Crank.report Survey

What Infrastructure Does Civil Society Need?

I recently had the chance to look closely at the work of a group of organizations. They were different sizes, working with different communities, over different lengths of time. The smaller organizations had a harder time. They had clarity on the job ahead of them and on their own community. They didn’t have the density of data over time that allowed them to talk about their impact.

I’ve been turning this over since, and I don’t have it worked out. Here are the pieces I’m trying to fit together.

1. What’s missing feels like the ability to try out different scenarios before you implement them in real life. It seems that is some of what data can give an organization. An ability to build scenarios, model an intervention and see its likely impact. That modeling is a tool to allow the community members to discuss and own desired impact.

2. A small organization already has three things, and they are not the same. Data, knowledge, and expertise. I’ve been thinking hard about how they differ. Data fits in a spreadsheet, who lives here, where the roads run, what the health needs are. Knowledge fits in a document, more and more a markdown file, the foodways of a place, how people actually move through a food pantry, what’s been tried before. Expertise is something else. It’s how we read the data and the knowledge together in a particular situation, and the decisions we make as we see them come together.

3. Two of those can be packaged. The third cannot and shouldn’t be. Data travels easily. Knowledge travels as an artifact, though it carries its context with it and lands differently somewhere else. Expertise doesn’t travel at all. At our best in civil society, expertise is what we in the community know and think—and, this matters, what we know because we work on things together. You can’t ship that. A skills file catches a little of it, the method, how a group tends to read a situation, and it helps. The rest lives in the room. So the job of any infrastructure is not to replace that judgment. It’s to give it more to work with.

4. The tools to hold data already exist, and the picture is still never complete. Esri’s Tapestry and Living Atlas, Google’s Data Commons, Esri’s work on digital twins—these show we can assemble and serve a rich data picture, and even use it to experiment before we act. That part is real and it’s here. But take a concrete problem: feeding your neighbors. To do it well you need to know who has a commercial kitchen, who has refrigeration, who has a loading dock, and what the community’s foodways actually are. Some of that comes from the commons—demographics, roads, health data. The rest we have to gather ourselves. And gathering it is exactly the work a small organization doesn’t have the capacity for. If every group builds its picture from scratch, we’ve rebuilt the problem we started with.

5. So the infrastructure I’m after isn’t the data. It’s the shared way in. We can’t make it every organization’s job to set this up. What the sector needs is a cheap, shared way to add local data and local knowledge to what the commons already holds—a common container and method, so one community’s effort isn’t thrown away, and so the group down the road can build on it instead of starting over. The markdown files are a hint of what the knowledge half of that might look like.

The way I’m using it, expertise is really two things: interrogating the data and the knowledge together, and then making a set of decisions aimed at doing better next time. What lets a community do that together—see what it has, see the alternatives, test them against what it knows to be true—I don’t know yet. We can build the data layer; Esri and Google largely have. Whether we can build the layer that helps a community actually decide together is the open question. It’s the part I’m really thinking on.

The organization will reportedly serve as both a leadership pipeline for students interested in public service and a research center examining issues affecting democratic governance, including political polarization, the ethical use of AI in government, criminal justice reform, and civil rights.

Nancy Pelosi to Form Institute at UC Berkeley Aimed at Teaching Civics and Bolstering Democracy

What a great retirement job, asset for the university, and for all of us who want democracy to work.

Today’s AI boom is unfolding through a highly concentrated ecosystem of hyperscalers, suppliers and private lenders linked by debt and increasingly opaque financing arrangements.

AI boom carries with it a big risk

All the more reason to practice principles of frugal AI: companies will go under, prices will rise — all in the effort to service that debt.

[Tyler Cowan] now describes ChatGPT Pro as approaching “the pocket-calculator level” of reliability for many of the ways he uses it. And because he has written online for more than 20 years and recorded years’ worth of public podcast conversations, he says, “I think the AIs know me better than almost any other human being on earth.”

