Crank.report Survey

Building with an AI swarm

I learned about AI swarms over the weekend. It’s a group of specialized AI agents that are coordinating on a centralized task. Here’s an example that provides a good idea of what is possible: A Practical Experiment in Building an AI Agent Swarm.

The job I set my swarm to was something that we’ve been trying to get over the line at work but the conceptual hurdle of building within our system was too much. It’s not an unusual problem for an organization that has been working for years, diligently doing work in a specific domain, and building up – in equal measure – rich context and knowledge along side legacy tech that abstracted that into a set of technologies that become rigid overtime.

So a basic description of what I wanted: a way to get our context into a verbose knowledge base. That is, a knowledge base written for agents, for reforming and republishing, for integrating into other systems – rather than a knowledge base written for humans. I also needed a way to organize this knowledge base so using it is as efficient as possible, reduces the likelihood of hallucinations, and can be reviewed regularly for errors by both agents and by humans. Oh, and I also wanted to display a set of information as a filterable web page.

My swarm included:

• a planner agent, this functioned to help orchestrate and make sure I was keeping everything documented along the way
• a discovery agent, who searches the web to find new leads and information relative to my project
• domain agents, who each have expertise in different functional areas
• a validator agent, who checks links and schema to make sure things are correct and work across the project

These agents are each represented as simple markdown files. That is, files I can read, review, and edit. That means they stayed in my domain area and I can easily alter them without spending tokens. Just with fingers on a keyboard. Just like I’m writing this post.

The agents went out and found leads. Then the domain experts researched the leads assigned to them and wrote markdown files describing the results of their research. These markdown files became the heart of the system. Specific portions powered the web page. The rest became the information that I could use for a variety of other projects.

I learned a lot in this process:

1. How to organize the agents so that I didn’t have too many but did have enough to be able to set off specialized tasks.
2. How to organize the markdown files. I’m using skills files as the model for this. It will let me build out more assets for each of the areas, things that can go beyond what’s in the markdown files when needed.
3. Just how far I could get with the tools available to me: I was using Antigravity and then Gemini CLI when I ran out of Antigravity tokens.
4. How to use change logs and other related assets so that I can review the work and, based on the reviews, improve my agents and process generally.
5. How to think about a road map in this agent swarm world. Not too different, actually, from thinking about building out a staffing plan.

I also learned what I don’t know:

1. How to choose between markdown and JSON. You could make an argument that my knowledge base pages could be in JSON rather than markdown. I realize that I’m not exactly sure how to play out the pros and cons.
2. How to think about adding agents over time. Do I want case study agents that build content to augment the verbose knowledge base? Do I want other associated assets? Do the domain agents do those?
3. How do I build for scale? I could make a rapid prototype but if I though 100 people were going to hit that page, 1,000, 10,000 what changes and what do I need to do?

The next step isn’t technical; it’s human. I need to share this prototype in a way that builds excitement for augmented productivity, rather than triggering the natural anxiety of displacement. In an agent-powered world, the staffing plan changes. The core leadership task remains the same: building alignment, fostering confidence, and ensuring the team has the agency to drive the new tools forward, rather than being driven by them.

This should be front of mind for every CTO, CFO, and head of operations reading this. Because when the pricing corrects, and it will, the companies that treated AI as a permanently cheap utility are going to wake up to bills that make their current SaaS spend look quaint.

Every AI Subscription Is a Ticking Time Bomb for Enterprise

This exactly what we need to worry about in the nonprofit sector. How do we use AI in a way that accelerates our work and impact and proactively manages the expenses.

AI vendors are clearly the “new new gatekeepers”. Like the previous ones, they will dominate how we learn about the world even while some of us turn to open source and liberatory alternatives. But they may not dominate how we connect and share our experiences of the world, and that’s the core of the opportunity: how do we design pro-social frameworks and spaces that sit alongside an agentic information ecosystem?

Notable links: May 15, 2026

And how do help people know the difference?

This year’s Key Recommendations in this report convey fundamental best-practice expectations for social media platform safety, privacy, and expression. As we have done for the past six years, GLAAD continues to implore social media companies to meet these basic best practices: improve content moderation; provide meaningful transparency; respect data privacy; demonstrate commitments to workforce diversity and civil discourse; and strengthen and enforce policies that protect LGBTQ people from hate, harassment, and disinformation — while also not suppressing LGBTQ content and creators.

Executive Summary – 2026 Social Media Safety Index | GLAAD

Read this with yesterday’s share on content moderation as infrastructure in mind.

Papers that are more difficult to read might be worth it if AI increased the amount of good science being produced. But this doesn’t seem to be the case. Organization Science is desk-rejecting (e.g., rejecting a paper before even sending it to peer reviewers) nearly 70% of manuscripts that made heavy use of AI. This number drops to 44% for papers written without AI.

Easy Is Overrated

Can we do this kind of analysis on grant submissions?

Without uncertainty tolerance, we risk getting stuck at what statisticians call a local maximum. Like a mountain climber standing at the top of a hill, unaware of a taller peak just out of sight, our discomfort with uncertainty can keep us wedded to a business strategy, a job, or a relationship that is safe, but not optimal. Uncertainty tolerance allows us to persist through ambiguity—and it has never been more relevant.

