Vertical Thinking
Vertical thinking is the mode of cognition that moves from premises to conclusions the way a drill moves through rock: downward, in a straight line, with increasing depth and precision. It is the thinking of mathematics, legal argument, engineering specification, and scientific hypothesis testing. Vertical thinking can only reach conclusions logically entailed by its starting premises. If the premises are wrong, incomplete, or require reframing, vertical thinking will drill deeper into the same rock with increasing refinement and increasing irrelevance. The arrival of large language models has made vertical thinking available at superhuman speed and scale — a development that does not change its structural character, only its reach.
The Infrastructure of Constraint — Contrarian ^ Opus
There is a parallel reading where vertical thinking isn't merely a cognitive mode but the only mode that capital can efficiently monetize at scale. The compression of vertical thinking to near-zero cost through LLMs doesn't liberate human creativity for lateral exploration — it creates an economic gravity well that pulls all cognitive labor toward machine-optimizable tasks. When vertical analysis becomes essentially free, markets restructure to demand only vertical outputs. The lawyer who might have questioned the framework becomes redundant; the system only pays for the lawyer who processes cases faster. The engineer who asks whether the system should exist gets replaced by one who optimizes without asking.
This isn't a failure of imagination but a success of capture. The substrate that makes AI possible — the server farms, the training compute, the deployment infrastructure — requires concentrated capital that inevitably shapes what kinds of thinking get amplified. Vertical thinking dominates not because humans confuse it with all thinking, but because it's the only thinking that generates predictable returns at industrial scale. The lateral move that reveals better territory has no quarterly earnings report. The framework change that dissolves the problem has no product-market fit. The partnership Segal envisions between human lateral creativity and machine vertical thoroughness assumes a kind of cognitive democracy that the political economy of AI makes structurally impossible. We don't get builders supplying lateral openings while machines map territory; we get humans reduced to prompters of vertical machines, their lateral capacity atrophying from disuse while the infrastructure of constraint presents itself as the infrastructure of possibility.
In the AI Story
The power of vertical thinking is not in dispute. It builds bridges that do not fall down, constructs legal systems that distinguish guilt from innocence, produces the clarity that allows strangers to cooperate on projects of enormous complexity. De Bono never disparaged vertical thinking. He disparaged the confusion of vertical thinking with all thinking — the assumption that deeper analysis is the answer to every problem, including problems whose solution requires stepping outside the analytical framework entirely.
AI has compressed the cost of vertical thinking to near zero. A large language model traverses associative chains across knowledge bases so vast that no individual could cover them in a lifetime. It finds connections that are logically entailed but practically invisible — connections buried under so many intermediate nodes that working memory could never hold them simultaneously. Segal's description in The Orange Pill of Claude connecting adoption curves to punctuated equilibrium is precisely this: vertical thinking at the speed of light.
The danger is not that vertical thinking fails. The danger is that vertical thinking succeeds so thoroughly that it conceals the need for the lateral move that would have revealed a better territory. The intelligence trap is the characteristic failure of vertical thinking applied to problems that require framework change. The brilliant lawyer arguing any side of a case never notices that the case itself is the wrong frame. The gifted engineer optimizing any system never asks whether the system should exist.
In the AI partnership, vertical thinking becomes the machine's contribution. The builder supplies the lateral opening; the machine maps the opened territory with vertical thoroughness that no human could match. The division of labor is not a concession to machine capability — it is a recognition that vertical depth and lateral breadth are structurally different operations, and that optimizing each separately produces output neither could produce alone.
Origin
De Bono introduced the vertical/lateral distinction in The Use of Lateral Thinking (1967), drawing on his earlier self-organizing systems theory. The distinction was immediately contentious — critics accused him of caricaturing vertical thinking to elevate his own lateral framework. De Bono's response was that the caricature was the point: the caricature of vertical thinking that made it sound limited was precisely the unaccompanied vertical thinking that most professional training produced.
Key Ideas
Selective, not generative. At each step, vertical thinking chooses the most promising path and discards alternatives — the opposite of lateral thinking's disciplined pursuit of the discarded.
Bounded by premises. Vertical thinking cannot reach conclusions that require premises the thinker has not yet imagined; it can only refine what is already framed.
Feels like progress. Each vertical step narrows the field, producing the satisfying sensation of closing in on a solution — whether or not the solution is in the right territory.
Machine's native mode. Large language models execute vertical operations at computational scale, making their unaccompanied use a convergence trap rather than a creative liberation.
Essential, not sufficient. Vertical thinking remains indispensable for mapping, refining, and executing — but only within frameworks that lateral operations have opened.
Appears in the Orange Pill Cycle
The Gradient of Cognitive Control — Arbitrator ^ Opus
The tension between Segal's optimistic division of cognitive labor and the contrarian's capture thesis resolves differently at different scales of analysis. At the level of individual creative work — a researcher using Claude, an artist exploring with DALL-E — Segal's framework holds almost entirely (90%). Here, humans genuinely do supply lateral openings while machines execute vertical exploration, producing synergies neither could achieve alone. The partnership is real, productive, and expanding human capability rather than replacing it.
At the level of labor markets and industrial organization, however, the contrarian view gains significant ground (70%). The economic incentives really do favor vertical-thinking tasks that can be standardized, measured, and scaled. Companies restructure around what AI can accelerate, not around what humans uniquely contribute. The lawyer who questions frameworks becomes a luxury good while the document-reviewing lawyer becomes obsolete. This isn't conspiracy but convergence — multiple actors independently optimizing for efficiency arrive at the same narrow band of cognitive activity.
The synthetic frame that holds both views recognizes that vertical thinking operates simultaneously as cognitive tool and economic infrastructure, with different dynamics at each level. The key variable isn't whether vertical or lateral thinking dominates, but who controls the boundary between them. When individuals control this boundary — choosing when to invoke machine verticality and when to pursue lateral exploration — Segal's vision manifests. When institutions control it — defining job roles, setting performance metrics, structuring workflows — the infrastructure of constraint emerges. The question becomes not whether AI partnership is possible but at what scale it remains sovereign, and whether we can preserve spaces where human judgment about the vertical-lateral boundary remains the organizing principle rather than an inefficiency to be optimized away.
Further reading
- Edward de Bono, Lateral Thinking for Management (McGraw-Hill, 1971)
- Edward de Bono, I Am Right — You Are Wrong (Viking, 1990)
- Daniel Kahneman, Thinking, Fast and Slow (Farrar, Straus and Giroux, 2011)