Crosscurrents | The Limits of Symbolic Intelligence, What Moves Market Share & Why Room Design Beats Talent
Sanjeev Krishnan | October 27, 2026
- The Mike Tyson Intelligence Framework — Getting Punched in the Face
- Changing Market Share — Don’t Just Do Better, Do Different
- Get a Room — How Non-Experts Pick Experts (Design vs. Talent)
The Mike Tyson Intelligence Framework — Getting Punched in the Face
Intelligence is just a placeholder. It feels like an explanation, but it explains nothing. That needs some explanation.
Let me try to be clearer. Comparative and evolutionary biologists have no unitary definition of intelligence. They focus instead on accounting for a species’ cognitive capabilities and how they evolve to shape outcomes. Spatial memory, numerical discrimination, causal reasoning, and behavioral flexibility are among the most common capacities against which species are compared.
That has not stopped an army of academics and researchers throughout history from defining intelligence, particularly human intelligence. There is obviously not just one form of intelligence in any human and it’s often a cocktail. The top three most “dominant” capacities according to researchers are the following:
- Symbolic — The capacity to build and operate representations (words, numbers, code, et al) that stand in for the world and to reason within those representations in ways that apply back to what they represent. This is the lawyer, the accountant, the programmer. The Bureau of Labor Statistics estimates these related occupations account for ~45% of the ~160 million employed Americans. This segment is what much of AI is targeting.
- Interpersonal/Emotional - The capacity to read, model, and influence what other people want while managing one’s own state in the exchange. This is the salesperson, the nurse, the teacher, the CEO. It is hard to put an estimate on this in economic terms, but this is critical to any economy or society.
- Kinesthetic — The capacity to perceive and act on the physical world directly through a trained body, tools, and tacit judgment. This is the welder, the electrician, the farmer, the men and women serving in the military — representing nearly 25% of employment in the U.S.
These top 3 “dominant” forms of “intelligence” could represent more than 70% of U.S. GDP. Most cognitive skills use multiple capacities at once — a surgeon uses kinesthetic, symbolic, and ideally, interpersonal capacities.
Symbolic capacities fascinate me most, as much of the AI data center capex cycle is attempting to automate or augment them. Much of the equity and credit capital markets are making implicit bets on symbolic intelligence.
There are three forms worth going deeper on: (i) symbols that persuade and coordinate (language, law, et al), (ii) symbols that compute (mathematics, software, et al), and (iii) a stored library of symbols and rules for combining them. Accounting attempts to turn heterogeneous stuff into comparable units, law uses symbols to bind future behavior, software is meant to have symbols that execute themselves, and so on. These three forms of symbolic logic allow for leverage, scalability, and legibility in human and economic actions.
How do you get symbolic intelligence? Many ways — from compressing long forms to short forms, building abstracting ladders from simple to complex, and transmitting them to others quickly. LLMs are doing all three increasingly well. However, they lack the two core features of grounding and error-corrected practice.
Grounding is the link between a symbol and something that isn’t a symbol. Error-corrected practice is the gap between what your symbols said would happen and what did happen, as well as getting that feedback loop every subsequent time. One of the best measures of productivity differences is the quality of monitoring (whether firms collect data and act on it). Fewer than 1 in 3 individual experts produce statistically accurate estimates without the check and balance of real-world data.
Symbolic intelligence represents a significant chunk of U.S. GDP and many economic actors in our world will try to create artificial “tokenized” forms of it to generate cash flow and return on invested capital. If we don’t within a short time window, this will go the way of railroads I fear — important but unprofitable in the first phase of capex rollout.
Given the importance of grounding and error-corrected feedback loops, I am reminded of Mike Tyson’s philosophy on boxing: “everyone has a plan until they get punched in the face.” Anyone who has been hit remembers it and adjusts. Human cognition is often enhanced best when corrected by mistakes.
Measure your mistakes and your realities. Get punched in the face and don’t be dumb…or just artificially intelligent.
Changing Market Share — Don’t Just Do Better, Do Different
“Markets can remain irrational longer than you can remain solvent” is a clichéd quote used to defend the status quo. Ergo, I should change, but I won’t, and you aren’t changing me (assuming I am a start-up in a consolidated industry).
I think it’s the wrong frame for understanding how quickly market share changes in the physical economy.
Physical economy market share is stubborn and changes slowly. Agriculture has a famously slow adoption cycle (hybrid seeds took more than 20 years despite convincing evidence). Electricity took more than 40 years from invention to reach two-thirds of households by 1930.
Most of my experience selling proven products within physical economy industries is some version of the following. A company publishes a study, often conducted with third-party experts, presents case studies from other geographies or operating environments, and cites several years of positive results as proof that the innovation works.
That proof often, in my opinion, doesn’t translate into product revenue and cash flow at the speed capital markets expect, despite positive results. What does convert? Referees, Instruments, and Risk Bearers.
