The community-first software era

I’m at the PublicSpaces Conference in Amsterdam this week. This year’s theme is “why a digital public space is essential for a healthy democratic society”. This is an idea space I care about deeply and I feel very privileged to have been invited to contribute.

I’ll be giving a talk to kick off the unconference today, and then tomorrow will be on a panel about Social media, Journalism & Democracy:

Can journalism use open social networks to combat disinformation, hate speech, and news fatigue, and free audiences from broligarchical algorithms, while still earning a living? What is the role of politicians, representatives, legislation, and the enforcement of judicial agreements? And what examples should we look to for inspiration and guidance? This panel dives deeper into the Open Social Web and the possibilities for journalism to contribute to the well-informed citizens needed to uphold and strengthen our democracy.

In September, I’ll move on from my role as Senior Director of Technology at ProPublica in order to spend a year more deeply exploring ideas that I think are important at the intersection of journalism, technology, and democracy. I’ll write more about this later, but a year to consider what is important is a gift.

In the spirit of the conference and my new direction, I thought I’d write about some threads I’m interested in pulling on right now.

We’re in the fastest era of technological change in decades. Consequently, we’re also in the fastest era of journalistic change — and possibly the fastest era of democratic change. Given that context, what’s going to be important?

Here are some ideas. In particular, I’ve been thinking a lot about how encoding and building on technology protocols (the rules that dictate how software interacts) and human protocols (the processes and norms that dictate how people interact) could help us build software that reflects our needs and values.

Communities build trust and loyalty

If you work in news, you could be forgiven for thinking that AI is ruining everything.

We rely on good journalism to understand the world around us, but those newsrooms need engagement, subscriptions, and membership to survive. AI intermediates those things by providing an aggregated, summarized view of its source material, and it fills social networks with slop. Newsrooms are seeing sharp declines in referrals from search and social. They’ve been investing in email newsletters because those feel closer to building a direct relationship with their readers, but AI is coming to intermediate the inbox, too.

At the same time, many incumbent newsrooms have failed to adapt to meet their readers where they are or to represent contexts and perspectives that are recognizable to them. Trust in news has significantly dropped in a short time period. Some legacy newsrooms maintain an almost bloody-minded culture of not changing, and have not invested in understanding who their readers actually are. It’s a mindset that was established when print was the only game in town — but the internet is a conversation, and that’s what modern readers expect.

The biggest exceptions are local news startups, which are building trust, evolving business models for journalism, and building far more representative editorial rosters. Most of all, they’re engaging with their local communities. Their constituents know them; that representation and those relationships are how trust is built. And their readers are more loyal because they know they can’t get the context and information they need from anywhere else.

Newsrooms rely on something called a “callout” when they want to learn more from their readers. More often than not, this is a simple web form: “Has your doctor pushed this prescription medication? Let us know.” But instead of a two-dimensional form, what if we built a short-term community space that safely brought readers in and allowed them to discuss in more depth with the journalists?

My bet is that two things will happen: the journalists will get better information, because it will arise in conversation, and those readers will build stronger, more transparent relationships with the newsroom. And stronger, more transparent relationships will lead to more trust and more loyalty.

The key, though, is short-term. Each space is ephemeral. Once the community outlives its usefulness, it disappears. Each one is tailored for its need: rather than a one-size-fits-all social media space, the features and design are adapted to the question or the context at hand.

The new generation of open social web protocols help us here: if these community spaces are built on AT Protocol, both the newsrooms and readers are able to keep those contributions and relationships even after the community space itself has vanished. The reader has independent control; messages are saved to a user’s PDS. AT Protocol gives us tools for identity and user-centered data persistence that we can use as core building blocks. Private data over AT Protocol is in the works, and will be helpful, too.

These aren’t group chats and they’re not long-lived social media sites or social networking instances. It’s not just about the content; the form of these communities, which is inherently self-contained, is important to have more focused activity and to build both trust and safety. Each one is different, which means focusing on how technology can improve user control and safety, but not how it might abstract or generalize social interactions across them. That also means these spaces probably shouldn’t be federated: if the conversation is removed from its form and context, it loses much of its meaning.

And, yes, AI could actually help here too — although none of this relies on it. Newsroom teams could create ephemeral communities from building blocks using natural language, and through use of underlying open social web standards, reusable code, and plug-in services for functions like trust and safety, know that the resulting community space will be trustworthy.

Beyond news, I bet that ephemeral, tailored community spaces can support lots of needs for lots of different kinds of organizations. Through facilitating genuine, transparent connections between people, they will help to build trust and loyalty in a world where those things are broadly in decline.

I think there’s something valuable here and I’ll be actively thinking about it over the next year. If this is something you’re interested in too, let’s chat about it.

Culture is critical but our tools don’t know about it

Culture is core to any organization. It governs the norms and conventions that dictate how work is done, how people communicate with each other, what is tolerated, who is welcome, who is accountable, and whose ideas can be heard.

But how organizations build culture is wildly inconsistent. Quite often, leaders will focus on the outcomes of their work — the product in a tech company, the journalistic process in a newsroom — but under-invest in the culture of the organization that gets them there. They tell themselves that they don’t need HR or a people officer; norms and formal processes that determine how work gets done will just get in the way.

The result of that under-investment is typically that people are unhappy, fewer ideas are heard, friction builds, and the actual work of the organization falls short of its potential. New ideas and hard truths aren’t heard by leadership, perhaps because people don’t have the safety or the avenues to speak up, or because leadership hasn’t established the norm that they will listen. Under-investing in culture always results in lost opportunities.

