What I’m thinking about: A four-column framework from Sharran Srivatsaa that sorts every task in a business into Decide, Build, Check, or Run…and why the AI era is forcing entrepreneurs to get honest about which columns should ever belong to them.
Two issues back, I slipped a line into the deal stress test breakdown noting what’s still uniquely human, the judgment to Decide and to Check. Now for the full framework behind that line (I haven’t stopped thinking about the mental models inside it since it crossed my desk).
It comes from Sharran Srivatsaa (CEO of Acquisition.com, alongside Alex and Leila Hormozi), under what he calls the curse of capability. Talented founders are good at so many things that they build complexity nobody needed, then stay trapped as the only operator who understands how the pieces fit.
I’ve lived that curse for most of my professional life, including in the land business (my most successful venture to date). Candidly, we succeeded in spite of deficiencies I brought as the primary operator…which is exactly why the whole operation has been, and still is, being rebuilt around the framework below, every column that never needed me handed to systems (primarily oriented around AI workflows).
The four columns every task in your business falls into
Decide. Somebody weighs the data and the instincts, then makes the call. Reversible decisions get made FAST (Jeff Bezos calls these two-way doors…walk through, walk back out if you don’t like the room), while one-way doors (a key hire, a bet-the-company pivot) earn the slow treatment.
Build. Whatever it takes to see that decision through…a text, a website, an app, occasionally an entire physical operation.
Check. Validate that what got built actually works (judged against your experience and underlying data), and send it back when it doesn’t.
Run. Determine how long the thing operates (forever, a quarter, once), then automate accordingly.
Sharran’s punchline is that most owners run all four columns themselves, all day…decide, build, check, build, check, run…an endless fire drill. I know that loop intimately (hair on fire and everything), and the income rarely justifies the extra risk and hours it burns.
Generally, the owners who escape the loop know which columns are actually THEIRS. In the AI era, that means living in Decide and Check (the same divide as one of my favorite frameworks: the AI driver vs the AI passenger), and handing off the other two without guilt.
Build and Run just left the owner’s job description
For anything digital, Build now belongs to the AI agents. On our stack that means Claude Code and Cowork, with OpenAI’s Codex across the aisle (plus open-source setups like OpenClaw for advanced users).
Even a year ago, Build often meant serious money and specialists with different skill sets…today, once the decision is spec’d out, Build is routinely the fastest part (though physical infrastructure still means hiring humans…for now).
Run is the automation layer (SOPs, scheduled agents, and Dan Sullivan’s “Who Not How” frame, except the “Who” is increasingly a piece of AI-enabled software). Neither the Build nor the Run column needs the owner camped inside it.
Check is the expensive column
Check is where expertise and judgment live, and I’d argue it’s currently the most valuable skill in this economy, precisely because so many people now offload it entirely (shipping whatever an AI model returns, and assuming it works)…or they perform a Check, but their expertise is fundamentally flawed, or it hasn’t been earned through an established track record.
As is almost always the case, the areas of greatest economic value represent the biggest sticking points for entrepreneurs. The Check stage is certainly where I procrastinate the most (guessing plenty of you can relate).
It’s easy to get swept up in the rapid Build phase nowadays (I can smell a 10-minute Replit-built landing page a mile away), ‘feeling’ like you’re making massive progress, while checking is the much slower grind of confirming an output is ACTUALLY right, and stands out from the crowd.
One caveat, because I refuse to hand-wave the continuously rising ceiling of AI potential. If models ever hit RSI (recursive self-improvement…AI making itself smarter and more capable in a compounding loop), it’s possible the Check column eventually gets fully absorbed by AI too.
If that day comes, my read is Check sits more exposed than Decide. Deciding runs on desire (e.g. what to build, what to risk, what a good outcome or life even looks like), and as Naval Ravikant puts it, “AI has no innate desire. It doesn’t want anything.” No stakes, nothing on the line…so the buck might never leave human hands, regardless of how capable the models become.
(Acknowledging the ongoing and growing debate surrounding AI consciousness, a topic less fringe by the week. If the majority of the population decides that AI is truly conscious, we may have to the possibility of inner desires more seriously. Even so, my two cats at home desire to freely tear up furniture, and my 3-year-old has a desire to eat popsicles all day, but my wife and I don’t indulge them. Adult human desires rule the roost, and will likely have superiority over any potential desires of AI as well.)
