Who are we Now?

This is based on my keynote at RubyConf 2026. They’ll publish a video eventually, I hope.

There’s a philosopher in Naples, Italy,  way back in the 1700s who understood something important to software today.

Giambattista Vico lived early in the Enlightenment, when mathematical and experimental thinking were all the rage. René Des Cartes and Francis Bacon were the hotness. People were all about, what can we know for certain through logic alone?

With this mindset, Viceo laid the basis for anthropology, ethnography, sociology, comparative law, historiography… a whole slew of fields. And he nailed my favorite feeling in software development.

Verum Factum: we can only really know what we made.

Giambattista Vico

When I write a program—or change it enough to be integrated with it—I understand it in a way that’s deeper than other tools. Like the software is an extension of me, because I know where and how it can flex, I can figure out what broke, and I feel powerful changing it. Vico named this feeling: verum factum. I hold the truths of this software, because I made it.

Verum Factum gives me confidence to ship to production. If something goes wrong, like people use it weirdly or an integration doesn’t work as planned, I know I can handle it. I understand the software’s language (I can read the logs).

This is software that’s alive and can change. As opposed to legacy software (as I used to define it), which is not alive in someone’s brain.

I used to call this the code-coder ecosystem, but now I call it a symmathesy. It is more than the sum of its parts, more than a product of set interactions. It is a learning system made of learning parts.

The software learns from me because I change it, and I learn from it when it runs: from tests that fail, from exceptions it throws, from data it changes. I learn what happens in the real world with real users. For this, the software has to run in production, and I need tools like tests and logs.

Scale this up to more developers, more services, way more users. Now I need stronger tooling. Continuous integration, real observability with aggregations and traces. Now the symmathesy includes all of these tools so that we can learn from the software. Meanwhile we learn from each other by talking. Maybe there’s a product owner or something who spends most of their time interfacing with users or the rest of the business, because we’re all part of larger systems.

As we add more people changing the same code, that verum-factum knowledge and confidence weakens, because I didn’t personally make all of this. But the team did, and collectively we can know it deeply.

To grow a symmathesy, it isn’t about adding. We make this symmathesy stronger by increasing learning between people–say, by pairing or ensemble–from people to the code, say by expressive code in domain language–and from code to people–say, with better tests or observability.

A symmathesy grows from the inside. To improve it, we have to be involved. There is no optimization from outside; there is no understanding it from outside. We are making it, we who participate, so it only makes sense from inside.

We don’t really know how other teams work, and management doesn’t either. Like when they say “why don’t you take that team’s procedures and then be the same?” and we look grrrr at each other.

I love that feeling, of being part of a symmathesy with other people, learning from each other and from what we’re doing. Of the code as an extension of myself, because I understand it, I can confidently change it. 

This is where I am a Symmathecist: I am improving this whole system from the inside, consciously.

I can be responsible: if something goes wrong, I know how to respond.

I can be accountable: I can give account, say why we did it this way.

It’s my favorite part of software development.

But now, agents! There was a turning point last year where I noticed that not only are they useful, it is irresponsible not to use them, at least in conjunction with my own code. They’re more thorough, as well as faster.

How am I supposed to be responsible for this system, when I don’t understand each line of code? How do I get that verum factum feeling back?

Who are we now, as software developers?

Here’s a long quote from Vico. It says: fortune is friendly to young men because they choose a profession that’s popular in their youth; but as tastes change, they later find themselves strong in such wisdom as no longer profits.

Is that us? Doesn’t have to be.

We can deploy confidently with agents. We can have symmathesy with agents. And we can get that Verum Factum feeling back.

One reason I know my changes are going to work in production is that I made them. Verum Factum, deep understanding from participation. We can understand law, society, and history with this. But Vico never said that was the only way to know a fact.

Two ways of knowing that our deploy is gonna work

There’s also Francis Bacon style. His idea was to learn about nature by putting it to the test! “Vexationes Artium,” artful vexations, like adding water to quicklime or stuffing a chicken with snow. He didn’t have the word “experiment” back then.

Verum Factum is something humans get because when we change something, we are also changed. That’s how a symmathesy knits together. Our work becomes part of us. 

Objective, reproducible, experimental knowing is different. For that, we stand outside the system and poke at it. This is great for any scope small enough to get specific.

The agent cannot have Verum Factum knowledge. It writes something, it has that in its context for a few minutes, then it compacts or restarts and that is gone. Sometimes it doesn’t recognize its own work even in the same session. It is not changed in the making.

