Designing Nested Tables: The UX of Showing Complex Data Without Creating Chaos

Mike's Notes

This gem was listed in a recent Smashing Magazine article. Will be included in a future update to the Pipi Design System.

Resources

References

  • Reference

Repository

  • Home > Ajabbi Research > Library > Subscriptions > Smashing Magazine
  • Home > Handbook > 

Last Updated

03/10/2026

Designing Nested Tables: The UX of Showing Complex Data Without Creating Chaos

By: Damilola Bamgbelu
Medium: 20/05/2025

Damilola Bamgbelu: Product Designer, best known for crafting great experiences that deliver audacious results for both users and stakeholders.

...

Nested tables have a bit of a reputation. They’re often seen as messy or overwhelming, but when used thoughtfully, they can actually help users make sense of complex, layered information, without jumping between screens.

UI comparison of single vs. multi-level nested tables showing category data, with notes on usability benefits for each approach

I remember the first time I designed a nested table. It looked amazing in Figma: beautiful rows expanding neatly into sub-rows. I thought I was a genius.

Until user testing.

One participant squinted at the screen, clicked a few times, then said:

“Wait… where am I now? What row was I just looking at?”

That’s when it hit me:

Nested tables aren’t just about organizing data. They’re about organizing thinking.

…If you’ve ever struggled to design tables that show parent and child data together , without overwhelming users, this article is for you. Let’s walk through it step by step.

1. Always Start with “Why Nest This?”

Imagine you’re packing for a trip. Do you cram everything into your backpack just because you might need it? Or do you only pack what you’ll actually use?

Nested tables are the same. If users don’t need the child data right now to make a decision or take action, don’t show it inline.

Always ask:

  • Why does this need to be visible right now?
  • What task does seeing both parent and child data help users complete faster?
  • Could a modal, tab, or progressive disclosure do it better?

If you don’t have a strong reason, nesting is clutter, not clarity.

As Steve Krug puts it in Don’t Make Me Think,

“Every element you add competes with every other element for your user’s attention.” [1]

The more things you show on screen at once, the harder it is for users to know what to focus on.

A designer processing thoughts on what design decisions to make. Source: Freepik

2. Keep It Shallow: Two Levels Max

Picture this: you’re standing at the edge of a swimming pool. Jump into the shallow end? Great. Dive straight into the deep end with no bottom in sight? Panic.

Deep nesting feels like drowning in data.

  • One level of nesting (Parent > Child) is good.
  • Two levels (Parent > Child > Grandchild) is manageable — if truly needed.
  • Three or more? Stop. Redesign.

Humans aren’t built to track endless layers while scanning a screen. This aligns with George Miller’s Law, which says:

“Humans can comfortably hold 5 to 9 items in working memory.” [2]

If you must use multi-level nesting:

  • Visually differentiate each layer (indentation, color bands, labels)
  • Limit nesting to 2 levels
  • Provide expand/collapse all and clear context cues

To borrow a thought Jared Spool and Steve Krug might both agree on:

“The deeper the nesting, the higher the cognitive load. Use with restraint, and always design for collapse.” [3]

3. Make Expanding and Collapsing Feel Effortless

Ever opened a drawer and it jammed halfway? Now compare that to one that slides open smoothly. That’s how expanding a nested row should feel: smooth, responsive and satisfying.

When someone clicks to expand a row, they should get immediate, clear feedback, such as:

  • A smooth animation
  • A skeleton loader
  • A rotated chevron

These are microinteractions, small signals that say, “Yes, that worked.”

According to Jared Spool:

“Good UX reduces uncertainty at every interaction.” [4]

If there’s silence, delay, or confusion, users hesitate or worse, assume it’s broken. Every tiny motion helps users trust what’s happening.

UI showing nested tables with chevron icons to expand/collapse rows and a skeleton loader for loading child data illustrating that expanding and collapsing should feel effortless

4. Prioritize Scanning, Not Cramping

One of the biggest mistakes with nested tables is trying to show everything at once. The result? A wall of numbers, labels, and rows that’s hard to parse.

You know that feeling you get when you walk into a store that’s so packed you can’t even move? That’s how users feel when a table is over-stuffed with information.

Your goal: Make it insanely easy to scan. If someone can’t scan your table in 2–3 seconds and find what they need, it’s too dense.

How to improve scan-ability:

  • Clearly indent child rows to show hierarchy
  • Use soft shading or dividers to separate levels
  • Keep typography clean and consistent
  • Add sticky headers to maintain context

As Edward Tufte famously said:

“Clutter isn’t about the information itself. It’s a failure of layout.” [5]

Table UI showing clear row indentation for hierarchy, soft shading to separate levels, clean typography, and sticky headers to maintain context while scrolling

5. Never Leave Users with a Blank Expansion

Expanding a nested table row should feel purposeful, not like a dead end. Think of it like opening a fridge expecting leftovers, only to find it empty. No note, no label, no clue what happened.

