall layer stack reads at a glance.
• GeoLens catalog browser: connect to a self-hosted GeoLens server, search its catalog, and add datasets as vector tiles, GeoJSON, or rendered raster tiles.
• Emerging Hot Spot Analysis: build a space-time cube from timestamped points and classify every cell as a new, intensifying, persistent, diminishing, sporadic, oscillating, or historical hot or cold spot, all client side.
• Mosaic time series: the Time Slider now steps through MosaicJSON and STAC collections of many COGs per date, on either a GPU or a WASM rendering engine.
• Copy and paste layer styles: give a whole set of layers one consistent look without restyling each in turn.
• Shareable tool links: deep-link any Whitebox tool with a ?tool= URL that opens the dialog preselected and pre-fills the form, with a Copy link button to build it for you.
• Smarter data loading: pick which layers to load from a multi-layer GeoPackage, import CSVs whose coordinates are in any projected CRS, and read a raster's real CRS, pixel size, and extent from the metadata dialog.
• Multiple AI profiles: define several provider, model, and credential setups, pick a default, and switch between them from the assistant panel.

Try it out

• Launch GeoLibre Web: web.geolibre.app
• GitHub: github.com/opengeos/GeoLi…
• Documentation: geolibre.app
• Release notes: github.com/opengeos/GeoLi…

#GIS #Geospatial #OpenSource #RemoteSensing #MapLibre #GeoLibre

Andrej Karpathy just dropped 12-page PDF on "Graph Engineering" for multi-agentic systems

the shift: Karpathy's loop runs 700 experiments and forgets all of them. A graph remembers forever

here's the full system:

step 1 → build one loop: generate, critique, revise. 630 lines, 700 experiments in 48 hours

step 2 → go parallel: agents in separate worktrees, same repo, different branches, no conflicts

step 3 → add a knowledge graph: extract entities, resolve aliases, assemble typed edges, query through subgraphs

step 4 → ground your evaluator: it checks claims against graph edges, not vibes

step 5 → plug the graph as shared memory. workers write to it. evaluators fact-check against it. Loops persist overnight

step 6 → the agent forgets. the graph does not. stop rebuilding context from scratch every session

Karpathy ran 1 agent in 1 direction. Anthropic's graph runs 1,000 with shared memory - same model, it's the architecture

this 11-page PDF changed how I'm building multi-agent systems today

read it now - then explore the full graph engineering article below ↓

Kimi's CEO🇨🇳 says every AI lab has the wrong obsession

Zhilin Yang, on why K3 beat the frontier: "Every lab, like Claude, thinks the model matters most. That's wrong. It's how you organize the people building it that wins."

Then Moonshot proved the philosophy. K3 sold out every plan on purpose, cutting their own revenue instead of throttling existing users. When Anthropic hit that wall in April, they cut users' usage 50% at peak.

His one big idea: long context is the AI era's RAM. The 128K-to-gigabytes jump, compressed into 2 years instead of 40.

The real moat: "your biggest advantage, perhaps your only advantage, is your organization."

Model, or the team behind it: which actually wins?

Transformer by hand ✍️ ~ 6 steps walkthrough below

Open the hood of a transformer and the parts list is overwhelming: embeddings, positional encoding, attention weighting, self-attention, cross-attention, multi-head attention, layer norm, skip connections, softmax, linear, Nx, shifted right, query, key, value, masking.

Which of those actually make the car run?

Two of them. Attention weighting and the feed-forward network. Everything else is an enhancement to make it run faster and longer, which is how we got from a car to a truck, and to the word "large" in large language model.

So I drew and calculated those two parts entirely by hand.

Goal: push five features through one transformer block, filling in every cell yourself.

1. Given

Five positions of input features, arriving from the previous block.

2. Attention matrix

Let us feed all five features to a query-key module (QK) and read back an attention weight matrix, A. The details of that module are a post of their own.

3. Attention weighting

We multiply the input features by A to get the attention weighted features, Z. Still five positions. The effect is to combine features across positions, horizontally: X1 becomes X1 + X2, X2 becomes X2 + X3, and so on.

4. First layer

Let us feed all five weighted features into the first layer of the FFN. Multiply by the weights and biases. This time the combining happens across feature dimensions, vertically, and each feature grows from 3 numbers to 4.

Note that every position goes through the same weight matrix. That is what "position-wise" means.

5. ReLU

We cross out the negatives. They become zeros.