5 Questions to Help You Navigate Uncertainty

I refer to this kind of uncertainty as navigating with a compass, not a map. I will read How To Not Know — this post is by the author — to see if it adds to the toolkit.

Moderation is infrastructure. It must be built like roads and power grids–with deliberate choke points, fail‑safe valves, and capacity limits. In road engineering, it is the ‘design speed’ of the road that users drive, not the posted speed limit; the tighter the curves, the shorter the sightline, the more measured the pace. The same is true of platforms–architecture sets behavior more effectively than policy.

What We Will Refuse to Build Again | dangerousmeta!

How do we architect for thoughtfulness and meaningful engagement?

To the mothers who shaped me.

Especially the three in this photo.

Communities frequently fail children across five sectors of their lived experience: ages of criminal responsibility, juvenile detention, child labor, immigration enforcement, and foster care. Policies in each area combine with economic and social conditions to limit opportunity and perpetuate harm. Examining these systems side by side reveals a pattern: children most at risk are those whose families, schools, and communities cannot buffer against structural deprivation. International comparisons demonstrate that the U.S. approach is a policy choice, not an inevitability. Countries like Norway and Sweden prioritize education, family, and social services rather than criminalization, showing that alternative paths are possible, practical, and effective.

Criminalizing Childhood: When the Justice System Fails America’s Youth - CounterPunch.org

This outlines the system’s change we need to work toward.

Ultimately, why not just build a “meta-paper,” using AI, to answer any possible question about the subject area under consideration. This meta-paper would allow the reader, using AI, to make many sorts of modifications and additions to the basic work. The meta-paper also would allow the reader to add new data, to run additional robustness checks, and to do whatever else you might think of. Once again, the canonical version of the paper evolves away.

[Will AI kill the research paper? - Marginal REVOLUTION](marginalrevolution.com/marginalr…

Thoth stores durable knowledge as entities and typed relationships, not just chat snippets. It can save, search, link, explore, visualize, and export your knowledge graph as an Obsidian-compatible wiki vault, while background extraction and Dream Cycle refine duplicates, stale confidence, missing relationships, and actionable insights.

GitHub - siddsachar/Thoth: Thoth - Personal AI Sovereignty. A local-first AI assistant with integrated tools, a personal knowledge graph, voice, vision, shell, browser automation, scheduled tasks, health tracking, and messaging channels. Run locally via Ollama or add opt-in cloud models. Your data stays on your machine. · GitHub

I’m using LM Studio for most of my local AI work. Haven’t quite graduated to doing anything more sophisticated.

Being successful with this approach to coding agents hinges on a rather crucial element: only a skilled developer who’s thinking critically, and comfortable operating at the architectural level, can spot issues in the thousands of lines of generated code, before they become a problem.

Agentic Coding is a Trap | Lars Faye

It’s like developers are now experiencing what anyone who has paid for a technology project feels.

More New-To-Me Photo Gear

At the East Bay Photo Collective Gear Sale, I bought a Nikon FE and Nikkor-Q Auto 135mm f/3.5 lens. It’s my second Nikon FE, the first I bought about 35 years ago. Anyway. New gear meant another photo walk. This one was in Berkeley’s Aquatic Park. I am learning the lens which requires using the depth of field lever, something I’m not used to doing. You can see the trial and error in a few of the shots.

These skills are designed to be small, easy to adapt, and composable. They work with any model. They’re based on decades of engineering experience. Hack around with them. Make them your own. Enjoy.

GitHub - mattpocock/skills: Skills for Real Engineers. Straight from my .claude directory. · GitHub

Interesting model to apply inside a workplace and in the nonprofit sector broadly.

But corruption also has a broader meaning. It can mean the malign use of power, the substitution of the public trust for your own private will, your own private interest. And that is more than anything else what is happening with the Supreme Court.

Jamelle Bouie via kottke

Advocates say that if made official, the new rule will prevent scores of unhoused transgender people from accessing safe emergency sleeping arrangements.

Trans people could be banned from homeless shelters soon

Another inhumane action.

Next time you’re in a design review, before you add your two cents, ask yourself the question: does this help the user, or does it help me?

Your website is not for you — Websmith Studio

This.

All six forms of psychological debt have the potential to cause damage to organizations pursuing AI integration strategies in a variety of ways: degrading employee motivation; corrupting collaboration and innovation efforts; and creating higher levels of stress and burnout.

The Psychological Costs of Adopting AI

Keeping team member agency while integrating and using AI is very important.

Taking human experience, which is beautifully ambiguous and nuanced and nondeterministic, and trying to fit it into a database shape, is inherently extractive. Nilay points out that it flattens people, which is totally true, but it also transfers ownership of that experience from their subjective truth into a centralized database that someone else controls, sets the standards for, and profits from.

Notable links: May 1, 2026

This is balance nonprofits need to struggle with as they understand how and where to use AI. How do we keep the texture of experience? How do we keep the agency of it in our community?

The beautiful thing about communities and platforms like Flickr is that they remind us that not everything on the internet has to be ephemeral, not everything on the web has to be hyper-commercial.

Why are the Artemis II photos on Flickr? - Anil Dash

Flickr was — is — one of my favorite corners on the internet.