The agent can totally use Vexationes Artium, experimental knowledge. From running a manual test to executing a property-based test suite, these can support or falsify our beliefs about the code. 

Skill at experimenting on code is great for people–because verum factum, even when it gives a feeling of certainty, is never perfect–and it’s great for agents. It’s HUGE for agents. It’s the difference between slop and engineering.

If we want agents to write working, reliable code for us, we have to double down, 10x down on our objective verification. We need to vexate that code in artful ways. And we have the agent help us with that, with its thoroughness.

Oh and then there’s software that incorporates LLMs in production. We have to get even more artful there. Evals become more crucial than code and prompts, since code and prompts can be re-created given good evals. 

So we turn our vigilance toil into more coding work. We build scaffolding, the code that helps us write and run production code safely. We can make so much of it now, and we need it!

More test code than production code? Absolutely. Instrumentation that doubles the length of a method? great. Part of its work is telling me, the developer, what it’s doing through the traces.

Then we automate all of these checks and reviews, including checking it again after it’s in production. And we get agents to help us with the checks and reviews, too. 

We build their world

Software development has learning loops at many levels. Look around, think of an idea, try it, check the results. TDD does this for a single change, CI for a feature. The full delivery loop includes deciding what to implement, and that’s still human-led in most software now. Agents are doing the smaller loops, now. It’s our job to make sure they’re learning—through their processes, context and direct feedback—enough to make our delivery loop smooth. We are in charge of their learning loops.

Note that since the model is not changed by writing the code, the LLM cannot participate in the symmathesy by itself. The unit of participation is larger:

It’s the LLM and the context you’ve constructed for it. From AGENTS.md and other documentation in the repo, to plugins with hooks and skills you have installed, to the tools you give it access to and the descriptions they come with – that’s all “the agent” for the purposes of learning systems made of learning parts.

I used to define legacy as “code that wasn’t alive in a person’s head”
Now I define it as “a codebase not set up for agents to work on.”

So if the agent screwed it up, what am I changing about its context or harness to improve the result next time?

Example: the agent took 8 commands to figure get this project running? [LINKS] Have it improve the documentation! OK now it’s down to 3 commands. Hanve it improve the scripts! Keep getting smoother.

That’s one way I still feel Verum Factum when working with agents: by tweaking the ways they work, watching when they’re having a hard time and changing their circumstances until they work better. I’m in charge of their experience.

We choose what to build

I’m also in charge of its goals. With the agent in the details, I get to be the architect, product owner, and lead designer. I define what “right” is. It solves the puzzles I write for it. There’s verum factum in this, too!

Side projects are freaking amazing right now. I can be like, hey, what if the home page had a bunch of color symbols bouncing around like gas molecules? And boom, it happened! Or rewrite this service in vanilla Ruby. Make it event-driven. Oh no, production is broken, dig around in kubernetes and fix it. 

There is joy in coding with AI, there is self-expression, at the application level. We don’t always get that at work, but hey, we can do some of the projects we always wanted to but didn’t have months to code. While learning about using AI to build software that AI can keep working with.

We learn from agents

When I realize I don’t understand how my software works, I cry and mourn my verum factum feeling… or I do something about it.

When coding isn’t the bottleneck, focusing on completing our task as fast as possible doesn’t help us. We need to be curious, and indulge our curiosity.

We can use AI to avoid learning the intricacies of the code, or to learn it even better. Take “I don’t know how this works” as an opportunity.

We can use AI to avoid asking our colleagues anything, or to ask them better, richer questions. Take “I wonder why” as an opportunity. Build relationships through easier questions, and the harder ones become easier to answer together.

AI makes some of our detailed-coding skills less precious than they used to be. And it demands new skills of us, like architecture and context engineering. It also makes most of these skills easier to acquire, if we use it for that. 

Cat Hicks found that AI is your enemy if you identify as “good at” things. But it’s your friend if you identify as “getting better at” things. That’s how symmatheses grow, too: always getting better at what matters right now.

So who are we?

We are still engineers.

We are not managers: prompting an agent is less like hiring that summoning a spirit.

And using agents is more like mushing sled dogs than coordinating a team. They never complain about work. They want to run, but getting them to run in a coherent direction that we want to go is a challenge.

Also, this is exhausting! Everyone else is pulling their sled really fast, the bar is higher every day, and there is so much to learn.