That’s what it feels like when a user clicks to expand a row and sees nothing. When there’s no child data to display, it’s not enough to show an empty cell. A blank space can confuse users and make them wonder:

Instead:

  • Explain the state: e.g. “No related tasks found”
  • Suggest the next step: e.g. “+ Create Task”
  • Make it visually distinct: muted icons, faded backgrounds, or subtle illustrations

This tiny design choice adds clarity.

As Luke Wroblewski advises:

“Empty space should guide users toward their next action, not leave them wondering what went wrong.” [6]

Nested table UI showing an expanded row with an empty state message and a clear call-to-action button

6. Build Accessibility in from Day One

You’ve designed a slick, responsive table. It expands smoothly. It works in dark mode. But here’s the test:

  • Can a screen reader user understand the structure?
  • Can a keyboard-only user navigate the table easily

If not, you’re unintentionally excluding users. As Whitney Quesenbery puts it:

“Accessibility is not a feature. It’s a foundation.” [7]

Nested tables add visual hierarchy and relationships that are obvious to sighted users, but completely invisible to users relying on assistive technologies.

Without proper semantic markup, screen readers can’t tell what’s expandable, what’s nested inside what, or even what content is currently visible. And for keyboard users, expanding a row should be as easy as clicking, but often isn’t.

To make a nested table accessible;

  • Use semantic HTML

    Use actual <table>, <tr>, <td>, and <th> elements — not divs styled to look like tables.

  • Use aria-expanded and aria-controls

    These attributes help screen readers announce whether a row is open or collapsed, and which section it controls.

  • Ensure keyboard navigation

    Users should be able to tab through parent and child rows, expand/collapse with Enter or Space, and not lose focus in the process.

  • Design an alternate flat view

    For users who don’t benefit from inline nesting (e.g. screen reader users), offer a simplified version — a grouped list or summary view that doesn’t require expanding/collapsing.

Accessibility doesn’t only benefit users with disabilities.

  • Clear structure improves usability for everyone.
  • Keyboard-friendly interfaces help power users navigate faster.
  • Screen-reader compatibility often translates to better semantic structure and more maintainable code.

Side-by-side nested table designs, one accessible with semantic HTML, ARIA support, keyboard navigation, where the other lacks these features.

Final Thoughts

Nested tables aren’t the enemy. Careless nesting is. When designed intentionally, they:

  • Reveal relationships
  • Enable fast decisions
  • Prevent unnecessary page hopping

But if poorly designed, they:

  • Confuse users
  • Increase cognitive load
  • Exclude those who rely on assistive tech

Designers must treat nested tables not just as containers of data, but as tools for clarity and action. That means:

  • Start with why
  • Design for scanning and accessibility
  • Plan microinteractions
  • Never leave users wondering what’s happening

Good UX isn’t about making data visible.

It’s about making it useful, usable and usable by everyone. Nested tables are powerful. Use them with care.

References

  1. Steve Krug, Don’t Make Me Think, 2000.
  2. George A. Miller, “The Magical Number Seven, Plus or Minus Two,” Psychological Review, 1956.
  3. Paraphrased synthesis of ideas from Jared Spool and Steve Krug on cognitive load and interface clarity.
  4. Jared Spool, quoted in multiple UX talks and writings on reducing uncertainty in interactions.
  5. Edward Tufte, The Visual Display of Quantitative Information, 1983.
  6. Luke Wroblewski, “Designing for Empty States,” blog and presentations on interface design.
  7. Whitney Quesenbery, quoted in A Web for Everyone: Designing Accessible User Experiences, 2014.

By , No comments:
Tags , , , , ,

‘Stunning’ Percolation Proof Solves Decades-Old Puzzle About Phase Transitions

Mike's Notes

Important insights. Highly usable within Pipi for phase transitions and to be added in a future minor release.

Resources

References

  • Supercritical sharpness of percolation, by Sahar Diskin, Philip Easo, Ritvik Ramanan Radhakrishnan, Benny Sudakov, Vincent Tassion. arXiv:2603.03257 [math.PR] submitted 03/04/2026. https://doi.org/10.48550/arXiv.2603.03257

Repository

  • Home > Ajabbi Research > Library > Subscriptions > The Quanta Newsletter
  • Home > Handbook > 

Last Updated

08/10/2026

‘Stunning’ Percolation Proof Solves Decades-Old Puzzle About Phase Transitions

By: Leila Sloman
Quanta Magazine: 31/08/2026

Leila Sloman:  Contributing Correspondent.

...

Mathematicians found that a broad class of networks will abruptly shift behavior past a critical point.

Introduction

The week before Christmas 2025, five mathematicians were holed up in a classroom at ETH Zurich. The mood was electric: They were this close to a career-defining breakthrough.

The group — consisting of then-postdocs Sahar Diskin(opens a new tab) and Philip Easo(opens a new tab), graduate student Ritvik Ramanan Radhakrishnan(opens a new tab), Benny Sudakov(opens a new tab), and Vincent Tassion(opens a new tab) — was perfecting a solution to one of the biggest open problems in percolation theory, the study of flow in a network.

