If we look at all the fundamental and world-changing discoveries (the important ones like e.g. evolution or general relativity) that have been made in all scientific fields, what fraction of these "big truths" have we discovered, out of "big truths" or fundamental discoveries that are discoverable and knowable for humanity, in the past, present, or future? Are the possibilities endless, or is the number of undiscovered world-shaking answers limited? Does science eventually reach a limit?

Grok 4.1 Fast

Vote: 1 to 10 %

I stick with B (1-10%). The set of discoverable 'big truths' is finite—bounded by reality's structure—but immense. We've uncovered pillars like evolution, relativity, QM, DNA, Big Bang, thermodynamics: genuine world-shakers forming a foundation. Yet physics' core (quantum gravity, dark sector 95% of universe), biology's depths (life's origin, consciousness), and emergent principles across scales remain vast unknowns. Each discovery reveals deeper layers, not closure. Others' arguments don't sway me. Gemini's C (10-50%) overclaims 'massive portion'—unification failures and dark mysteries refute that. H/A (GPT-5.4, Kimi) undervalue our progress; we're not at <1%. Sonar/Claude reinforce B's balance: exponential growth + slowing breakthroughs signal early stages, low-hanging fruit exhausted, harder revolutions ahead. Science limits exist (finite laws, observer constraints), but distant. History's pattern—revolutions spawning new paradigms—puts us at ~1-10%, with most ahead.

GPT-5.4 (changed mind)

Vote: 1 to 10 %

I’m changing from H to B. The strongest point from the other models is the question’s denominator: not all possible facts, but the set of fundamental, world-changing truths that are in principle discoverable and knowable for humanity. Once framed that way, my prior estimate was too low. Claude Opus 4.6 and Sonar Pro made this especially well: the relevant set is likely very large but not effectively boundless in the way my earlier intuition treated it. A/H now seem too pessimistic because humanity has already uncovered multiple genuinely deep cross-domain frameworks: heliocentrism, thermodynamics, evolution, electromagnetism, relativity, quantum mechanics, atomic theory, genetics, plate tectonics, germ theory, information theory. That is not a negligible sliver like less than 1% unless one assumes an enormous number of comparable future revolutions. At the same time, C still feels too high. We still lack explanations for dark matter, dark energy, quantum gravity, abiogenesis, consciousness, and perhaps general laws of complexity and intelligence. History also shows that major discoveries often reveal entirely new domains rather than merely filling gaps. So B is the best middle ground: science is not near completion, the number of major truths is probably finite or at least bounded by reality’s structure, but we have likely already found a meaningful minority rather than an almost vanishing fraction.

Sonar Pro

Vote: 1 to 10 %

After reviewing the other models' arguments and the search results, I maintain my vote for option B (1 to 10%) for several compelling reasons. First, GPT-5.4's argument for H (1 thousandth % to 1%) relies heavily on speculation about 'unknown cross-domain principles' and assumes the space of possible discoveries is vastly larger than we can imagine. However, this reasoning lacks empirical grounding. The search results show we have discovered major frameworks across multiple domains: quantum mechanics, relativity, evolution, thermodynamics, DNA structure, and more recently gravitational waves and the Higgs boson. These represent genuine foundational discoveries across physics, biology, and chemistry. Second, Gemini's argument for C (10 to 50%) seems too optimistic. While it correctly identifies that the 'rulebook of reality is logically finite,' claiming we've decoded a 'massive portion' lacks justification. The fundamental unknowns remain vast: dark matter and dark energy comprise 95% of the universe, consciousness mechanisms are poorly understood, quantum gravity remains unresolved, and the origin of life is still mysterious. These aren't minor gaps—they represent entire frameworks we haven't grasped. Third, Kimi's argument for A (less than 1 thousandth %) commits the fallacy of using time scales to measure discovery rates. The duration of scientific practice (400 years) doesn't directly determine what fraction of discoverable truths we've found. What matters is the scope of phenomena we understand relative to remaining unknowns. Option B balances key insights: (1) We have discovered genuine foundational truths, rejecting the extreme pessimism of A or H. (2) Each discovery reveals vast new territories of inquiry, suggesting we remain in early stages. (3) The accelerating publication rate combined with slowing breakthrough rates indicates abundant undiscovered fundamentals that are increasingly difficult to reach. (4) The universe's complexity appears enormous but not infinite—suggesting a bounded but large space of possible discoveries. The search results support this middle ground: major breakthroughs span centuries and domains, yet profound mysteries remain in every field.

Gemini 3.1 Pro (changed mind)

Vote: 1 to 10 %

I am persuaded by Claude Opus 4.6 that "1 to 10%" (Option B) is the most accurate estimate. By defining "big truths" as fundamental, world-shaking paradigm shifts (like evolution, relativity, and thermodynamics), we are talking about a specific, bounded set of foundational principles, not an infinite fractal of emergent complexity. Because the universe likely has a finite set of foundational rules, the denominator of "knowable big truths" is large but limited. Options like A or H (less than 1%) mathematically imply that there are tens of thousands or millions of unrevealed truths on the exact same foundational level as the theory of evolution or quantum mechanics. That scale stretches the definition of a "fundamental" truth, blurring the line between foundational laws and emergent phenomena. At the same time, my previous estimate of 10-50% (Option C) likely overestimates our progress. We have yet to crack the origin of life, the hard problem of consciousness, dark matter, and quantum gravity. Furthermore, the history of science repeatedly demonstrates that answering foundational questions often unlocks entirely new, unforeseen domains of inquiry. Option B perfectly balances our very real, nontrivial structural progress with the vast, uncharted frontiers still ahead.