How do I make AI create ease & possibility for me rather than be a lever for someone else to burn me out even faster?

avdi grimm

We have to stop asking “How can I do this faster?” and ask instead, “How can we do this better?”

In our learning loop, it used to take a long time to TRY something in software. Now that part is fast. Agents can help some with checking results and looking around (doing research), and they can come up with a zillion ideas. They can’t tell us which ideas are valuable, nor which ones are aligned with the wider business we are working in. That part is on us.

[hartmut rosa stamp]

And background conditions are not stable right now, in the whole world and in software.

So we need more time talking to people, and more time being part of the system around us.

Verification can give agents confidence in the code we’re deploying. But it can never make them accountable. Nothing can make them accountable, because they can’t be held to account: they’re gone after they finish an execution. Nothing can make them responsible, because they’re not around to respond. We people stick around, we are part of the system ongoingly. We feel the consequences. 

Agents can participate in the symmathesy of the software team, but they cannot be Symmathecists. They cannot consciously participate in the growth of the whole system, because they aren’t part of the wider system, nor do they feel the consequences. They’re like the code: they can get better at their part, but that’s it. The team as a whole, and our team’s participation in our organization or company—this is ours.

Who are we now, as persons?

AI can’t be accountable, it can’t suffer consequences, because it is not conscious. What do I mean by consciousness?

Well, I like this widely-accepted definition:

There is something it is like to be me, and I can’t get away from that. My choices now affect what it is like to be me tomorrow. I am changed by what I do. That’s where Verum Factum comes from!

As people, we always change. And yet, is it just me, or is that change coming faster and faster?

Acceleration everywhere

Technology is always advancing. I can’t say whether it’s advancing faster and faster, that’s hard to measure (although software is, at least inside Honeycomb). It is established that technology adoption is speeding up by orders of magnitude in the last hundred years.

The telephone took 75 years, the internet took 7 years to be widespread, ChatGPT took 2 months. The expectation of learning and changing is… it’s dizzying. 

I find myself asking: now that AI is ubiquitous, where does that leave us, as persons?

I find myself looking for a narrative to fit in, because my expectations of life from a few years ago don’t fit anymore. And not a narrative that is built in the values of right now.

Vico had this talent: one of his innovations was looking at language and its evolution, noticing that in earlier times people didn’t think the way we do now, they had different circumstances, societies, concerns and patterns of thought. This led him to a different story than the Enlightenment’s story of incessant progress.

Vico’s opus, New Science, turns history from a series of facts into sweeping patterns. He drew human history as a series of ricorsos, repeating: from barbarism, to an age of gods when religion was strong, to heroes, to an age of men where reason ruled. Eventually men get so focused on reason that a “barbarism of reflection” arrives, religion is lost, and society breaks down. 

Vico’s first Age of Men starts around Aristotle, when reason became a virtue, a central pillar of humanity. Then this broke down with the fall of the Roman empire and we had the Dark Ages.

Vico’s second Age of Men has the Enlightenment. (Hartmut Rosa marks this as the founding of modernity.) People lost interest in history, or when they looked at it, they made it all about themselves. 

Taking it farther, we get the techno-optimists of today. 

[Andreeson stamp]

We have had technological progress ever since the Enlightenment. But it is not a steady progress. Instead, it is accelerating. 

If only technology was accelerating, then we could have more free time. We’d get all our coding done in a few hours and go hiking and play games six days a week. But no! The pace of life accelerates at least as fast. 

And to go with that, the pace of social change speeds up. A few hundred years ago, you knew what you were gonna do with your life because it was kinda fixed based on who you were born to. Then we got to choose our careers: “What do you want to be when you grow up?” My grandparents could go their whole life with one job.

Nowadays, one career doesn’t last you a lifetime. What do my kids want to be? Whatever the heck they can get hired for when they get out of school. And a few years later, whatever is the new hotness. Software development is not the only career that’s been utterly changed by AI, automation, the internet.

Technology, pace of life, social change: all increasing in pace relentlessly since the Enlightenment, the founding of modernity. This is Hartmut Rosa’s theory of social acceleration. (video)

The Enlightenment, when we realized again that the greatest function of Man was reason. When rational thought became the only kind of knowing that is respected. Vico’s work was pretty much lost outside of Naples for two hundred years. 

The Rational Animal is no longer special

Both humans and animals could evolve, grow, participate in communities. But only humans could use language and reason.