Regular readers will know that I often recommend Corey Ford’s writing and coaching. He’s been influential in building my understanding of how central culture is in an organization’s success. One of his superpowers is boiling down sophisticated ways to build culture into repeatable mantras that make them easy to follow — which, in fact, is one of the tools he advocates for.

But here’s the thing: mantras are protocols, too.

The productivity tools we use are one-size-fits-all. Yes, you can build custom workflows into tools like Jira, but it’s cumbersome, and the result superficially represents the way the organization works, if it represents it at all. Some tools, like Salesforce, have built an entire cottage industry around customization: you either need to be an expert or you need to hire a consultant and spend tens or hundreds of thousands of dollars.

Because we all use these tools to do our work on a daily basis, the way they work dictates the way we work. The assumptions made by their authors become deeply ingrained in our own organizational cultures — particularly if we haven’t done the work to establish a strong culture ourselves. We import their values, assumptions, and cultures wholesale.

But they’re not applicable. A tool built for a tech company in Menlo Park should not dictate the culture of a newsroom in Alabama. That’s not necessarily a value judgment: they’re just different organizations with different contexts and different constituents and communities. Most importantly, they’re built for scale; the one-size-fits-all growth that doesn’t let you build deep, meaningful relationships. Adopting those values, even unconsciously, is one way a newsroom can lose trust with its community.

So instead, imagine a way that an organization could actually encode its values in ways most never have. Its leaders build mantras — atomic units of culture — that encode how they work, how they talk to each other, and so on. Consider Corey’s: make space for every voice, feedback is a gift, and one consultative decision-maker per lane, for example. Mantras could be available to pick from; organizations would write their own; some would make theirs available on an open-source basis for others to build on.

That mantra dashboard becomes readable by everyone in the company, which is an important step towards establishing shared norms and processes. That would be game-changing enough for most organizations. But it also becomes readable by our productivity tools, which read a machine-optimized version of each mantra in order to adapt the way they work to the way we want to think.

Here, for example, an automated system reads the protocol one consultative decision-maker per lane, and builds it into the design of a productivity tool. The system is set up with clear decision-making lanes that each have a single owner, but with facilities to share their thinking openly and consult others using the process defined in make space for every voice before arriving at a conclusion.

We’re in a world where everyone can roll their own bespoke software. Some organizations have the capability to do this with engineers; others will use AI, or generators, or some other means. This ability to create software that is more tailored for us gives us the opportunity to enforce and encode our values and norms in ways that include the tools we use.

The human work of building and establishing a culture must always come first. But our machines can now follow it, too.

What if software is Duplo now

Both of the above examples have discussed creating more bespoke platforms, often with the use of AI. I do think that’s the core way software is changing: the era of shrink-wrapped, one-size-fits-all products that are optimized for scale is coming to a close.

But it’s also true that any organization that thinks it’s going to vibe code its own tools is in for a world of hurt. There are unforeseen maintenance costs, hidden design considerations that expert teams have worked on for decades, research outcomes and underlying science. Building software involves multiple highly-skilled disciplines; building great software that really works is hard.

An LLM that has been trained on the outcome of all that work can cargo cult a software product, but it can’t reproduce the underlying skill. Unless it is itself in the hands of an expert, its work can never be as good. And for most organizations, vibe coded software will be spiritually the same as Microsoft Access databases in enterprises twenty years ago: these tools will proliferate invisibly, nobody will know how to maintain them, they will create privacy and security risks, and ultimately will create more friction than they solve.

Protocols live in the middle ground. Borrowing from Corey, think of them as mantras for how entities — people, software, networks — interact with each other. Each of them should start with a human-first need. They need to be deeply considered. Building them is deeply human work that must be informed by research, study, experimentation, and collaboration.

Some protocols are purely technical: AT Protocol and ActivityPub are great examples. But as I discussed in the last section, our norms and values can be encoded as protocols that dictate how software works, too.

Once we have them, we can use them as building blocks for new things. Skilled engineers, designers, and product teams can create more sophisticated software building blocks, too. And then we can combine them in ways that more closely represent the needs of our organization and context. Instead of using an LLM to build inherently unmaintainable software messes, we can connect well-built building blocks according to protocols and recipes that have been developed by experts.

The resulting software would fit an organization’s needs more deeply. Because those underlying protocols would be shared, different tools that are built this way could work together more easily.

This combination of protocols, building blocks, and recipes would have a mix of underlying models: some would be open-source, some would be proprietary. They would all allow for far more remixing, customization, and interoperability than we experience using cloud software today.

If building software by hand is like an industrial manufacturing process (or a movie studio), what I’m talking about here is more akin to building with Duplo. It’s also kind of the Unix philosophy, if you squint a bit, although that was solely about a modular approach to technical systems. Here I’m advocating that we turn human norms into protocols that help automated systems to build software based on our needs and values.

It’s undoubtedly less flexible than forging each atom yourself; it’s also safer, builds on the work of experts, and allows for far greater maintainability. Most importantly, it allows organizations to put their values first — and forces them to encode their cultures and assumptions. That’s a benefit in itself, not least because it allows both organizations and the software that supports them to be responsive to the needs and values of their communities.

The community-first software era

Technology companies can build our underlying software; they shouldn’t dictate our culture. But to change that dynamic, we have to define what our culture actually is.

Building and using software that better supports democracy means, in part, building and using software that better supports communities. If we are to do that, we have to be clear about what our values are, and we need to have mechanisms to build them into the tools we depend on. We can no longer depend on one-size-fits-all. Platforms can be bespoke, they can be ephemeral, and they can shift according to our changing needs.

New technologies like LLMs give us the ability to create and customize those platforms. For them to be effective, we need to ensure that what we build is safe and maintainable; that it leverages the right underlying disciplines and expertise; and that, most of all, it puts the needs of real people and real communities first.