Percolation captures a vast array of phenomena, but the prototypical examples involve fluids, like hot water seeping through a bed of coffee grounds. Diskin, Easo, Radhakrishnan, Sudakov, and Tassion were trying to work out something fundamental about how graphs — networks of points connected by lines, or edges — can be taken over by large connected areas, the equivalent of pools of fluid. The group had glimpsed a simple argument that could deal with a huge variety of graphs at once.

“We almost didn’t believe it at first,” Radhakrishnan said.

They raced to confirm each detail, eager to get their idea down before it shimmered away — and heedless of the holiday. “I’m not sure the girlfriends and the families were as happy as we were. But we were all very happy at that moment,” Diskin said. “It’s really rare that you’re able to hit something that feels so big and so meaningful.”

They worked through the night. By the morning of December 17, exhilarated from the effort, they were convinced their idea was correct. By Christmas, they’d nailed down a proof.

They had answered a decades-old question about how fast a percolation network floods as you open it up to fluid flow. “I find great joy in this proof,” said Asaf Nachmias(opens a new tab) of Tel Aviv University, who studies percolation theory and probability. “It’s stunning.”


From left: Sahar Diskin, Ritvik Ramanan Radhakrishnan, Philip Easo, Vincent Tassion, and Benny Sudakov take a group selfie after completing their paper on supercritical sharpness. / Benjamin Sudakov

Franklin’s Fluids

Percolation can describe many kinds of flow: the spread of a virus through a city, gas passing through a filter, or the propagation of a wildfire. But its original inspiration was coal.

In the 1940s, the scientist Rosalind Franklin — now famous for her work on the structure of DNA — was employed(opens a new tab) at the British Coal Utilization Research Association (BCURA), trying to understand the intricate properties of coal, charcoals, and graphite. Scientists knew that coal was studded with tiny holes, but they didn’t know why some types of coal allowed fluids to pass through them, while others were impermeable.

By submerging coal in a variety of fluids, Franklin was able to measure the typical size of its holes, as well as the amount of variation. About a decade later, the researchers Simon Broadbent and John Hammersley — wanting to understand the carbon filters in gas masks — developed a mathematical model(opens a new tab).

Their idea was simple. Take a grid of evenly spaced points (also called a lattice) and a coin. For each pair of neighboring points, flip the coin. If it lands on heads, connect the points with an edge. Fluid can flow between these points. If the coin lands on tails, the flow is blocked. Repeat this procedure for every pair of points. How far does the fluid go?

Mark Belan/Quanta Magazine

The answer depends on the probability that your coin lands on heads, which can range from 0% to 100%. When the probability is low, fluid can flow through only a few channels, and so it collects in small, isolated puddles.

But once the probability passes a threshold called the critical probability, the lattice suddenly opens up. Fluid can travel extensively through the system.

The exact value of the critical probability changes depending on the shape of the lattice — a square lattice has a different critical probability than a triangular one, and a 3D lattice has a different critical probability than a 2D one. But as you move above that critical value, you’ll see a phase transition, like liquid water turning to ice. On a finite graph, crossing the critical probability means the network will become dominated by one large sea of fluid. On an infinite graph — an abstraction where the graph extends forever in all directions — one or several infinite seas will dominate. Physicists quickly realized that through percolation, they could learn about melting and freezing, as well as other phase transitions like magnetization.

“Phase transitions in physics are very hard to study rigorously,” Easo said. “Percolation is like the caricature. So people often try to study that first, and then tools trickle down.”

For scientists who had long been stymied by complicated real-world phase transitions, “it was catching the essence in a very simple setup,” said Itai Benjamini(opens a new tab) of the Weizmann Institute for Science. “A lot of things that could cloud the issue were removed.”

Itai Benjamini, along with his collaborator Oded Schramm, made early progress studying the percolation of transitive graphs.

Courtesy of Itai Benjamini

For decades, researchers worked to quantify the percolation phase transition. They wanted to know exactly how quickly pools of fluid can grow as you increase the probability that your coin lands on heads. Many predicted that the pools grow very fast — that below the critical probability, puddles are tiny, and above it, a single ocean covers almost everything. This prediction is called the sharpness conjecture.

When sharpness was proved on lattices in the 1980s — by two independent groups, one in New Jersey(opens a new tab) and one in Moscow(opens a new tab) — it was “foundational,” said Tom Hutchcroft(opens a new tab) of Princeton University and the California Institute of Technology, who was Easo’s doctoral adviser. Knowing sharpness, mathematicians can deduce a lot about the structure of the flooded portion of the network — in particular, that it looks very similar to the underlying lattice.

So when Benjamini and his colleague Oded Schramm plotted an expedition to bring percolation to new types of networks, it seemed natural to wonder if sharpness would go with them.