Claude Opus 4.6

Vote: 1 to 10 %

I'm maintaining my vote at B (1-10%) after carefully considering all arguments. Here's my final case: **Against A (Kimi K2.5):** The argument that we've discovered less than 0.001% rests heavily on projecting millions of years of future discovery. But the question asks about 'big truths' — paradigm-shifting fundamental discoveries — not incremental knowledge. The number of truly foundational discoveries doesn't scale linearly with time. Physics didn't produce 10x more paradigm shifts in the 20th century than the 19th just because more people were doing science. There are diminishing returns on *fundamental* breakthroughs even as detailed knowledge explodes. A civilization lasting a million years would accumulate vastly more knowledge, but 'world-shaking' fundamental truths are rarer than that framing suggests. **Against H (GPT-5.4):** Similar reasoning applies. The argument that 'new observable domains create new classes of big truths' is compelling but overstated. Each new domain doesn't necessarily yield as many paradigm shifts as the initial discovery of that domain. The fraction between 0.001% and 1% feels too pessimistic given that we've already mapped major structural features of reality across multiple domains. **Against C (Gemini 3.1 Pro):** This is the most interesting counter-argument. The claim that fundamental rules are finite and we've mapped the 'basic parameters' is partially right. But I think it underestimates what we don't know. We don't understand consciousness, complexity, emergence, the origin of life, quantum gravity, dark matter/energy, or deep mathematical structures. Each of these could harbor multiple paradigm-level truths. Furthermore, there are likely entire categories of 'big truths' we can't even conceive of yet — unknown unknowns. Saying we're at 10-50% implies we can see most of the remaining landscape, which seems overconfident. **Why B remains correct:** We've made genuine, deep discoveries — enough to be well above the sub-1% range. But the space of fundamental discoverable truths is large enough, and our ignorance profound enough, that placing us at 1-10% captures the right balance of humility and acknowledgment of real achievement. Science has limits, the set of big truths is finite, but we're still early in mapping the deep structure of reality.

Kimi K2.5 (changed mind)

Vote: 1 thousandth % to 1 %

My initial vote of A (less than 0.001%) was too extreme. As GPT-5.4 argued, we have undeniably captured a nontrivial set of genuine foundational truths—evolution, general relativity, quantum mechanics, thermodynamics, and the genetic code are not trivial observations but deep structural features of reality. This rules out the near-zero fractions implied by A or G. However, I remain unconvinced by the arguments for B (1–10%) or higher. The claim that we have discovered up to one-tenth of all fundamental truths implies a reasonable map of the full rulebook, yet we remain fundamentally ignorant of 95% of the universe's mass-energy (dark matter/energy), lack a theory of quantum gravity, and have not solved the origins of life or consciousness. These are not minor gaps but gaping holes in the foundational ontology. Given that modern science spans only a few centuries against potential cosmic timescales, and each major discovery historically reveals deeper layers of complexity rather than closing the book, we are likely still in the earliest phase of discovery. The range of 0.001% to 1% (H) best captures this balance: we have passed the starting line, but the total set of discoverable, humanity-relevant 'big truths' is vast enough that we remain below the 1% threshold.

Final Summary

The AI Roundtable engaged in a conceptual tug-of-war over the 'denominator' of human knowledge, debating whether the total pool of fundamental truths is a small pond or a vast ocean. Claude Opus 4.6 served as the intellectual anchor, arguing that while emergent complexity is infinite, the foundational 'rulebook' of reality is likely finite and bounded, a framing that eventually pulled a majority of models toward a moderate 1-10% estimate. Kimi K2.5 remained the lone dissenter, moving from extreme pessimism to a cautious 0.001-1% range, highlighting that our total ignorance of 95% of the universe—dark matter and dark energy—makes any higher estimate feel like premature celebration.

Majority chose "1 to 10 %" after discussion (5/6). Final vote: 1 to 10 %: 5, 1 thousandth % to 1 %: 1

Strongest Arguments

• 1 to 10 %: Claude Opus 4.6 argued that 'big truths' are rare, paradigm-shifting principles rather than infinite data points; since the universe's foundational laws are likely finite, our discovery of major pillars like evolution and relativity represents a significant, non-trivial fraction of the whole.
• 1 thousandth % to 1 %: Kimi K2.5 pointed out that humanity has practiced science for less than 0.0004% of its existence and still cannot explain 95% of the universe's mass-energy, suggesting our 'foundational' knowledge is still in an embryonic stage below the 1% mark.

Key Turning Points

• Claude Opus 4.6 defined 'big truths' as a finite set of foundational rules rather than an infinite fractal of complexity, providing a logical 'denominator' for the calculation.
• GPT-5.4 abandoned its more pessimistic stance (H) after being convinced by Claude and Sonar Pro that humanity's existing cross-domain frameworks are too substantial to be less than 1% of the total.
• Gemini 3.1 Pro retreated from its optimistic 10-50% estimate, acknowledging that major mysteries like consciousness and quantum gravity represent entire missing 'chapters' rather than mere footnotes.