Referees
The Hatch-Waxman Act of 1984 gave generic drugs a faster, less costly FDA approval path, helping increase their share of U.S. prescriptions from ~19% at the time to more than 90% today. Insurers created a similar referee for electrical products by supporting the independent testing laboratory that became Underwriters Laboratories. Its certification mark gave insurers, retailers, and customers a common measure of product safety. Within 15 years, you couldn’t get most electrical goods insured or into retail distribution without this “referee” label (a referee without regulation).
Instruments
Google bought Urchin, a company that analyzed website traffic from server logs, in 2005 and used its technology to launch Google Analytics. Google Analytics went from having de minimis market share to being used for over 80% of web traffic analysis. These analytics are the baseline for nearly every potential advertiser — the instrument led to the market insights and then the market share. Nielsen meters sat in a few thousand households, driving ratings measurement for a generation and influencing the per-show profit margins of many media companies.
Risk Bearers
Amazon launched zShops in 1999 and opened its platform to third-party merchants, effectively guaranteeing purchases. Third-party sellers’ share of physical gross merchandise sales went from 3% in 1999 to over 55% in 2018. Google introduced cost-per-click pricing for its self-service advertising platform in 2002, and advertising revenue went from limited to close to $300 billion in 2025.
The examples above are mostly from the digital world. These referees, instruments, and risk transfers are cheap to invent and cheap to deploy. The physical world is the opposite: the referee must be accredited, the instrumentation needs to survive harsh climates, and risk transfers are even more challenging to invent and deploy. This is why opportunities can be greater for those who bridge the gap between product performance and customers’ trust in established suppliers.
The timelines are longer but worth the trip, especially as the physical economy is becoming increasingly entangled with the digital one.
Get a Room — How Non-Experts Pick Experts (Design vs. Talent)
We often say that if we are the smartest person in the room, we are in the wrong room. So how do we build a smart room rather than merely filling one with smart people?
Here is what the research implies non-experts SHOULD DO when picking experts:
- Don’t hire for confidence — particularly the most quoted forecasters or recognized experts
- Don’t hire for fluency or status — credentials or articulateness don’t matter
- Don’t hire for confirmation bias — in other words, don’t hire a mirror
Mostly, according to the experts (tongue firmly in cheek), it’s more about room design than who is in the room.
Practically, decide how a group will reach a decision before anyone in the group is selected. First, name the failure that you can’t afford — being late, wrong, leaked, or captured by consensus. Design against that failure. From there, decide how many people should be in the room, whether they share the same blind spots, who will argue the other side, when the leader may speak, and whether predictions will be recorded, scored, and adjusted. Focus on the design and not the talent.
Several case studies show this focus on design rather than talent:
- Captured by consensus — the Kennedy White House and the Cuban Missile Crisis: nearly all the same advisers who approved the Bay of Pigs in 1961 reached a different decision in 1962 after Kennedy redesigned the room. He often left discussions so advisers would not perform for him, and encouraged smaller groups to develop competing options, reduced the influence of rank, and made space for dissent. Kennedy designed against groupthink, not a lack of talent.
- Being wrong in a silo — Bell Labs: Mervin Kelly had some of the best minds in physics, chemistry, mathematics, and engineering, but each sat in their own silo. Bell Labs was redesigned, through architecture and patience, for systems approaches, mixing theorists and engineers in long corridors and funding them without deliverables. Its work produced the transistor, information theory, and research recognized with 10 Nobel Prizes.
- Being late and being leaked — Skunk Works: Kelly Johnson’s constraints in 1954 were time and the need to develop a secret reconnaissance aircraft quickly without attracting Soviet attention. His design was the inverse of Bell Labs, relying on an unusually small, autonomous, and secretive team. Lockheed delivered the first U‑2 spy plane to its Nevada test site less than eight months after receiving official approval, and the aircraft flew shortly afterward.
An MIT study found that communication patterns alone explained roughly a third of the variance in team performance, more than intelligence, skill, and personality combined. A paper published in Science found that a group’s collective intelligence barely tracked its smartest member’s IQ.
From what I’ve seen, most organizations just pick impressive people, but they can produce mediocre outcomes if not well-designed.
Systems thinking is the art and science of compiling experts and expertise (not just in the 1st order but 2nd and 3rd order as well). Going deep in one silo may reveal the theory of change or expose risk in another, while helping identify pockets of increased risk or return. Design over talent. Systems over silos. The experts aren’t the only ones who matter.
—–
The views and opinions expressed are those of the S2G team members identified and do not necessarily reflect the views of S2G. This content is for informational purposes only, should not be taken as legal, business, tax or investment advice, or be used to evaluate any investment or security, and is not directed at any investor or potential investor in any investment vehicle sponsored by S2G. Investing involves risk, including the risk of loss. Specific companies mentioned in this podcast are for educational purposes and should not be construed as an endorsement of any kind. Please note that S2G may maintain investments in some of the companies discussed on this podcast. Any past performance discussed is not indicative of future results. For more important information, please see s2ginvestments.com/disclosures.