quantum of sollazzo

quantum of sollazzo

Archives
Sponsor
Subscribe
November 17, 2026

683: quantum of sollazzo

Quantum of Sollazzo

Forwarded this email? Subscribe here  ·  View it online

No. 683  · 

Quantum of Sollazzo
 

Things that caught my attention

This issue is going to sound a little US-centric but there you go. I found it very interesting to find that the US Census Bureau's Business Trends and Outlook Survey now asks firms about Super Intelligence, or SI, instead of Artificial Intelligence, or AI. This is based on President Trump's September Executive Order in which he famously started referring to Super Intelligence.

Now, I will leave the politics out of this, obviously. But one point to me is that I don't care about AI or SI because, as I've always said, "AI" is an imperfect and shifting label. Whenever I give talks, I remind people that I started university at a time where the phrase "artificial intelligence" referred to rules-based systems, or logic programming. Neural networks were a curiosity, which had become a bit more of a reality by the time I finished university in the mid 2000s. When I started my first job that had AI in its title in the NHS, artificial intelligence mostly referred to deep learning based techniques, and its powers were harvested most spectacularly for image recognition. The sub branch called Natural Language Processing was mostly focused on things like translation, where things were often far from usable. Now, the phrase AI refers mostly to large language models. So I'm not particularly precious about a label that keeps shifting, and if we want to call it superintelligence, or human-like intelligence, or whatever other definition may suit the current flavour of it, that is absolutely fine and, perhaps, more appropriate.

However, it is also true that definitions matter, especially when we're talking about a census or surveys or statistics. As Simon Willison laments in his blog post, that change of definition may cause an issue in how old surveys can be related to new surveys.

What label is best for you?

·

I wanted to have a go at US Midterms electoral maps, and specifically at hooking together a hex map of seats, together with arrow maps of swing. I played a bit with Claude Code, and I think I've settled on this prototype (with source code available in Quantum's GitHub repository).

This was inspired by seeing the arrow map the BBC developed, which was based on a fully geographical map. As you may remember, I've played with Spikes Map before, for the same purpose.

The key question for you: does having arrows on top of a hex map work better than a geographical map? Does it add value and make the swing easier to understand and measure?

A note on swing: calculating it makes most sense only in 2-party situations, and where both parties were present across two elections with the same electoral boundaries. It loses its power the further you move from this setting. In the US context, it's broadly ok, but there are situations where Independents win. In specific cases, like Montana, there may have been controversy: in this case, the Independent candidate was actually supported by most of the Democratic party against their own candidate; so I opted to show the swing as if the candidate was a Dem.

An issue highlighting that data is not quite neutral, as much as you want it to be.

Swing maps

'till next week,
Giuseppe @puntofisso.bsky.social

Sponsored by Jane Street

Jane Street depends on all sorts of messy, real-world data to understand financial markets and the global economy: think world news, decades of weather patterns, deidentified credit card spending, or packet captures of stock exchange market data feeds.

We're hiring Data Engineers to turn datasets like these into reliable inputs for trading. Working closely with our researchers, you'll evaluate unfamiliar datasets, build robust ELT pipelines, develop deep domain expertise, and decide what's worth exploring next. 

The job requires a mix of engineering, data analysis, and product sense. If you love the detective work of investigating a weird dataset and figuring out what it actually means, we want to hear from you. No financial background is necessary.

Want to hear what the job is really like? Eric Mannes, who leads our alternative data team, talked with Ron Minsky about it on Signals & Threads.

We have openings in New York, London, and Hong Kong.

Jane Street

Topical

*sshole · sshhole.com

This is cool. While there are doubts about how really impactful this vulnerability is, the way the story is told and visualised is pretty perfect, and that's how it earned its place in the Topical section.

*sshole · sshhole.com

ATP Finals Simulator

Using data from TennisViz and Courtside Advantage, and Claude Code, Danilo De Rosa has created this ATP Finals Simulator (this newsletter should reach you as I'm about to head to one of the matches in Turin...)
Source code available here.

