Crank.report Survey

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.

This is a product mindset problem, not a technical problem. The technical part—actually writing the code—is relatively straightforward because the scopes are small and the components already exist. What’s hard is having the mindset that is able to choose the “right” software to build, and figuring out how to do it without breaking the things that are already working.

Boring Tiny Tools - Vaughn Tan

More thoughts to share on this article. For now, let me say this exactly the approach I’ve been taking for myself. I’m not sure that it’s the only one needed for org transformation.

The process of building the app reinforced something I’ve been thinking about for quite a while: coding is a specific skill, but it’s only one part of a much larger process. Great developers aren’t necessarily great coders, though they can be. Apps must be envisioned, their specifications defined. The act of trying to describe an app to an AI coding engine is a clarifying one. The more you describe the app, the harder your brain has to work, because it’s always more complicated than you think it’s going to be. The decisions you make determine what the app comes to be. It’s authorship of a sort, but defined in a way that takes the writing of code out of the equation, which is weird, since the act of coding has usually been an inextricable part of the process of making software.

Road to WWDC 2026: What’s a developer? | Six Colors

This, from the same article. I’m using this to build small utilities for myself — like the app I’m using to post this. I’m also using it as a way to think through and share more complex products before handing them over.

We now live in an era where, if you can dream an app, you can probably build it. Especially Mac utilities. And who cares more about native Mac software than Mac users? Certainly not those companies that gave up on Mac development and focused all their energies on giant cross-platform code bases to attract venture investment and big payouts.

Road to WWDC 2026: What’s a developer? | Six Colors

I love being able to make my own tools.

The ho-hum that is coming, the one already arriving at the edges, is something different. It is not the slowing of capability. It is the migration of AI from topic to infrastructure. It will go into the background. From the thing you think about to the thing that makes everything else work. Consider when the iPhone 18 marries Gemini into its operating system. (Of course, knowing Apple they will bungle it up.)

AI models are having their iPhone moment. What’s Next?

But it’s actually worse than that because even subscribing to Office 365 doesn’t fix the problem. You need a newer version of Office, which necessitates a newer version of macOS, which may necessitate getting a new Mac—all to fix what seems like an artificial problem.

Michael Tsai - Blog - Bricking Microsoft Office 2019

Harder and harder to keep your hardware and software going and secure.

The old plan wasn’t bad, but it also wasn’t the best plan that could be made with the information available. I debated what to do with it. Then I thought about me three months from now, and asked myself – do I want to regret not making this change then?

Making a New Plan | Accidentally in Code

This is such a common challenge in leadership.

Do you demonstrate consistency and with that help give others confidence in their ability to act?

Or.

Do you change as you get new information and so stay on the edge of what’s possible?

AI Doesn’t Have ROI | Hacker News

And an interesting conversation to go with that last post.

These users — much like the users of effectively every subsidized AI subscription — never really knew how much anything they did cost, because Microsoft intentionally hid the actual cost of prompts and allowed users to spend obscene amounts as a way of boosting growth for GitHub Copilot.

This problem is industry-wide.

Every single user of every single AI subscription service is having their tokens subsidized and the actual cost of AI obfuscated. As a result, every frothy, fluffy hype-piece about Claude Code or AI in general is a kalopsia — the belief that something is more beautiful than it really is.

AI Doesn’t Have ROI

This. This is what I worry about on behalf of nonprofits and the systems the may be setting up on these subsidized (and for them further discounted) accounts.

In the study, Nvidia and Microsoft outlined three distinct kinds of blind goal-directed (BGD) activity it often saw in AI agents. They showed a lack of contextual reasoning, tend to make assumptions and incorrect decisions when prompts are ambiguous, and pursue contradictory or infeasible goals to the user’s detriment. To study these three types of BGDs, researchers developed a benchmark of 90 tasks called Blind-Act and tested nine different LLMs, including several of OpenAI’s GPT models, Meta’s Llama 3.2, and two of Anthropic’s Claude models.

[Nvidia and Microsoft Researchers Say AI Agents Don

Coding agents really did change everything. These are tools which burn vastly more tokens, but are also quickly becoming daily drivers for the work carried out by extremely well-compensated professionals. Right now that’s still mostly software engineers, but a coding agent is a tool that can automate anything you can do by typing commands into a computer… so they are clearly applicable to a much wider set of skilled knowledge workers.

I think Anthropic and OpenAI have found product-market fit

This is what I’m finding. I’m using them to structure skills files which then interact with python scripts to things I need done. I’m not a coder. But I’m using these agents in the same model.

The lapse comes as the Department of Justice’s Office on Violence Against Women withholds $150 million in taxpayer dollars meant to be distributed in fiscal 2025 to help survivors of domestic violence, sexual assault, stalking and trafficking.

The 19th reached out to over 250 organizations and municipalities that had previously received funding from the Office on Violence Against Women to ask how the delay in fiscal 2025 funds and the stalling of fiscal 2026 grant applications has impacted them. The 48 that responded shared stories of layoffs, diminished services for survivors and sacrifices among advocates, who have taken voluntary pay cuts to make sure survivors get the life-saving support they need.

How a missing $150 million is impacting domestic violence services

These organizations are already chronically underfunded.

The API features an MCP and a CLI and ultimately opens the door to more members from our community building tools and apps on top of Buffer. Opening up Buffer has a huge potential impact for our customers and community members, and we couldn’t be more excited to see what you build.

Buffer’s API is Open for Building

Opportunity here to build shared tools for civil society orgs using Buffer.

It is the procurement model. Companies that signed up for a productivity tool are discovering they signed up for a metered utility, and the meter runs when nobody is looking. The fix may be straightforward: capped budgets per engineer, tiered access for high-leverage roles, agent runtime quotas.

Many of the larger buyers are already there. But the implication is that the era of “give every employee a Claude Code seat” is closing, and what replaces it will look more like AWS billing than like Office licences.

Microsoft’s quiet Claude Code retreat and the real cost of enterprise AI

Important for nonprofits (a) to budget; (b) to develop sector resources.

That is why, if we want this technology to go well, it is enormously important that there be people outside those incentives—people who care about things going well and insist on safety, who are paying close attention, who are willing to say hard things, who are willing to be our earnest, thoughtful, critics. It is through dialogue and mutual effort, through the push and pull, that humanity will achieve great things.

Anthropic co-founder Chris Olah’s remarks on Pope Leo XIV.

Sounds like a civil society call to action to me.

Former officials and industry leaders fear the Cybersecurity and Infrastructure Security Agency no longer has the capacity to help utilities, banks and other critical infrastructure operators prepare for a coming wave of AI-fueled cyberattacks.

CISA takes backseat in White House AI cyber response

I just read AI Agent Traps which has given me a more expansive reference for a set of cyber risks.