6. Second layer

Let us bring it back down: 4 dimensions to 3. The output feeds the next block, which has a completely separate set of parameters, and the whole thing runs again.

You have just calculated a transformer block by hand. ✍️

The takeaway: the two parts are doing two different jobs, and neither one alone is enough. Attention mixes across positions, so a feature can see its neighbours. The FFN mixes across feature dimensions, so each position can think about itself. Horizontal, then vertical. Then that pattern repeats N times, each block with its own separate set of weights. That is the Nx from the list up top, and that is what makes the transformer run.

💾 Save this post!

#AIbyHand #Transformers #DeepLearning

Ilya Sutskever explains the 1% human training loop, and why fully autonomous AI is the wrong target

"The thing you really want is for the human teachers that tell, that teach the AI, for them to collaborate with an AI."

"You might want to think of it as being in a world where the human teachers do 1% of the world and the work and the AI do 99% of the work."

"You don't want it to be 100% AI"

"But you do want it to be a human-machine collaboration which teaches the next machine."

Most teams are trying to remove the human from the loop. Ilya's design keeps a tiny human control surface and uses it to teach the next generation.

The leverage is not zero humans. It is 1% human judgment steering 99% machine work.

• Ilya Sutskever (@ilyasut), AI researcher, on @dwarkesh_sp

Graph Convolutional Network by hand ✍️ ~ 12 steps walkthrough below

Graph Convolutional Networks (GCNs), introduced by Thomas Kipf and Max Welling in 2017, are the tool for data shaped like a graph: social networks, recommendations, biological networks, drug discovery, molecular chemistry.

I drew and calculated a simple GCN entirely by hand.

Goal: run a two-layer GCN, then a small classifier, on a five-node graph, filling in every cell yourself.

1. Given

A graph of five nodes, A to E, with edges between some of them.

2. Adjacency matrix (neighbors)

Put a 1 wherever two nodes share an edge, in both directions.

3. Adjacency matrix (self)

Add 1s down the diagonal, one self-loop per node. That is just adding the identity matrix.

4. Messages

Multiply each node's embedding by the weights and biases, then ReLU. Negatives become 0.

5. Pooling

Multiply the messages by the adjacency matrix. Each node gathers the messages of its neighbours and itself.

6. Visualize

Node A pools [3,0,1] + [1,0,0] = [4,0,1].

7. Second GCN layer

Messages again: weights, biases, ReLU.

8. Pooling again

Pool over each node and its neighbours, once more.

9. Visualize

Node C pools [1,2,4] + [1,3,5] + [0,0,1] = [2,5,10].

10. Fully connected layer

Weights, biases, ReLU. This time there are no neighbours to pool, just the node itself.

11. Linear layer

One more: weights and biases.

12. Sigmoid

Squash each score to a probability (≥ 3 → 1, 0 → 0.5, ≤ -3 → 0). That is the classification for each node.

You have just classified every node in the graph by hand. ✍️

The outputs:
A: 0 (very unlikely)
B: 1 (very likely)
C: 1 (very likely)
D: 1 (very likely)
E: 0.5 (neutral)

The takeaway: a GCN layer is two parts. The top part pools each node with its neighbours through the adjacency matrix. The bottom part is an MLP that transforms each node on its own. A transformer layer has the same two parts, with an attention matrix where the adjacency matrix was. Both matrices do one job, mixing across positions: attention over tokens, adjacency over nodes.

In my class I call the GCN the transformer's little cousin: a bit more stubborn, because its attention is fixed by the graph rather than computed from Q, K, and V. Draw the two side by side and the resemblance is hard to miss.

💾 Save this post!

#AIbyHand #GraphNeuralNetworks #DeepLearning

The circular diagram presents the C major scale, its seven diatonic triads (I–vii°), and the corresponding relative modes in a single, integrated view. At its center lies C, from which the primary triads radiate outward:

CM (I), Dm (ii), Em (iii), FM (IV), GM (V), Am (vi), and Bdim (vii°),

each formed by stacking every other note of the C–D–E–F–G–A–B scale. Encircling these are the seven modes:

Ionian, Dorian, Phrygian, Lydian, Mixolydian, Aeolian, and Locrian

aligned with their respective parent roots and annotated with their characteristic scale degrees and interval patterns.