ATP Finals Simulator

Tools & Tutorials

Data to Viz | A collection of graphic pitfalls

Yan Holtz has released this helpful collection of 61 flashcards covering common data visualization mistakes, organized into categories like misleading charts, color, design, maps, and labels. Downloadable as a ready-to-print PDF.

Data to Viz | A collection of graphic pitfalls

GitHub - NangoHQ/nango: Connect your agents & product to 1,000 APIs.

Nango is an open-source platform offering prebuilt tools to connect thousands of APIs for building agent and product integrations. It's allegedly used by Service Now and other big players.

Walker Playground

Lemmings in the browser. You probably have no idea what I'm talking about if you're under 30.

Walker Playground

XGBoost is still good

Spotify data scientist Andres Vourakis links to several GitHub repos with production models based on decision trees in large companies, suggesting that "XGBoost (and similar tree-based models) are still some of the most valuable models in production ML today".

What is Data Contract CLI? | Data Contract CLI

Datacontract CLI is an open-source Python tool that lints, tests, imports, and exports data contracts using the Open Data Contract Standard (to this newsletter's readers: not the same thing as the Open Contracting Data Standard).
It connects to sources like Snowflake, BigQuery, Databricks, and Postgres to verify actual data complies with contract schema and quality rules.

What is Data Contract CLI? | Data Contract CLI

learn-harness-engineering: Harness engineering beginner tutorial, from 0 to 1

This is a free course teaching "harness engineering": building instructions, state, verification, scope, and lifecycle systems around AI coding agents for reliable task completion.
It comes with lectures and hands-on projects.

Welcome! – From Spreadsheet to R

This is a beginner-focused workshop teaching spreadsheet users with no programming background how to work with R.

Not everything needs an LLM

Jina Yoon (PostHog): "Whereas LLMs return free-form and overly confident prose, decision models make model uncertainty explicit. In a calibrated model, answers that score 0.8 are right about 80% of the time. That means a score of 0.5 is a coin flip.
We've been experimenting with them at PostHog and several use cases have made it to production. This is everything we've learned about using decision models so far: when to use them, how to use them, and mistakes to avoid.
"

Not everything needs an LLM - by Jina Yoon - build mode

Work on our free address matching tool is going well

The Ministry of Justice's Robin Linacre reports on their free address matching tool.

Work on our free address matching tool is going well

The Watchmaker: Building a Mechanical Watch in a Database

"By the end, a database had designed a 26 mm watch movement of 34 parts, checked that no two parts share space, worked out by itself how to take it apart, and run it for a minute. Every tick was computed in SQL."

The Watchmaker: Building a Mechanical Watch in a Database · Trevor Dwyer

e-means: 10x faster k-means at equal loss - Blog

e-means is "a modern, robust k-means algorithm developed at Meta FAIR. A replacement of Lloyd's algorithm, e-means delivers a ~10× speedup in both epochs and wall-clock time while achieving lower loss."

e-means: 10x faster k-means at equal loss - Blog

Postal codes and tessellations

"From postal points to areas using Voronoi polygons".

Postal codes and tessellations | One World

Britecharts 3.0: Accessible D3 Charts for JavaScript and React

Britecharts 3.0 packs a lot of useful features: screen-reader data tables, keyboard navigation, dark mode, TypeScript types, and tools for coding agents.

Britecharts 3.0: Accessible D3 Charts for JavaScript and React

Data Thinking

Data to deployment

Today's data market is immature with no standardised units, pricing, or quality metrics, making it "the Wild West", argues Yoni Rechtman.

Data to deployment

Dataviz, Data Analysis, & Interactive

What it takes to get Europe on a bike

Another brilliant analysis and chart by The European Correspondent. Cities like Riga, Malacky, and Turin show cycling growth follows political will, EU funding, and public pressure after tragedies.

What it takes to get Europe on a bike

Building a Music Engraving System From Scratch | Note Universe Blog

A solo developer spent two years building a custom music engraving engine to power a sight-reading practice app (now using Claude Code for development).