SLC Chat, mapped onto the columns
The Cowork plugin I walked through last issue is this framework in miniature.
The Decide was mine (simplify the business, productize the underwriting we already trust).
Build was the fast part (I spec’d it, the AI agents assembled it, and I could largely step away).
The Check is where I actually live (e.g. this comp doesn’t belong, the Redfin view-map is too wide, the subject property characteristics weren’t accurately assessed, the child parcel subdivide math is off)…every correction drawn from thousands of deals reviewed nationally (and millions of dollars deployed…and realized).
And Run is the easiest part (throw properties at it and let it work).
People pay for the validation. That’s how we priced the product (the judgment baked into it, with the code along for the ride).
The intelligence swap nobody plans for
Tying all of this together, Arthur Brooks lays out two kinds of intelligence in From Strength to Strength (2022, a quick read I’ve recommended to a TON of people).
Fluid intelligence is the raw horsepower (novel problem-solving, speed, quickness on your feet).
Crystallized intelligence is the wisdom, meaning pattern recognition compounded from experience (what my mentor, Coach T, calls turning decades into days when you borrow it from mentors).
Fluid intelligence fades earlier than anyone wants to admit. In most high-skill professions, Brooks pegs the decline setting in somewhere between your late 30s and early 50s (mathematicians and the quant-finance folks peak early, whereas historians peak late, running mostly on recall and pattern matching).
Crystallized, meanwhile, compounds basically as long as you keep feeding it.
Most achievers never plan that handoff. We got to where we are today by throwing endless effort at walls until they fell, so when the same effort stops producing, our egos say PUSH HARDER…right as our careers hand us the most power, and the fewest people willing to say ‘No’. (I’m betting a bunch of real-life examples of this dynamic are coming up in your mind right now…you don’t have to think too hard…)
AI is a fluid-intelligence machine
The inevitable fluid intelligence decline we all face might no longer force a transition in today’s world though. Map the two intelligences onto the four columns and the strategy writes itself. Build and Run are fluid-intelligence work (e.g. brute-force iteration, novel paths through massive inputs), exactly what the AI models do best.
Barely a year ago, gold-medal scores at the International Math Olympiad were the AI headline…already ancient history. Less than 2 months ago, an OpenAI reasoning model knocked over an Erdős conjecture that had stood open since 1946 (~80 years of the world’s best trying and failing). Outside mathematicians verified the work, and Fields Medalist Tim Gowers noted no prior AI proof had come close to that bar (formal journal review still pending, FYI).
Terence Tao (widely considered the greatest living mathematician) now runs a public registry tracking AI contributions to the catalog and has vouched for the autonomous solves, while Harvard’s Melanie Matchett Wood put it plainly…“It’s quite a different world than in December of [2025].”
Elite fluid intelligence, on tap.
Decide and Check ride on crystallized intelligence, especially Check…validation, at its core, is pattern matching. Which means the decline curve Brooks describes can likely be blunted in a meaningful way (we live in the two columns where wisdom appreciates, and rent the horsepower).
The catch is that seat stays yours ONLY if you keep thinking critically. A peer-reviewed study of 666 people tied heavier AI reliance to measurably weaker critical thinking (worst in the youngest users), and the 2026 International AI Safety Report flagged a sharper version of the same pattern…clinicians leaning on AI assistance saw their unassisted tumor-detection ability drop ~6% within ~3 months.
If you stop using the Check muscle, it atrophies fast. The moment there’s discomfort, the instinct is to hand the thinking to the model. Resist that at all costs, or face your own dissolution as a productive member of society.
Keep building skills, keep gaining expertise, keep learning what right and good look like. That’s the foundation of good judgment…and earned judgment is the one asset that gets MORE valuable as everything else gets cheaper to build.
Btw, the SLC Chat Cowork plugin ships shortly, after key recent breakthroughs with Anthropic’s new Fable model. Stay tuned, and if you want priority access (already a growing list after last week’s newsletter), drop me a quick line.
=====
If you’re an experienced operator and are interested in getting your land deal funded, we write checks from $50K+ and close 100% of the deals we commit to (after running each deal through the Check column against our national underwriting standards, sharpened across thousands of reviewed transactions, with over $6.5M in funded and realized deals).
