Check-In Vol. 1: Fable Dropped, Twice

AI is in the news every day. Valuations, model benchmarks, funding rounds, existential debates. Most of it is written for investors, researchers, or people who care about which AI lab wins. Very little of it is written for the hotel operator, the property manager, or the resort director trying to figure out what any of it actually means for their business on Monday morning.

Check-In is my attempt to fix that. Each week I pick five AI developments I think matter, and I translate them from hype into something operators can actually use. No VC framing, no doomsday scenarios. Just what happened, why it matters for your business, and the question you should be pressing your vendors on.

I started Conduit because hospitality runs on human connection, and these models are getting genuinely good at that. But none of it matters if operators don't have the context to make smart decisions about which tools to trust. That's what this is for.

— Punn, Co-Founder of Conduit

---

This week: the AI labs race to Wall Street, a new messaging channel, a shift in how guests search, a compute infrastructure bet that changes the industry's long-term trajectory, and the model that launched then was blocked within a week.

---

1. The AI labs are all going public

What happened. In the span of eight days, the two biggest AI labs filed for IPO. Anthropic filed a confidential S-1 on June 1 at a $965 billion valuation. OpenAI followed on June 8 — last valued at $852 billion, with analysts expecting a public valuation above $1 trillion. This comes as SpaceX, now merged with xAI, is already on its IPO roadshow at $1.77 trillion. The three companies that have defined the current AI era are all going public at the same time.

Why it matters. The AI tools you're evaluating are no longer scrappy startup bets. They're infrastructure. When the companies building the models go public, they get capital, enterprise accountability, and pressure to be reliable at scale — none of which small players face.

For hospitality, this means two things. First, the vendors you choose today may be running on models that become regulated, audited public utilities within a few years. Second, any AI tool that isn't built on top of a well-capitalized foundation model is a consolidation risk. The midfield is clearing out.

The one question to ask your AI vendor: Which foundation models are you built on, and what's your contingency if that model's pricing or terms change post-IPO?

---

2. Meta's Business Agent is live on WhatsApp

What happened. Meta rolled out AI-powered business agents natively inside WhatsApp. Guests can now interact with a business — book, ask questions, get follow-ups — without leaving the app. No link. No redirect. No download.

Why it matters. WhatsApp has 2.5 billion users. In markets like Southeast Asia, Latin America, the Middle East, and Southern Europe, it is the communication layer. For a resort in Bali or a villa operator in the Algarve, this is the moment the platform your guests already live in became a booking and service channel.

The catch: Meta's agent is generic. It doesn't know your property's cancellation window, your early check-in policy, or that Room 12 has a view worth paying extra for. The operators who win won't just activate the channel — they'll make it feel like talking to a knowledgeable concierge, not a chatbot reading from a FAQ.

What to do now: Audit your top-10 guest origin countries. If any are WhatsApp-dominant, move this up your Q3 priority list.

---

3. 37% of your future guests don't search anymore

What happened. A new Skift Research study found that 37% of travelers under 45 now start their trip planning with an AI assistant rather than a search engine. That number was in the single digits two years ago.

Why it matters. Your SEO strategy was built for Google. The traveler who types "boutique hotel near Amalfi Coast" into a search bar is a different creature from the one who asks Claude, "Plan me a 10-day Italian road trip for two in September, wine country preferred, under $400/night."

That second query doesn't return a list of links. It returns a curated recommendation — with your property either in it or not. Being findable in the AI era is a different skill than being rankable in the search era. It favors properties with accurate data, genuine reviews, and the technical hooks that let AI systems actually reach them.

The shift in one sentence: Search rewarded pages. AI rewards properties.

---

4. SpaceX went public as an AI infrastructure company

What happened. SpaceX IPO'd on Nasdaq at $1.77 trillion — the largest IPO in Wall Street history. The S-1 filing doesn't describe a rocket company. After the February xAI merger, SpaceX now controls the full AI stack: Grok models, Starlink global connectivity, terrestrial data centers for training, and orbital AI compute satellites (the AI1 constellation) launching in 2027.

Why it matters. A single entity now controls the model, the connectivity network, the training hardware, and the global delivery infrastructure — and it has $1.77 trillion in market cap to defend. The competitive dynamics of the AI industry just shifted.

For hospitality operators, the immediate implication is less obvious but more important: the cost and availability of AI compute over the next five years will be shaped by infrastructure decisions being made right now. The operators whose AI vendors depend on a single infrastructure player are quietly accumulating concentration risk they haven't priced in.

The longer view: orbital inference — AI computation delivered via satellite rather than a cloud data center — becomes viable as early as 2028. For remote properties, island resorts, or markets with poor terrestrial connectivity, this isn't a science fiction scenario. It's a procurement consideration.

The question to ask your AI vendor: Which compute infrastructure does your platform depend on, and what happens to your pricing if that infrastructure consolidates?

---

5. The world's most capable AI model launched, then was blocked overnight

What happened. Fable 5, Anthropic's most capable model to date, launched earlier this week to broad public access — a significant step up in reasoning and instruction-following from anything previously available. Days later, the US government issued an export control directive suspending access to Fable 5 and Mythos 5 for all foreign nationals, including foreign national Anthropic employees. The stated reason was a potential jailbreak method. Anthropic reviewed the technique, found it exploited minor, already-known vulnerabilities, and called the ban "legally unsound" — but is complying. Separately, Microsoft limited employee use of Fable 5 over data retention concerns.

Why it matters. This is a stress test that most AI vendors — and their customers — weren't prepared for. A model that launched to broad access earlier in the week was suspended by the end of it. No transition period. No advance notice.

For hospitality operators, the question isn't whether this specific model affects your specific vendor. It's whether you know what your vendor does when the frontier model becomes unavailable. The operators who discovered the answer this week are the ones who had no fallback.

The model governance risk has three layers:

• Regulatory suspension — what happened with Fable 5: a government directive removes access overnight
• Vendor lock-in — if your AI platform is built directly on one model with no abstraction layer, any disruption to that model is a disruption to your guest experience
• Data retention exposure — Microsoft's concern: what does your AI vendor retain, and under what jurisdiction?

What good looks like: Your AI vendor should have a multi-model architecture with documented fallback paths. You should know, before an incident, which model you'd drop to and whether the guest experience degrades gracefully or fails completely.

---

What this week adds up to

Take these five signals together: AI infrastructure is consolidating fast, regulators are moving, and the difference between operators who've thought about resilience and those who haven't is starting to show.

The Fable 5 suspension and the SpaceX IPO are superficially different stories. But they point at the same thing: the open-field phase of AI is over. Capital is concentrating. Governments are acting. The vendors that survive the next few years are the ones with durable infrastructure, not just impressive demos.

For hospitality operators, the window to make considered choices — rather than reactive ones — is narrowing. Ask the hard questions of your vendors now, before an incident forces the conversation.

---

Check-In is Conduit's weekly digest for hospitality operators — five AI signals, translated into what to actually do about them. Published by Punn Kam, Co-Founder of Conduit.