Building a Music Engraving System From Scratch | Note Universe Blog

The AI Sell-Off Compared to Well-Known Stock Market Crashes: S&P 500 vs 1987, 2008 and Covid | Company Tracker

"The recent AI sell-off brought a lot of fearmongering headlines. Put next to past stock market crashes, it's only a mild correction for the S&P 500, 3% below its record. Tech fell harder, with the Nasdaq down 8% and chip stocks down 28%, as much as they fell during Covid (35%). We'll be tracking the sell-off here with daily data updates."
(via Fair Warning)

The AI Sell-Off Compared to Well-Known Stock Market Crashes: S&P 500 vs 1987, 2008 and Covid | Company Tracker

The Lancet Commission on 21st Century Global Threats to Health

"Explore how 17 major threats could shape health worldwide through 2100, and how much of that burden policy could still avert."
Interesting data (note it requires registration). By the Institute for Health Metrics and Evaluation (IHME) at the University of Washington.

The Lancet Commission on 21st Century Global Threats to Health

What's banned where? 100 maps of bans by country (2026) | One Map

One Map publishes a new daily map of national bans, reaching 100 maps covering 2,428 bans across 195 countries. Chemical weapons, torture, landmines, but also fur farming.

What's banned where? 100 maps of bans by country (2026) | One Map

India’s App Economy in 18 Charts

"The apps Indians have taken to since early 2024, and the ones they have left behind."

India’s App Economy in 18 Charts - by Manish Singh

Lo que paga cada vecino de Madrid en impuestos / What every Madrid neihgbour pays in taxes

The 124.237 buildings of Madrid, in 3D, coming from cadastral data

Impuestos en Madrid: 14.508,36 € por habitante

TV’s Biggest Show Is a Sitcom From 2007

A few interesting takes from Daniel Parris. Sitcoms have dropped from 13% of new TV releases to just 1.7%, yet remain the most rewatched genre, driving "comfort viewing."

AI

AI Can Draw Blueprints Now

An architect tested 13 AI models and found Anthropic's Claude Opus 5.5 (Sept 2026) can produce permit-quality construction drawing sets.

AI Can Draw Blueprints Now

Testing whether AI agents need a government change-of-address directory

Tom Loosemore tested AI agents Muse, Duet, Instinct, OpenClaw with 8 house-move scenarios to see if a government change-of-address directory would help.
No spoilers.

Testing whether AI agents need a government change-of-address directory

Fable 5.2 and Jagged AGI

Ethan Mollick tries Claude Fable 5.2, using it for a diverse set of expert-level tasks: filmmaking, hardware hacking, business planning, literary scholarship, and research. It shows "almost AGI" capability, according to Mollick.

Fable 5.2 and Jagged AGI

AI R&D Is Accelerating Towards Potential RSI in 2027

METR's report finds Anthropic's AI capabilities are improving at roughly 2x the pre-October 2025 rate, mostly due to AI-driven R&D. It looks like things are going faster than AI 2027 had predicted...

AI R&D Is Accelerating Towards Potential RSI in 2027

The missing quadrant in responses to AI

Sayash Kapoor, Rishi Bommasani, and Arvind Narayanan in their "AI as Normal Technology" blog announce the launch of STABL, a nonprofit studying how societal bottlenecks, not just capabilities, shape AI's real-world impact.

The missing quadrant in responses to AI

Did you like this issue?

Buy Me A Coffee

 

You're receiving this because you subscribed to Quantum of Sollazzo, a weekly newsletter covering all things data, written by Giuseppe Sollazzo (@puntofisso). Have a product or service data people should know about? Sponsor an issue.

The Quantum of Sollazzo grove now has 50 trees at Trees for Life, helping offset this newsletter's carbon footprint.

Proofread by Andy Redwood at Proof Red, who also makes motion graphics about climate change.

ProofRed Logo

© 2012–2026 Giuseppe Sollazzo · London · UK  ·  Unsubscribe

Don't miss what's next. Subscribe to quantum of sollazzo:
Older → 682: quantum of sollazzo
Powered by Buttondown, the easiest way to start and grow your newsletter.