An “Infovore” shares his chats

I’m looking forward to trying the how to listen to a symphony example. Of course, I will try it on Gemini.

Physical media fosters community. Flea markets, record fairs, used bookstores, and game swaps provide spaces for exchanging recommendations, sharing stories, and discovering unexpected finds. These interactions create connections formed through personal exchange rather than automated recommendations.

If You Can’t Hold It, You Don’t Own It

As someone with a house full of books, full of records, I appreciate this comment so much.

On browsing knowledge

Prompted by one of TechSoup’s board members, I’ve been walking around saying, “Websites won’t exist in two years.” Every time I say it, I keep whittling down the amount of time—18 months, 12 months, 9 months. It’s getting closer both because of time passing, and because the time to the next iteration of everything is being compressed.

My thinking here is still pretty disorganized. These are the pieces I’m trying to put together:

1. The “Webb” Store (See what I did there?)
   Imagine a dynamically built web store with an agentic backend. The store is built just for me—tires that fit my truck, the fishing lures I like, shirts from my favorite brands in my exact size. I still get the dopamine hit of shopping, comparing, and finally deciding to buy. The “shopping” isn’t just transactional; it’s about the knowledge that helps us choose. Once the choice is made, an agentic backend runs off and does the heavy lifting across different retailers.

2. Surfing AI-Friendly Knowledge Vaults
   Now look at the data side. Imagine surfing knowledge files in a tool like Obsidian. Let’s say we organize our expertise in “knowledge bundles”—markdown files with agreed-upon YAML front matter. We define the edges of these bundles so we can surf across them as a knowledge graph based on relationships, not just explicit hyperlinks. My files, my expertise, owned by me, but contributed to a collective map.

3. Putting the Two Together
   When you bring actions (the Webb Store) and the ability to surf knowledge (the Obsidian graph) together, we are engaging with the digital world in a completely different way. It isn’t SaaS—which is how so many of us are trapped using AI right now. It isn’t just a database. And it certainly isn’t the internet as we know it today. It’s a shift toward owning our data and using it to meet resource and referral needs in a way that bypasses traditional corporate gatekeepers.

4. Many Fronts, Bespoke Backends
   This paradigm can have many front ends—it can be something we chat with, type with, or look at. It adapts to the user. On the backend, bespoke agentic processes handle the execution based on the relationships defined in our knowledge graphs.

5. The Stakes for Civil Society
   My brain feels stuffed trying to understand, describe, and shape this because the stakes are incredibly high. There are massive dangers here—the kinds of systemic vulnerabilities highlighted in DeepMind’s AI Agent Traps paper. But the real opportunity is for ownership and shaping by civil society. If we get this right, we can organize for true community connection, coordination at scale, and decentralized ownership with an ability to surf across borders without being mined for profit.

Informed by

• Google’s Project Jarvis
• Vercle v0
• Google’s Open Knowledge Format
• Google’s Knowledge Catalog
• Filesystems are having a moment
• Oracle’s Comparing File Systems and Databases for Effective AI Agent Memory Management
• Google’s Disco
• DeepMind’s AI Agent Traps

So it must be that a key ingredient to blogging is simple: have a willingness to state something that seems obvious to you but nobody else is saying it.

Blogging Can Just Be Stating The Obvious - Jim Nielsen’s Blog

Noted.

simonw/browser-compat-db

Too geeky for me still. It will get there.

The AI boom needs electricians.

It also needs line workers, substation technicians, grid engineers, mechanical contractors, welders, construction crews, and commissioning specialists. These are not jobs that can be filled instantly with a software update or a new financing round. They require training, experience, and a steady labor pipeline that the power sector does not currently have in abundance.

That is an important reminder that the AI boom is not only a digital story. It is also very much a physical infrastructure story.

Silicon Valley’s AI Dreams Face a Blue-Collar Reality | naked capitalism

And it needs the infrastructure built in a way that powers communities. Not just compute.

