What I’m thinking about: The world is moving faster every week…real estate is moving slower. Navigating that gap is where the next decade gets won or lost.
Roughly a month ago, the institutional capital folks I trust most were calling this the “worst overall market environment” they’d seen in 20+ years (taken as a whole, not just RE). Iran war freshly kicked off, consumer sentiment at its lowest reading since data started being recorded in the early 1950s, private credit in turmoil, stocks on the kind of volatility tear where the only honest answer to “what happens next” is “no idea,” and of course, arguably the most hostile market for real estate buyers in US history (see below, existing home sales lowest since 2009).
Fast forward to a week ago. Same contacts, sentiment up roughly 5% off that floor.
Not because anything got dramatically better (e.g. consumer sentiment is still the second lowest ever, just beating out last month’s reading)…but because the market has effectively thrown its hands up on trying to price every new piece of geopolitical noise day-by-day.
(This almost exactly mirrors a year ago when tariff policies began to be widely applied, and rapidly changed.)
Wars rarely have a long-tail impact on US markets anyway (we’re geographically insulated from kinetic conflict, and generally American markets are considered a global safe haven in times of uncertainty…a reminder that as rough as things may get domestically, the rest of the world is often facing even more difficulties.)
Meanwhile, the Magnificent Seven and AI hyperscalers (arguably the best-run companies in the history of the world) keep printing strong earnings, quarter after quarter, during a period when they were trading at some of their lowest multiples in years.
=====
Cleanest analogy I have for this is the news cycle.
Less than 20 years ago, you’d watch or read a sports recap the day after the game and that was fine. Today, if media personalities aren’t recording the moment the final whistle blows (if not before), they’re going to be outcompeted by someone who is. Attention spans and reaction loops have collapsed in lockstep, and the market (the biggest information aggregator on earth) is just following our collective shortened attention.
People are like, “Ok, bunch of fear and uncertainty about a potentially unprecedented breakdown in global order…[pause]…let’s just get back to making money again.”
(To be clear, the market is perhaps the most sophisticated instrument humanity has ever created, but it sure feels like the above, right? All the fallacies and emotions of human decision-making are baked in.)
Real Estate Is the Stubborn Exception
Real estate (the largest asset class in the world) is on basically the opposite trajectory, and that’s where RE investors/operators are feeling the disconnect more than anyone. We’ve been in a protracted buyers market across most of the country since mid-2022, with no clean bottom in sight (see the graph at the top again, a slow-walking trough). I walked through every major US real estate down cycle a few weeks back, and the current setup is genuinely unprecedented (mostly driven by a mix of high prices + higher equity cushion for current borrowers + safety nets prolonging the pain).
Stocks, news, and AI cycles (see below) are compressing into each other. Real estate just drags (not unexpected for less liquid assets)…that’s the tightrope dynamic we all have to continue to navigate currently.
The AI “Bubble” Conversation Just Shifted
For much of 2025, the loudest macro question was whether all the AI hyperscaler capex was a 2000s-style telecom (or late 1800’s railroad) bubble waiting to pop. Worth taking seriously (the largest companies on earth are deploying gobs of cash into compute, and historically when capex of that scale fails to produce productivity gains, measured in real revenue, the market inevitably faces a large correction).
Over the last few months, the conversation has been shifting from, “Do we have an oversupply of AI infrastructure,” to “Can the AI providers keep up with demand.” Tools like Claude Cowork and Claude Code (btw ChatGPT Workspace Agents and GPT-5.5 both dropped a few days ago, positioned squarely around agentic workflows) are now obvious productivity levers to anyone actually deploying them (recent 6-work-days-into-hours breakdown here).
Well-funded companies are spending hundreds of thousands to millions of dollars a month on tokens for AI compute (note Anthropic’s tripled it’s ARR from $9B to ~$30B in just the first 4 months of 2026), and the gap between the haves and the have-nots is going to increase rapidly.
An undeniable proof point came from Ramp recently (the rapidly scaling fintech now valued at $32 billion, that specializes in corporate cards and expense management).
