We’ve Entered the Antebellum Era of AI
Re-publish from Medium
— — -
Everyone in the Bay was building claws last week…
— I raced home & rewired my smarthome instead.
I spent this week in the Bay Area. Every event, every dinner, every hallway conversation was full of people building claws — Claude, Cursor, wrappers on wrappers, agents that call agents. The ambient energy was frantic and familiar. It reminded me of 1996, when everyone was suddenly building websites, most of them terrible, all of them consequential. I know because I was there — I founded Tag Media around that time, building NASA’s website among many others, before we were acquired by Omnicom and merged into Razorfish. We were inventing the internet. A history that is now repeating.
This week, somewhere around day three, probably over bad conference coffee, I coined a new phrase:
The antebellum period — the years before the American Civil War — wasn’t a time of ignorance. People knew something was coming. They could feel the tectonic pressure. They just kept building the old structures anyway, because what else do you do when the rupture hasn’t arrived yet? You optimize the plantation, you expand the railroad, you print the pamphlets.
That’s us. Right now. Everyone knows the next wave is going to break differently — ASI, AGI, reasoning models that can actually run companies, tools that make today’s “AI-powered” apps look like AOL keywords. And yet here we all are — building claws, training on vibes, shipping dashboards.
- Guilty.
What matters in every pre-rupture system isn’t what’s being built, but who captures the value. The real leverage sits with whoever controls the memory, the context, and the rights to act on what gets produced — not the agents doing the work, but the systems that own their output. This isn’t about better tools; it’s about who owns the layer where labor compounds.
The difference is: some of us are building infrastructure that will survive the rupture. And some of us are just decorating the antebellum house.
> “Everyone knows the next wave is going to break differently. And yet here we all are — building claws, training on vibes, shipping dashboards.”
> — Titus, March 2026
— — -
Meanwhile, back at the Manor house
I raced home from the Bay last night — not to decompress, but to update my Claude Cowork setup with a new outpost skill so I can text it further instructions from wherever I am. Which, honestly, is a perfect microcosm of where I’m operating right now.
Kimono has shipped more code features and updates in the last 60 days than in the entire six months before that. Combined. That’s not a marketing line — it’s the kind of number that makes you stop and ask what changed. The honest answer: the tools changed. Not just what we’re building with, but how we’re building. AI-native workflows compound in genuinely nonlinear ways.
60 days of shipping = more than the previous 6 months combined.
200 lapsed users reactivated in 10 days.
The reactivation campaign is a story I keep telling because it’s proof of concept for Kimono’s entire thesis. I ran an entire marketing campaign — ideation, segmentation, copy, sequencing, execution — using Airtable, ActiveCampaign, and Claude Cowork. No agency. No growth team. Just me, a well-structured contact database, and a sufficiently stubborn disposition. 200 lapsed users came back in 10 days. I’m pushing to connect Kimono to Topmail.so to make it even better.
That’s not AI hype. That’s a Founder doing what used to require a team of five plus interns, in a fraction of the time. The tools aren’t replacing the judgment — they’re compressing the distance between the judgment and the output.
— — -
Even the Catastrophic Stuff Got Fixed
Our head of AI, Sutha, found a bug recently that could generously be described as “existential.” A 32-bit integer primary key on our busiest table — a 2.1-billion-row time bomb, from some early engineering choices, left for someone to find. Someone being Sutha, at 5 PM on a Friday, staring at an AWS RDS console that wouldn’t tell him anything useful. It got solved. He wrote it up on Substack, and you should read it — not just for the fail/bug, but for the broader argument he makes about what AI agents are doing to managed cloud infrastructure. Short version: AWS might be Novell NetWare, and the CNE certifications are being retired as we speak. Food for thought, even if I’m not sure I agree. Amazon was early in AI and might have a plan, but RDS sure doesn’t show it today.
When your debugging loop runs at the speed of your discovery loop, the whole character of engineering changes. You stop sitting on problems. You stop accumulating technical debt like sediment. Things get solved because the cost of solving them has collapsed. Sutha’s piece is the CLI argument in long form, and it’s one of the sharpest things written about infrastructure this year.
🦞 🏯 🔧
— — -
I Also Reprogrammed My House
This one is partly personal indulgence, partly a data point I can’t stop thinking about.