Exclusive: ICE obtains local voter files in Texas and North Carolina

More related than relevant. This article started me thinking about community efforts that brought together non-voters with voters so the legal voter could cast a representative ballot.

The consequences will be dire for utility or for privacy, and possibly both. It’s hard to understate this point: future statistical releases will either be useless compared to past ones, or they will be incredibly unsafe.

Banning noise will be a disaster for statistical data products - Ted is writing things

This articles holds so much for me to learn more about. Also, something to watch with regard to the privacy and safety of the communities we support.

We don’t get chances to rewrite the rules very often, but this is one of those times. Last one was in the early 1990s with the advent of the web. My plan is to give all the new power back to the web.

On AI | scripting.com

This. I think the opportunity is ours for the taking. The window, however, will not be open for long.

There are three ways to do AI coding at home without spending like a company, and which one fits depends mostly on how much you trust the next year of hardware and model releases.

AI Coding at Home Without Going Broke | Stephen Bochinski

This is about coding but frames some of the decisions around Frugal AI well. The fact is for many of us, the downloadable slower models are fine for or work and tasks so I believe that option — having your own set up — is becoming more viable.

Data centers now seek amounts of electricity that used to be associated with entire cities — raising questions about who pays for new infrastructure, who gets access to scarce power and how quickly projects can connect to the grid.

The power decisions that could shape the next century

We have to figure out how to regulate and establish this system so that it benefits community.

Finished reading: Goodbye to Berlin by Christopher Isherwood 📚I didn’t realize this novella, almost a collection of vignettes, was the basis for Cabaret. The slow way a society changes is instructive reading.

Does knowing more stuff mean the model is better at the tasks we pose to it? I can certainly imagine how a coding model with deeper knowledge of modern libraries and patterns could crunch through coding tasks more effectively.

Claude Fable 5 | Simon Willison

On Frugal AI Usage

I worry a lot about what will happen when the music stops and the bill for the trillions of dollars spent on AI, data centers, chips, and memory makers comes due. Those payments are going to come from the people who can afford to stay locked in to the providers they’ve chosen with the services they’ve built.

And we all know those organizations are not going to be in civil society.

So this begs the question: how do we benefit from the boom in heavily subsidized AI without becoming a victim to it later? How do we avoid getting locked into skyrocketing usage pricing or giving up tools we’ve spent time building our workflows around?

In other words, how do we articulate principles of frugal AI usage?

First, a definition. Frugal AI usage means limiting the compute power to the minimum required. Right now, with user license-based pricing, this won’t net you a direct savings. Later, when it converts more completely to license plus usage-based pricing—where you pay by the token, the basic unit of data an AI processes—it will matter a lot. Frugal AI also reduces the environmental harm of AI use.

So back to the principles:

1. Use AI to build automations, but keep AI out of the repetitive execution. You don’t want to build a tool that relies on AI for the whole thing. It is time-consuming, expensive, and not as reliable.
2. Decompose your process into steps. Determine if a step is best accomplished by deterministic processes (fixed, predictable rules) or probabilistic processes (AI guesswork).
3. Use a hybrid approach with local models. When you are iterating, testing, or working on well-documented processes outside your personal skills, move the work to a local model running right on your machine.
4. Keep your context, instructions, and common prompts in lightweight, portable structured files. These markdown files. JSON. They live on your hard drive and are easy to edit and reference.
5. Use the principles of progressive disclosure. Only provide the required information to the AI at the required times. Because context equals cost, don’t burden the AI with rules it doesn’t need yet.

A Real-World Example: My Python-PPTX Effort

I can use a small project to demonstrate this. I used a command-line AI tool to build a workflow to help me make brand-compliant PowerPoints according to the rules I follow when making decks. It’s made up of a PowerPoint template in our brand style, a rules document saved as a markdown file, and a python script that generates the PowerPoint. Simple.