Their team publicly shared their internal AI tool, “Glass” (built on Claude Code), which gives every Ramp employee a single role-aware, self-improving entry point to AI agents that already know the company’s tools, data, and workflows. (I highly suggest you take five minutes to read the post linked above for a non-technical breakdown on how Glass operates. It will transform your view of what enterprise AI is capable of.)
Critically, Glass collapsed the learning curve for AI, so 99% of Ramp employees use AI daily, substantially above the baseline most corporations are still at.
As soon as I saw Glass, I wanted to implement it, and was disappointed when I saw Ramp deliberately chose NOT to release Glass externally. They view their internal AI capability (smartly) as a moat.
Glass has set the new standard for what AI-native actually means, and you’d better believe that the best companies have taken notice and will be attempting to emulate, if they haven’t already (dev speed is insane now, what used to take 3 years now takes 3 months).
Where SLC Chat Stands This Week (And One Result That Surprised Me)
I’ve been heads down on SLC Chat (more context here), stress-tested against my voice, our data, and the thousands of deals we’ve reviewed as funders.
I pulled a deliberately tricky test question last week to gauge accuracy: “Walk me through what downside protection actually looks like on a specific deal. Subject is 8 acres in a rural Southeast county, asking $48K. Two recent sold comps at $7,200 per acre and $6,800 per acre, both similar access, flatter topo. One older comp at $5,400 per acre. Show me the math.”
Honestly, this is a poorly worded question on purpose (no DOM, no definition of ‘recent’ or ‘older’, no comp acreages, no detailed characteristics breakdown for the comps, especially the ‘older’ one). I wanted to see how the bot handled missing context, and what additional info it would request from the user.
On Claude’s mid-tier Sonnet model, SLC Chat got it roughly 95% there (correct math, quality caveats, mostly matched my voice).
On Opus 4.7 (Anthropic’s top-tier model), the response was…effectively perfect. Voice match and nuanced breakdown just about indistinguishable from how I’d answer it myself.
To be clear, this result does NOT mean SLC Chat is “100% accurate across all questions” (an impossible bar with the way LLM’s are currently constructed). What it IS, is “for certain detailed comping and downside-protection questions where my brain has logged thousands of reps, the bot is routinely answering at a level that requires NO edits from me.”
(Obvious for anyone who reads this newsletter, I am a STICKLER for accuracy. Our reputation is built on nuance and getting comping right, across a comprehensive set of land scenarios. If SLC Chat couldn’t protect that, we wouldn’t ship it.)
Check out the brief sneak peak in the P.S. below.
=====
The stock market cycle is compressing on a near-weekly basis…AI is compressing even faster…real estate is going to take its time regardless. We’re playing all three timelines at once, and that’s the actual opportunity (the operators holding the long-term frame while moving at breakneck pace on the AI side are going to look around in a few years and realize how far ahead they got).
If you’re an experienced operator with routine deal flow looking for a capital partner whose underwriting edge keeps sharpening (now amplified by world-class AI tools and workflows backed by proprietary formulas), please reach out. We write checks from $50K+, close 100% of deals we commit to, and bring national underwriting experience built across thousands of deals.
P.S. See the 3 min sneak peak of SLC Chat that legit blew away my expectations. Voice calibration and content QA are ~90% done. Setting up the delivery mechanism to publicly release shortly. Hit reply if you want early access.
P.P.S. Callan and the Uncommon Business team are running a series of free masterclasses ahead of their Accelerate to Automate launch (you’ll hear about this next week). I’ll be popping on the May 5 session to share specific Cowork build-outs aimed at land investors (and any operators who want to see what AI looks like applied tightly to a niche like ours). If you’ve been on the fence about diving in, this is a no-cost way to see what’s actually possible right now.
Use my link to sign up. FYI Callan and her team had so many thousands of people show up to the first Bootcamp session last week that the attendance broke Zoom. AI implementation is at an inflection point…interest has never been higher because the tools are have become the single biggest leverage point a business owner can utilize right now.