But now computers can use language and reason. It can prove mathematical theorems that we couldn’t figure out. This is not unique to people anymore.

So maybe it’s time to lean back into our uniquely-human (or uniquely-living-being) way of knowing. Maybe it’s time to value Verum Factum more than our modern culture does.

From Vico’s Naples, let’s go 100 miles north and 300 years into the present, and turn to a figure from the Age of Gods. In May of 2026, The Pope released an encyclical (some sort of official statement?) about AI. It’s really quite good, and beautifully written.

How to be More Human

Curiosity! We talked about that already.

Play! Our goal-driven behavior won’t lead us to new, surprising ways to be and work. 

Together! We grow, learn, and work better in groups. A person is a person through other persons. The more we can be people together, the more we can lean into our humanity.

For instance: can we not text chat in Slack all day? If I want reams of text, I know where to get it.

On calls, use your cameras! Communicate with faces and hands, not only words. Get animated, show emotion. Play music.

Draw. Communicate visually and in more dimensions in a shared Mural, Figjam, Honeycomb canvas, whatever. Share paper. 

Work physically together! I work remotely, and while I love my job, I wish it could be in person. Alignment is somehow much easier and richer in person.

When you’re together: share meals. Go for walks. Have routines and rituals. Do boring things together. Communication happens without pressure.

Two kinds of experience

In German, there are two words for experience.

Erlebnisse: Adventure! This is one-time experiences, where something happens, we see something new.

Erfahrungen: Ongoing life. This is repeated experience. We do it regularly, with the same people.

The weird part is: you think that trips and excitement are what change you. It’s the opposite: the regular stuff is what sinks in and becomes part of us. The adventures aren’t even what sticks in our memory!

Erfahrung is the experience that changes us. And experience that changes us is what agents cannot have. (I guess you could retrain them? but then it’s a new model, and that is not the course of work.)

After reading this (in Rosa, referencing Walter Benjamin), I notice that of all the new restaurants I’ve tried, I can remember only a few, faintly. The restaurants that we go to often, those I can bring to mind richly. 

What if repetition is more valuable, more richly human, than novelty? Then I want to change some of my strategies for life. I’ve gained more appreciation for the boring stuff, like taking kids to school or making dinner. I value travel less.

As persons in this age of AI, we are more than rational. We are experiencers (Erfahrung), and we are participants.

Work in the same place longer! And in the same industry. The wider system context that we absorb as people, the AI doesn’t have that. It only has the narrow information we fed it or it retrieved, and generic training data that isn’t suited to the situation. We are the full participants.

(I don’t like that conclusion either. I like switching jobs and industries and learning entirely new domains. But there it is.)

But! Even as we participate in our symmathesies of team and organization, work is not community. Most of what we call community these days, isn’t.

Two kinds of human association

In German, there are two words for associating with other people. These are extremes, and most real interactions fall somewhere in between. (This comes from Ferdinand Tonnïes.)

Gemeinschaft: community for relationship. We are in this together because here we are. This is based on family, place, religion, tradition, friendship—things that you can’t just swap out. This is where you can be yourself, because you’re there because of who you are. You can move or find a new friend group, but new ones are not as Gemeinschaft as where you’ve been for decades.

Gesellschaft: association for purpose. We are in this together because it somehow benefits you and somehow benefits me. This is transactional. Your participation is conditional on other parties getting what they want. An extreme: I buy something from you on eBay. This is useful for commerce, government, employment.

This tells me something else I don’t like:

Do not show up with your whole self at work. That is impossible for some and a bad idea for most of us. 

If we can bring our own imagination and perspective and humor safely and productively, great! In our teams we often find friendship, in our teams we are in the same circumstances–but that is contingent on a transactional relationship with the company.

And if your self-worth is based on your performance review, you’re doing it wrong. That is a review of a Gesellschaft association, of what the company wants from you, not of who you are.

We find real community in Gemenschaft. So here I am on a mostly-tech blog telling you to spend more time with your family. To get to know your neighbors and gather with them regularly. To appreciate your religious congregation, if you have one where you’re valued.

Because the tech culture we are in, at work and more and more in this country–it thinks Gesellschaft is the only association. Economics models people as in it for themselves. 