This unified layout enables musicians to quickly visualize harmonic relationships, making it a practical reference for composition, improvisation, modulation, and transposition across keys.

there are enough technology takes.

e.g. what models will do, what businesses get built, what commodification/open weights means, what moats no longer exist, what your firms venture capital thesis is, etc. all kinda interesting, but increasingly overdone to hell. twitter has turned most of this stuff into pure hive mind territory. you can see this when you talk to ppl irl, they bring up the same damn talking points.

but what there is not enough of is cultural analysis around technology.

how does ai change taste? ambition? status? dating? creativity? work? loneliness? education? trust? identity?

the model capability discourse matters. but the more interesting questions lie in what happens to ppl once those capabilities become widely distributed.

Stanford professor Judy Fan went on stage at MIT and broke down why humans are so good at making the invisible visible...

And why AI hasn't actually learned to "see" the way we do.

It completely changes how you think about Human Intelligence v/s Artificial Intelligence:

1. Nature never gave us straight lines or sharp corners. The number line, the coordinate plane, even basic geometry are all human inventions. We created tools that do not exist in nature simply because we needed a way to think more clearly.

2. The coordinate system Descartes invented solved a problem that had stumped mathematicians for centuries, doubling the volume of a cube. Once invented, this tool became so indispensable that virtually every math curriculum on Earth still depends on it.

3. Humans have been doing this for at least 30,000 to 80,000 years. The story of human progress is inseparable from the story of marking up our environment, from cave walls to Galileo's telescope to Feynman diagrams of particles we will never see with our own eyes.

4. Every major scientific breakthrough relied on a visual tool that made something invisible visible. Darwin needed side-by-side illustrations of finches to see variation that was otherwise too subtle to notice. Cajal needed detailed drawings of neurons under a microscope to map how the nervous system was wired.

5. Fan's research group studies something deceptively simple: how people decide what to put into a drawing and what to leave out. When two people played a drawing game, sketchers used far more detail when the target object had close competitors than when it stood alone, all the way down to using fewer strokes and less time when more detail was not necessary.

6. People are not just copying what they see. They are making constant judgment calls about what level of detail actually serves the goal of communication, and they do this naturally without ever being taught the theory behind it.

7. There is a real difference between drawing something so someone can identify it and drawing something so someone can understand how it works. In one study, participants drew explanatory diagrams that emphasized moving, causal parts of a machine while depictive drawings emphasized background and overall appearance, even though both were drawing the exact same object.

8. Explanatory drawings were genuinely better at helping someone figure out how to operate a machine, but worse at helping someone identify which machine it actually was. You cannot optimize a single drawing for both goals at once. Communication always involves tradeoffs.

9. AI vision models trained on photographs generalize surprisingly well to simple, sparse sketches, suggesting that resemblance based recognition is not just a story we tell ourselves. It is something modern neural networks can replicate with real accuracy.

10. But there remains a large, measurable gap between how confidently AI models recognize sketches and how confidently humans do, even when both groups answer the same questions about the same images. Humans are simply far more reliable and far more consistent in their judgments.

11. When researchers compared human-made sketches to AI-generated sketches under tight stroke budgets, both were similarly recognizable at higher budgets, but diverged sharply as the budget shrank. Humans and AI systems simplify drawings in fundamentally different ways once resources get scarce.

12. Reading a graph is not one single skill. It involves perception, knowing where to look, mapping that visual information onto the actual question being asked, and then translating that mapping into an answer. Each of these steps can independently break down, and people fail for very different underlying reasons even when they land on the same wrong answer.

13. When tested directly against humans on graph reading tasks, leading multimodal AI models, including GPT-4V, showed a meaningful performance gap. Even when a model's overall accuracy approached human levels, its pattern of mistakes looked nothing like how humans actually get things wrong.

14. People choose entirely different types of charts depending on what specific question they are trying to answer, not out of a generic preference for bar charts or scatter plots. Their chart choices closely tracked which visualization would genuinely help someone answer that specific question correctly.

15. Two of the most widely used graph literacy tests in education research turned out to correlate strongly with each other, suggesting they measure overlapping skills. But when researchers dug into the actual error patterns, the standard categories used in textbooks, like "find the maximum" or "identify a cluster," failed to explain why people got things wrong nearly as well as a more basic, underlying four-factor model did.

16. The deepest goal behind all of this research is not just academic curiosity. It is to eventually help students and everyday people develop genuine literacy with the visual tools that science and modern decision-making increasingly depend on, because every generation should be able to see further than the last by standing on the visual tools the previous generation built.

Follow @yasminekho for more ideas on thinking better, becoming clearer & building a more intentional life.