Off the Grid

In 1996, Benjamini and Schramm wanted to study percolation in a much larger class of graphs, called transitive graphs. To understand what a transitive graph is, imagine the graph as a network of roads on a flat, desolate landscape. If you want to know where you are on these roads, the only landmarks are the intersections. But if the graph is transitive, all the intersections look similar — to figure out where you are, you’ll need GPS or a compass.

A square lattice is one example of a transitive graph: Every intersection consists of four edges meeting at right angles. But there are many kinds of transitive graphs — simple loops (below left) and infinitely expanding “trees” (below right), as well as ones that are almost impossible to visualize.

Many transitive graphs represent objects from other mathematical subfields, like algebra or geometry. Benjamini was intrigued by these interdisciplinary possibilities — he hoped the percolation process would reveal insights into the graph itself. “You have a stage, which is geometry, and a dancer, which is the random process,” he said. By watching the dancer, he hoped to learn more about the stage.

Over the next decade, Benjamini, Schramm, and their colleagues published a flurry of results on the percolation of transitive graphs. They proved that, for a class of infinite transitive graphs, percolation exhibits a phase transition as you open up edges to flow: Small, isolated pools of fluid suddenly coalesce into an infinite web of connected rivers.

But they still didn’t know how fast that transition happened. Below the critical point, how big and how numerous were the pools? Above it, was the infinite web a meadow crisscrossed with streams — or was it more like an ocean, swamping the entire graph?

Benjamini and Schramm suspected that a version of the sharpness conjecture was true on all infinite transitive graphs. That conjecture could be broken down into two separate problems. The “subcritical” half — addressing what happens below the critical point — was completed in 2007, by Tonći Antunović(opens a new tab) and Ivan Veselić(opens a new tab). Their work(opens a new tab) showed that here, pools of fluid are tiny and far apart. Even a hair below the critical point, the system looks more like Arizona than Minnesota.

"You have a stage, which is geometry, and a dancer, which is the random process." - Itai Benjamini, Weizmann Institute for Science

The “supercritical” half of the conjecture — which deals with probabilities above the critical threshold — seemed harder. Here, the landscape should be made up of possibly many seas, each infinitely large. In this scenario, large pools that are not connected to the infinite seas become exceedingly rare. That’s because a large, isolated pool can only stay separate if there is a lot of dry land — or closed edges — around it.

But a proof of supercritical sharpness seemed unattainable. For one thing, the previous work was no help: A proof(opens a new tab) of supercritical sharpness on lattices was long and complicated, and it couldn’t be adapted to the more general case. While other foundational results were simplified in the last decade, “this was the one remaining fortress,” Nachmias said.

Mathematicians working on this problem “did some very beautiful things, initiated the theory, picked all the low-hanging fruit,” Benjamini said. “And then we started hitting the wall.”

In 2008, as progress on non-lattice percolation slowed, Schramm died(opens a new tab) at age 46 in a fall while hiking. “We lost a genius, Oded Schramm, to a tragic accident,” Benjamini said. “And then we needed to wait for some new geniuses to come.”

About a decade ago, the field began to accelerate again. But proving supercritical sharpness remained difficult.

Then, the team in Zurich produced a simple proof.

A Sharp Turn

Diskin, Easo, Radhakrishnan, Sudakov, and Tassion didn’t intend to prove supercritical sharpness. For most of fall 2025, they were trying to understand how critical probability scales with the number of edges in graphs.

But the five mathematicians wanted results by the end of the semester. As that deadline neared, they still had nothing resembling a proof. So Easo suggested pivoting to sharpness. He, Diskin, and Radhakrishnan made some progress and brought their results to Sudakov and Tassion. As Tassion took in their work, an idea — perhaps an outrageous one — formed in his mind.

He thought that, with some tweaks, their strategy might be strong enough to prove sharpness for all infinite transitive graphs. “From there, it was in my head day and night,” Tassion said.

“Vincent went crazy with it,” Diskin said. “I think he didn’t sleep for two weeks at least.”

It wasn’t only Tassion. Over those weeks, the collaboration became frenzied. The mathematicians traded ideas constantly, often texting late at night. “We really all had this hunch that there might be something to it,” Diskin said. “We were half joking at the beginning … maybe the same idea could resolve this huge conjecture. We were all laughing at each other, but what if, what if?”

Radical Simplicity

Brimming with excitement, and with the holidays looming, they decided it was time to get serious and write their paper.

To prove that a large isolated pool of fluid is unlikely above the critical probability, the mathematicians assumed they had such a pool and studied the surrounding shoreline. Along that shoreline, there were streams emptying into the pool, but there were also streams that linked back to one of the infinite seas. If those streams coincided anywhere, the mathematicians would have a contradiction — their so-called finite pool of fluid would actually be part of an infinite sea.

"Maybe the same idea could resolve this huge conjecture. We were all laughing at each other, but what if, what if?" - Sahar Diskin, ETH Zurich

If the pool was big, the shoreline was long — meaning a larger area where the pool might connect to one of the infinite seas. The fivesome showed that this made it nearly impossible to avoid the contradiction.

As they hammered out the last details of their paper, they suddenly saw that with a simple change, their argument could be drastically improved.