I start the python script in Terminal. I get a prompt asking for the topic of my presentation or the path for relevant files. Then—and for the first time—it invokes GenAI and, using my rules file, it generates the content. Finally, it hands control back to the Python script to put that text into the approved template and deposit the result in a specific folder on my hard drive. From there, I open it up and edit by hand. That whole process takes about 3 minutes.

Here’s how the frugal AI principles apply:

• I used AI to build the automation, not run it entirely. I didn’t ask the AI to handle the deck design or layout. Python is a well-established tool for generating PowerPoints, so I used that. Our template is deterministic. I don’t even use AI to create images for the deck; I just instruct it to put in placeholders with a design brief style instruction for the kind of image that might go there.
• I used a hybrid model approach. I shifted tools based on the phase of the project. I used a big, cloud-based AI for the initial design and brainstorming. But when I went from design to actually building and fixing the script, I switched to a local model on my machine. I could go back and forth, get it working, and never need to touch the internet. Zero tokens used for that portion, no internet bandwidth burned, but I still got the benefit of AI.
• My rules live in a markdown file. I can invoke that same file in different processes with different tools—my local model or any of the big commercial LLM tools out there. If I want to change my rules—say, go from a single-sentence headline to a descriptive phrase—I just open that document and edit it. No AI needed. Change the rule once, and I change it everywhere that markdown file is referenced. This keeps me from being locked into any single provider.
• I practice progressive disclosure so I only use what I need, when I need it. Let’s imagine I make my little project more complex: I want a set of rules for how numbers, charts, and graphs are displayed. I could put that all in my main rules document, but then that text is longer, and I’m burdening the AI’s instruction set with things that are only needed some of the time. Instead, I make a separate markdown file for my numbers rules. Then I update my script: if the information is general, just use my global rules. If the information includes sets of numbers, only then do we pull in the numbers rules document.

While this example is small, scale this approach across an organization running tens, hundreds, or thousands of workflows a day, and the savings on token context windows become staggering. You add context only when you need it, which takes less compute, costs less, and makes hallucinations much less likely.

I’m still working through these rules, these examples, and how to play it out in organizations and across the sector. I welcome any feedback.

That is a real, non-imaginary, discontinuous forward leap in capabilities. This did not happen because AI is magic. It happened because Fin already had exceptionally high engineering discipline, fast feedback loops, and a culture of experimentation and measurement.

If you want to know what engineering teams founded pre-AI can expect to achieve by embracing AI, there you go. This should be well within reach for the rest of us.

AI enthusiasts are in a race against time, AI skeptics are in a race against entropy

This engineering — and I’d add operational — discipline is so necessary and has to be put in place to make this leap and to diffuse the opportunity to innovate with AI inside the organization, across teams.

I described this to a colleague as the disappointing power of statistics. With traditional rule-based systems, you had been forced to build models that emerged from careful understanding of the problem. You had to articulate the structure of the domain before you could write code that operated on it, and the act of articulation was where most of the learning happened. Data-driven methods broke that requirement. You could now process large datasets with techniques that surfaced patterns and produced solutions without any of that prior knowledge work having been done. The result was capability without the comprehension that traditionally came hand-in-hand.

What I see in AI today is the same problem, accelerated and scaled to a level my MSc students could not have imagined ten years ago. The basic questions I started raising in supervisions have become questions about the constitution of knowledge itself.

The Disappointing Power of Statistics - by David Millard

This feels so spot on. Driving only with data, only with AI abstracts away the human component. That part holders judgement, experience, choice.

Welcome to Schrödinger’s UPS Vortex, the quantum rift within which your box is on a truck passing through Memphis, in a warehouse in Topeka, or on the outer rim of the galaxy, where it’s being worshipped as a god by a species of semi-intelligent space protozoa. I once got a text from UPS saying they’d picked up my package the day after the package showed up on my porch. (This technically makes me immortal.)

The 40 Most Rage-Inducing Problems in Tech - The Ringer

God. Please fix shipping updates.