But the thing is, any personal utility doesn’t bring real happiness, meaning. That’s all found with other people. And it isn’t “I’ll give you meaning if you give me meaning”! it’s in: let’s watch this show together, let’s do this yard work or wash the dishes together, let’s help our neighbor clean out the sewage-filled basement together. That is Erfahrung, and it builds Gemeinschaft.

Our work has some meaning, but it is not who we are. It’s too Gesellshaft.

Let’s save optimization for the robots.

Who are we now, as humanity?

What does all this modernity, acceleration, and now AI make us? This is not determined.

Is AI bad for us?

My kids refuse to use AI because they are artists, so they care about artists, and AI steals from artists. Also, they say that AI will keep them from learning, because they see peers using it to avoid learning. My kids are in college, and they are going to be at a huge disadvantage in the job market because they don’t know how to use AI.

”Is AI bad for us?” is not a helpful question. AI is here. As people, we don’t get to choose our circumstances, most of the time. Every big change is usually both good and bad. Ours is to choose how we respond.

Buy art from artists. Donate to community theatre.

Care about the environment? Let’s tackle fossil fuel extraction.

Care about water use? Let’s talk to golf courses.

Data centers are probably more significant than plastic straws, but they’re still small. 

I think on some level, I want to find side-effecting reasons that AI is bad, because deep down I sense that it’s bad for us as people and as a society.

AI can be both good and bad. I like using it for coding and for learning and for generating silly images. Also, AI can accelerate the splintering of society that started with the internet. 

I thought the internet would be all good–and it’s fantastic, gosh I love watching a video on how to pot a cactus or fix my dryer. Buying Vico’s “New Science” on Kindle for $2 and reading it immediately.

Democratization of knowledge is not all it’s cracked up to be. All the information is no information! We need institutions that prioritize and vet, institutions we can relatively trust, that check each other. Without these, we all live in different worlds based on the information on our various screens. These institutions are in bad shape right now.

We have instant communication with anyone on the globe. I love having friends overseas and all over, being able to stay in touch with some of you after meeting at conferences. But that is not the same as someone who can pick me up at the airport, and help me with a flat tire. I don’t even have to like the neighbor or family member who helps me with stuff, it’s still meaningful. That’s Gemeinshaft.

The internet splinters our reality. Whatever beliefs you’re inclined toward, it will confirm them.  AI will splinter us further.

Do I think AI is bad for us? Sure. Does that mean I shouldn’t use it? No.

Systems change from the inside.

The world is not as I would prefer it. I get to respond.

Who are we now, as humanity?

This is not determined.

There is such a thing as responsible governance of AI.

It doesn’t have to leave millions out of work.

It doesn’t have to separate us opaquely.

It doesn’t have to own our data.

We are not helpless here! I may be helpless. You may be helpless. But we are not helpless. We are participants, the system is made of us.

[mary parker follett stamp]

Can we change the world? Not alone. And not if we sit on some of our lazily absorbed beliefs.

Politics

If we believe “All politicians lie, they’re all corrupt”—then we give up our power.

Politicians are not all the same. When some nasty ones spread this belief around in the 1980s, and it caught on, and now we’ve grown up with it—they won. This belief keeps good people out of politics, and it keeps bad people in, because we give up our power.

But not everyone! Beth Anders-Beck[LINK] for instance, running for Congress in Massachusetts right now (2026). My friend Caterina Clayton, [LINK] running for state congress in Missouri. Others [LINK to Ben Darfler?] who are elected and work on their city council. They want to make government work. Government can work, if we elect more and more people who want to govern well. Who are in this with us for Gemeinschaft: we are in this together, this is our country, and it is ours to improve.

Do all politicians make promises and then fail to keep them? Yes. That is not lying, that’s hope! Do all politicians act differently behind closed doors than on stage? I sure hope so. That’s communicating with your audience, and negotiation. Within these, A person can act with integrity, moving the world closer to those promises you want to reach, or for individual profit.

We can have nice things!!

Provenance of media
Liability for harmful content
Transparent, uncustomized algorithms
Fewer spam phone calls
Shorter lines at the DMV

Is it hard? yes. Will we measure the results of our efforts this quarter? no. This year? maybe a few. In our lifetime? for the most important changes, no.

Verum Factum means that we have the potential to understand our own society. We make this. Pay attention to what we are making.

Conscious participation, looking to improve and responding to what happens: this makes me a Symmathecist.

I strive to be an effective Symmathecist in my family, in my communities, and in society. Because

Authentic realism does not give up on changing the world.