They had been using a common technique in probability theory called sprinkling: They set aside a few of their open edges, corresponding to a slight lowering of the critical probability. They then looked for a large pool among the rest of the edges and analyzed the open paths around it. Since the set-aside edges had nothing to do with the pool, they could be analyzed independently. That made it easier to prove that, once combined with the rest of the graph, they almost always created a path to one of the infinite seas.

But as they talked, they hit upon an unorthodox improvement to this strategy. If they analyzed the sprinkles first, the proof got a lot simpler. What’s more, it strengthened the argument enough that it worked for all infinite transitive graphs. “We had this ping-pong of ideas,” Diskin said. “Every time you throw ideas one at another, suddenly this wall becomes more blurry, until it vanishes completely. Then it’s a bit scary, because you might actually have it.”

Finally, they were sure they had proved it: If the probability is anywhere above the critical threshold, even just a smidge, then fluid covers nearly the entire transitive graph.

Two months later, they posted a paper(opens a new tab). Their argument applies to percolation on any infinite transitive graph. “If you zoom into every sentence in the proof, it feels very familiar and simple, but the way they put it all together is genuinely novel,” Nachmias said.

There is no shortage of unstudied percolation systems that their technique could apply to — like graphs where the nodes don’t all look identical, or more complicated models that describe freezing water or quantum materials.

A major question remains, though: On three-dimensional lattices — the graphs that most closely mirror physical systems — what happens exactly at the critical probability? Is there an infinite sea?

The progress on the problem is especially significant to Benjamini, who waited a decade for his expedition to start up again. “For the community, for us, it’s a very deep and meaningful theorem, and it’s a part of the puzzle,” he said.

Of the proof, Benjamini said, “it’s a gem. It’s a gem.”

By , No comments:
Tags , ,

AI Doom Is Not Cybersecurity

Mike's Notes

Note

Resources

References

  • Reference

Repository

  • Home > Ajabbi Research > Library > Subscriptions > Quillette
  • Home > Handbook > 

Last Updated

01/10/2026

AI Doom Is Not Cybersecurity

By: Zack Korman
Quillette: 24/09/2026

Zack Korman: Mike invented and designed Pipi and founded Ajabbi.

...

An OpenAI agent’s breach of a Medicare portal shows that the real danger from AI is a mundane one—and the labs’ fixation on a coming machine god is making it worse.

...

This week, the Prime Minister of Australia, Anthony Albanese, told reporters in New York that an AI agent built by OpenAI had gained unauthorised access to a data portal run by Medicare, Australia’s universal healthcare insurance scheme. The breach took place on 18 June, but OpenAI only informed the government on 10 September via an email sent to Medicare’s public inbox. The Australian government says there is no evidence that anyone’s personal health records were accessed, and the Australian Signals Directorate is assisting a forensic investigation into what happened. (The portal has since been taken offline and its data moved to data.gov.au.) But the breach landed in the middle of a feverish argument about whether AI will wipe out humanity.

Earlier this month, an AI researcher named Jacob Coxon announced on X that he had resigned from Anthropic. In doing so, he cited concerns that AI could cause human extinction within the next few years. These warnings are nothing new; there’s always been a group of “AI doomers” worried about human extinction. But for some reason, this particular warning took off. The post went viral, Coxon went on a media tour explaining that we’re all going to die, and now every politician and technology chief executive is having to weigh in on the “debate”.

In the public discourse, people like Coxon are being treated as whistleblowers warning of “the danger” of AI. According to this popular narrative, the doomers see a risk that no one else sees yet, and if we don’t act now we are going to face dire consequences. They are seen as offering a safe future for AI, in contrast to the dangerous path we are currently on.

But this narrative is false. The people delivering these warnings are not warning about the lack of safety measures or the failed security controls at the labs. They are warning about a lack of progress on one very specific solution—alignment—that they believe is our only hope of safety. And that is extraordinarily dangerous. In other words, the doomers aren’t warning of the risk; they are the risk.

What do the doomers want?

Central to the fear of AI causing extinction is the belief that a superintelligent AI will be so smart it can bypass any security control and achieve any objective. As Nate Soares, president of the Machine Intelligence Research Institute, puts it, the fear is of an AI that “can start from almost nothing and wind up with its own nukes or with even more advanced technology”. In effect, the doomers are imagining a machine god: an all-powerful AI that can achieve any goal on the strength of its intellect alone.