Some of you know I spent time at Samsung leading smart home initiatives and led, amongst others, the teams behind the smart fridge and the FRAME TV with Yves Behar. That era left me with a fairly elaborate Matter-over-Thread smart home setup. It’s either a blessing or, depending on the week, a sophisticated form of domestic suffering that may have contributed to my second divorce
I have Claude Co-work re-programming my Matter-over-Thread smarthome with my new Homey… you should see the new automations with my Frame TV, ecobee, and lightswitches, plugs, and the aquarium feeder when I’m out of town.
Last week, I reprogrammed the whole thing. Not just tweaked — rebuilt the automation logic, restructured the device mesh, rethought the whole architecture.
It took an evening. With AI assistance. Something that previously would have required a dedicated weekend and Stack Overflow fury-scrolling.
That’s the antebellum tell: the thing that used to be a project is now a Tuesday. The question is what you do with all those recovered Tuesdays.
— — -
What Kimono Is Actually Building Here
All of this — the contact reactivation, the shipping velocity, the cowork setup, even the house — is a live proof of concept for what Kimono is building. We’re not just a privacy-first AI platform. We’re infrastructure for the moment when your data stops living in a dozen fragmented silos and starts working for you.
The infrastructure question isn’t abstract — it’s the whole game. The “Plaid for PII” framing gets at it. Plaid didn’t replace your bank — it made your financial data *interoperable*. Kimono does the same for identity, relationships, and personal data. In a world where everyone is building AI-powered everything, the question of who controls the underlying data becomes the question. We have a strong opinion about who that should be.
Kimono sits at that layer — not just an interface, but living, evolving memory. The continuously updated graph of who you know, what matters, and how it all connects. In a world of stateless tools and disposable outputs, the advantage shifts to systems that remember, that carry context forward, and that can act on it. That’s the infrastructure — not just generating answers, but owning the layer where relationships, data, and decisions compound.
In addition to our growing user community, Kimono now has a generalist AI chat. DM me to join. It’s early, it’s good, and it’s a taste of what’s coming. We have 4k on a waitlist for the product, but we prioritize chat members who are smart and contribute.
Our first prototype, a digital business card replacement, is Kimono.ID is infrastructure, not decoration. Your data, working for you — before the rupture arrives. See it live and connect with me: kimono.id/Titus-public and join the waitlist.
— — -
I Lost an Argument & I’m Telling You About It
A week ago, Sutha, Alek, and I had a proper debate about whether Kimono should build an MCP — a Model Context Protocol integration, for the non-technical readers. My position: MCP, obviously. It’s the hot thing, everyone’s building them, they’re flashy and demonstrable and easy to explain at a conference. VC’s are asking…
Engineering voted unanimously for CLI. Command line. Machine to machine. The quiet pipe that nobody talks about at conferences and everybody actually uses.
I didn’t go quietly. I stayed up late with Claude, Perplexity, and OpenAI — yes, all three, at the same time, like some kind of insomniac AI council — trying to build the case that I was right. Marshaling arguments, running scenarios, looking for the angle that would flip the vote.
I sent a concession text the next morning.
Here’s why it matters enough to put in a blog post: CLI — machine-to-machine, agent-to-agent, silent API infrastructure — will be at least half of Kimono’s revenue within the next year. Not the chat UI. Not the consumer app. The pipe. The part that other systems call when they need verified identity data, relationship context, and permissioned personal information without asking a human to click anything.
That’s the antebellum argument, right there, inside our own Slack. The visible, demonstrable, conference-friendly thing versus the infrastructure that actually moves the money. Sutha and Alek saw it first. I needed a late-night AI tribunal to catch up.
The good news is: I’m now the CEO who listens to his team and only loses arguments once per topic. That’s the bar I’m setting for myself, and I’m keeping it.
— — -
So, What Do We Do in the Antebellum?
Build infrastructure, not decorations. Move faster than you think is appropriate. Don’t confuse the tools that exist today with the ceiling — we’re still in the foothills. And pay attention to who owns the data underneath all of it, because that’s where the next war gets fought.
The claw-builders in the Bay were full of energy, and I don’t mean this as a dig. Some of them are building the railroads. Some are painting the parlors. It’s not always obvious which is which from the inside.
I think I know which one Kimono* is. We’re building the thing that survives the rapture.
More soon. 🦞
— — -
Titus
Richard D Titus is the founder & CEO of Kimono. A serial entrepreneur he also previously served as Controller, Future Experience at the BBC (iPlayer, Mobile service), co-founder of Razorfish (LA), founder of Schematic, and the ex EVP of product experience (Frame TV & Smart Homes) at Samsung Electronics. He writes here and at rxdxt.substack.com and is [@richardtitus]on X, and @rxdxt on Instagram and most other places (if not Titus).