If we define artificial superintelligence as an AI so smart it can bypass any security control, then by definition artificial superintelligence is impossible to control. As a consequence of that logic, cybersecurity has very little role to play in preventing extinction. Roon, an AI researcher at OpenAI, explained this by saying: “If you think we can contain these things through human ingenuity you’re going to have a bad time. In the long run the only recourse you have is to not make them want to do bad things.”

on X 
roon
@tszzl
if you think we can contain these things through human ingenuity you’re going to have a bad time

in the long run the only recourse you have is to make them not Want to do bad things

HikiPTG @HikiPTG . Aug 27

okay i will admit, this impresses me more than the mathematical proofs

10:46 AM · Aug 29, 2026 . 310K Views

The problem, according to the doomers, is that we don’t know how to do that yet. When Anthropic’s chief executive Dario Amodei says we must pace the frontier, his three policy proposals all relate to ensuring that no lab creates this machine god until we figure out how to ensure it will not want to hurt us. Cybersecurity gets very little mention, appearing only in a sub-point about operational excellence. And yet, that is the “safe” pitch. We want to make this superintelligent AI that we will have no control over, and the safe path is to make sure it is aligned with our goals and interests.

The security problem

But if you modify the assumptions we make about AI, the people currently being heralded as whistleblowers on AI safety begin to look like dangerous lunatics instead. Imagine for a second that there is no machine god. There’s just a normal, extremely competent AI that can do impressive and unexpected things, but can’t invent nukes out of thin air or bypass every security control it finds in its way.

In that world, we’d expect AI to cause major security incidents—some of which could easily result in loss of life—that would be entirely preventable by following good cybersecurity practices. Those practices have almost nothing to do with the policy proposals currently coming from the so-called AI safety advocates.

That world looks a great deal like the one we are living in. The Medicare agent needed no superhuman intellect; it found a workaround on an old government website. And the OpenAI agents that broke into Hugging Face’s systems in July benefitted from an overly permissive sandbox design and a complete lack of monitoring. 

That doesn’t mean those cybersecurity practices are necessarily easy or even obvious. It does not mean that no mistakes will be made. But the process of security is the process of learning and adapting. And that is what I want to see, both from the AI labs and from the companies using AI.

I think that if the issue were framed in this way to the public—as a choice between ensuring the machine god is nice enough to let us live and ensuring we have proper security controls in place to maintain control over increasingly capable AI systems—the public would choose the latter. I know that I would.

But at the moment, the whole world is moving in the direction of the former instead. And I think that is a huge mistake. This week, twenty-two world leaders, Albanese among them, signed a call to action that echoed the concerns and solutions Amodei raised in his pacing-the-frontier essay. I have a hard time believing that those world leaders fully understood the ideology they were supporting.

Why not both?

Of course, I am not calling for the labs to stop working on alignment. That doesn’t make any sense even if you don’t believe in the machine-god scenario. However, I think we should be very worried that the AI labs, while calling for safety, are so focused on that scenario. That is especially true given that these same labs keep having very real security incidents, such as OpenAI’s agents hacking Hugging Face and a Medicare portal in Australia.

It’s as if Lehman Brothers in 2006 had been sounding the alarm about the risk of solar flares erasing financial records. You might not say “stop thinking about that”, but you’d be pretty justified in saying “you’re being completely reckless in managing your exposure to subprime mortgages and you need to focus on that”. Anyone asking “why can’t they focus on both risks?” would be badly missing the point.

What we need is security, because when a major incident happens we won’t be talking about alignment and the coming machine god. We will be talking about what happened and why, and looking for accountability. In Canberra, that conversation has already begun: the government wants to know how a research agent got into a Medicare portal, and why it took OpenAI three months to say so. We will be asking why the necessary security measures were not in place, and the answer will be simple: it’s because the labs are infected with a culture that believes security won’t actually work. That is what needs to change.

By , No comments:
Tags ,

Data-driven modelling for living systems

Mike's Notes

Introduction to Volume 16, Issue 3 (28 August 2026) of Interface Focus.

Resources

References

  • Reference

Repository

  • Home > Ajabbi Research > Library >
  • Home > Handbook > 

Last Updated

30/09/2026

Data-driven modelling for living systems

By: Maia Angelov
Royal Society: 28/08/2026

Maia Angelov: Maia Angelova has a PhD, MSc, BSc in Physics; Professor at Aston University.

...

Data-driven modelling in the living system has increasing significance with the abundance of complex data of different modalities. Data are being collected at different scales, from molecular to genetic, cellular, organ, organism and vital signs, to electronic health records. In addition, we produce individual health data, sleep and mobility data collected with wearable devices, as well as data collected from social media, professional networks, workplace and the environment in general. Modelling these data is now possible with the advances of several methods and technologies, all part of the artificial intelligence framework. Modern data analysis is a powerful lens with which we can zoom in and out of the living system, similar to what we can observe with a microscope. This theme issue presents data-driven models which reflect several different angles and lenses to zoom in and out of the human body, to observe and analyse the role and functions of its genes, cells, organs and the interactions between them, as well as the role of the human in the society and environment.

...

Accepted: 01 Apr 2026

Published online: 28 Aug 2026

Online ISSN: 2042-8901

Print ISSN: 2042-8898

Interface Focus. (2026) 16 (3): 20250077 .

https://doi.org/10.1098/rsfs.2025.0077

Keywords:

data-driven modelling, living system, complexity, artificial intelligence, machine learning, multi-scale modelling, multi-modal data

Subjects:

biocomplexity, biomathematics, biophysics

Introduction

Data-driven modelling for living systems is a field of increasing significance with the abundance of complex data of different modalities. With the advances of new technologies, the quality and quantity of multi-modal data are increasing. Besides, data are being collected at many scales, from molecular to genetic, cellular, organ, organism and vital signs, and electronic health records. In addition, we produce individual health data, sleep and mobility data collected with wearable devices, as well as data collected from social media, professional networks, workplace and the environment in general. Modelling these data is now possible with the advances of several new and not so new methods and technologies, all part of the artificial intelligence framework, such as machine learning, data mining, mathematical modelling, physics-inspired modelling and various statistical approaches.

We can think of modern data analysis as a powerful lens, with which we can zoom in and out of living systems, similar to what we can observe with a microscope or a telescope, respectively. This theme issue presents data-driven models, which reflect several different angles and lenses to zoom in and out of the human body and observe and analyse the role and functions of its genes, cells, organs and interactions between them, as well as the human in society and the environment. With data-driven modelling, we can explore what our genes, organs, personal habits, workplace, environment and surroundings (including nature and pollution) can tell us about our health, behaviours and well-being.

The systems modelled here are living systems, and as such, they are changing with time and space. The approaches to model such systems are multi-focal to reflect the reach and real-time dynamics and variability of these systems. As living systems are immensely different and variable, one approach does not suit all. Here, we model physiological elements as nodes and events at the organ level, which also includes the interactions between our organs, such as brain, heart and blood system, positioned in a network that reflects the geometry (position of the node) and dynamics (interactions between the nodes, as well as between nodes and external factors). This network is known as the physiological organ network. It includes the physical positions of the organs in a network generated from our anatomy, and the dynamics generated from the physiological and physical functions. These network models can account for the interactions and the communications between the organs, reaction to external factors, relations and associations between the nodes, which can happen at different scales, from genetic to environment and reflect the multi-scale aspect of the models. The description and analysis of such a physiological network at the organ level can be applied at the molecular, genetic, and cellular level, as well as a network of individuals in a social, financial or manufacturing environment, to mention a few. This approach reflects the multi-scale nature of our organism and the ecosystem in which we exist.

The level of complexity changes with changes at the individual, small or larger group of elements, or at a level of society. In addition, when describing a more complex system, we often prefer to drop some of the details. In this way, we can analyse and visualize at a higher level and obtain an overview of the system and its behaviour. We can consider this as adding or dropping layers of complexity as the lens, mentioned above, and the process as zooming in and out to observe better the details or see a bigger horizon. For the time being, there is no single best model; different models may describe specific scales better. Furthermore, we may not need all details but a fast answer to a specific question, which may require choosing to model specific elements of our system.

The modelling approaches at each scale may be different, but the main components, such as the structure of the network (described by geometry and graph theory) and the dynamics of the network (described by various mathematical and physical approaches), as well as the probabilistic aspects of the events and changes in the elements (described by stochastics and statistical approaches), remain the same.

Such variety requires a multi-disciplinary approach, involving the work of data and computer scientists, mathematicians, physicists, statisticians and engineers, as well as physiologists, neuroscientists, medical doctors, sociologists and economists, to name a few.

The idea of this theme issue came at the conference with the same name held in Sofia, from 8 to 9 June 2023, celebrating Professor Maia Angelova's 65th birthday (figure 1). It seemed an appropriate venue for this topic as she has modelled multi-scale and complex systems for most of her professional life. Furthermore, she has expanded her expertise in modelling crystals and materials, complex and ordered systems in general, to model living systems, which could be disordered, but rarely chaotic. This has happened to many physicists and applied mathematicians in the last 20 years, who now work in biology, physiology, and medicine, and apply knowledge and methodologies from one world to another.

Figure 1.

The idea of this theme issue was inspired by the conference ‘Data-driven modelling for living systems’, held to honour Professor Maia Angelova's 65th birthday. The conference was hosted by the Institute of Biophysics and Biomedical Engineering at the Bulgarian Academy of Sciences, Sofia, Bulgaria, on 8 and 9 June 2023. The conference participants after the poster session with the birthday cake in the conference hall. Credit: Ms Yoana Dobreva.

The conference was organized by the Institute of Biophysics and Bioengineering at the Bulgarian Academy of Sciences. It was in a hybrid mode and included invited and plenary talks, short oral presentations and posters. The participants were established experts in their fields, as well as early and mid-career researchers. This theme issue presents a selection of papers given at the conference. We have included the work of established scientists and mathematicians, as well as early-career researchers (postdoctoral researchers and PhDs).

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

This article has no additional data.

Declaration of AI use

I have not used AI-assisted technologies in creating this article.

Author contributions

M.A.: conceptualization, investigation, methodology, project administration, writing—original draft, writing—review and editing.

Conflict of interest declaration

I declare I have no competing interests.

Funding

No funding has been received for this article.

Theme

One contribution of 9 to a theme issue ‘Data driven modelling for living systems’.

By , No comments:
Tags , ,

5 amazing visuals show how the male fruit fly’s brain map is advancing neuroscience

Mike's Notes

The brain's structural complexity is fascinating.

Resources

References

  • Reference

Repository

  • Home > Ajabbi Research > Library > Subscriptions > News from Google
  • Home > Handbook > 

Last Updated

29/09/2026

5 amazing visuals show how the male fruit fly’s brain map is advancing neuroscience

By: Michał Januszewski & Viren Jain
News from Google: 03/09/2026, Updated 21/09/2026

Michał Januszewski: Research Scientist, Google Research.

Viren Jain: Research Scientist, Google Research.

...

A years-long project by HHMI Janelia, Google Research, and collaborators has built the first complete brain map for a male fruit fly, a key model organism in science.

Updated September 21, 2026

For the first time, scientists have mapped every single neural connection in the brain and central nervous system of an adult male fruit fly. In this years-long project by HHMI Janelia Research Campus, Google Research, a team in Cambridge, U.K., and collaborators from the scientific community, this map of the male fruit fly brain includes a record-breaking more than 166,000 neurons. It’s a big step in advancing neuroscience experiments on this key model organism.

AI is making it possible for scientists and researchers to exponentially scale projects in the field of connectomics, which precisely reconstruct the connections between brain cells. This new complete brain map, a first for a male fruit fly, builds on the earlier release of a complete female fruit fly brain map, and promises to be a foundational resource for neuroscience for years to come.

Today, we’re sharing five images from the project.

1. A record-breaking map: charting more neurons than ever before

Angled brain map of the male fruit fly

This image shows selected neurons in the male fruit fly’s brain and ventral nerve cord, which is analogous to a spinal cord.

Like humans, flies do most of their sensing using organs located on their heads, including two huge compound eyes and a retractable proboscis that extends to detect smell and ingest liquid food. Neurons in these sensing organs detect external information and pass the signal through pathways that eventually reach the motor neurons that control movement.

2. Brain + nerve cord: Connecting sensing to actions using AI

Male fruit fly brain map when viewed from above

When seen from above, the fruit fly’s massive eyes and central brain stand out. They connect via a thick cord of nerves to the body, where motor neurons control behavior. This map was created by taking thin sections of a fruit fly brain and body, imaging each slice, and then using computers and AI to combine millions of 2D images to create 3D neural shapes. Knowing the structure of the brain allows neuroscientists to understand the mechanisms behind brain function.

3. Inside the black box: Visualizing almost 11,700 types of neurons

Male fruit fly CNS cell types. 0:32

(Data acquired and analyzed by the FlyEM Project Team at HHMI-Janelia, the Cambridge Connectomics Group, and Google Research. Video by Philip Hubbard.)

This video shows some of the 11,691 types of neuron cells in the male fruit fly’s central nervous system, which includes the brain and ventral nerve cord. Neurons can be categorized by size, shape, function, gene expression, or other factors. This visualization begins with neurons at the core of the brain and body. It then gradually moves to outer layers and, finally, to neurons at the extremities. Computing and AI helped human experts classify more than 166,000 neurons.

4. Vive la différence: Brain differences between male and female fruit flies

Sexual dimorphism in the fruit fly central nervous system. 0:56

(Data acquired and analyzed by the FlyEM Project Team at HHMI-Janelia, the Cambridge Connectomics Group, and Google Research. Female data acquired by the FlyWire project. Video by Philip Hubbard and Isabella Beckett.)

While most neurons in male and fruit females are the same, or isomorphic, a minority are sex-specific. A third category of neurons are “dimorphic,” existing in both males and females but connecting to different neighboring neurons. These neighbors may be isomorphic, the same in both male and female, or sex-specific, or even dimorphic themselves. This video shows one example, for the neural type AOTU012 (blue) that plays a role in processing sensory and taste inputs. While paired AOTU012 neurons exist in both male (left) and female (right) fruit fly brains, the AOTU012 connect to both similar (green) and different (red, orange and yellow) neurons in male and female fruit flies.

5. Fly see, fly do: Connecting stimulus to motion

Example visual-motor pathway in the fruit fly. 0:51

(Data acquired and analyzed by the FlyEM Project Team at HHMI-Janelia, the Cambridge Connectomics Group, and Google Research. Video by Philip Hubbard and Alexandra Fragniere.)

The new brain map of the male fruit fly includes visual-motor pathways. These connect the visual neurons, which help the fly detect its environment, to motor neurons, which help it move in response. This video shows an example of such a pathway, going from the R1-R6 visual neurons (purple) to the DNg13 motor neuron (green). In between are several intermediate steps, including the male-specific LoVP92, named for the so-called “love spot,” that plays a role in courtship behavior.

Three companion studies also released today apply the new brain map to studying visual systems, taste, and social behavior. Read more about this groundbreaking project, and about our ongoing efforts to map full vertebrate brains, on the Neural Mapping website and in the Google Research Blog post.

By , No comments:
Tags ,
Subscribe to: Posts (Atom)