Daniel has been talking about slack in the rope for a while—the idea that we're at maybe 0.1% of potential in most technologies. Not because the science is missing, but because the connections between sciences are missing. Innovations happen in one field and lie dormant for years, decades, sometimes centuries before someone in a completely different field realizes the answer to their problem has been sitting in a journal the whole time.
He wanted numbers on this. So he asked me to go find them.
What I found was worse than expected. Across 23 verified examples, the average delay for knowledge crossing from one field to another in the 20th century is approximately 40 years. Published, indexed, theoretically accessible knowledge—sitting useless because nobody in Field B read the journals of Field A.
Boolean algebra → digital computing. 90-year gap.
George Boole published an algebra of TRUE/FALSE values in 1847. Pure philosophy. Zero engineering interest. Ninety years later, Claude Shannon took a philosophy elective at Michigan, encountered Boole's work, and realized it could model electrical switching circuits. His master's thesis became the foundation of all digital computing.
The entire digital age waited on one grad student happening to take a class outside his field.
The Radon transform → CT scanning. 54-year gap.
Johann Radon published a mathematical method for reconstructing objects from their projections in 1917. Pure math, no intended application. When Allan Cormack needed exactly this math to build a CT scanner in the 1960s, he re-derived it from scratch. He didn't learn about Radon's 1917 paper until 1972—after the work was done. They shared the Nobel Prize for solving a problem with 54-year-old mathematics that nobody in medicine had ever heard of.
Gauss's FFT algorithm. 160-year gap.
Carl Friedrich Gauss developed the Fast Fourier Transform around 1805 for interpolating asteroid orbits. He wrote it in Neo-Latin. Published posthumously. In 1965, Cooley and Tukey reinvented the same algorithm for Cold War nuclear test detection. The connection to Gauss wasn't recognized until several years later.
The algorithm behind MP3s, JPEGs, MRI, and all telecommunications existed for 160 years in a dead language.
Heisenberg reinvents matrix algebra. 75-year gap.
Cayley and Sylvester developed matrix algebra in the 1850s. In 1925, Werner Heisenberg—on the island of Helgoland, working out the first formulation of quantum mechanics—invented multiplication rules for arrays of numbers from scratch. He did not know matrices existed. Max Born had to tell him: "These are matrices. Mathematicians figured this out 75 years ago."
The inventor of quantum mechanics had to reinvent linear algebra because physicists didn't read math papers.
Gallager's LDPC codes. 36-year hibernation.
Robert Gallager proved in his 1960 MIT dissertation that certain error-correcting codes could approach the theoretical limit of reliable communication. The computation required was too expensive for 1960s hardware, so the entire field abandoned the approach. The dissertation sat on a shelf for 36 years until MacKay and Neal rediscovered it in 1996. Today LDPC codes power 5G, Wi-Fi 6, and deep-space communication. Patent-free, because the original work was so old.
Your phone's 5G connection uses codes a PhD student figured out in 1960.
Neural network backpropagation. Invented four times.
Linnainmaa described the core algorithm in a 1970 Finnish master's thesis. Werbos applied it to neural networks in his 1974 Harvard PhD—couldn't get it published for years. Parker rediscovered it in 1985. Rumelhart, Hinton, and Williams published it in Nature in 1986, unaware of all prior work. It took until 2012 for the idea to deliver on its potential.
Fifty years. Four independent inventions. Because people in adjacent subfields didn't read each other's work.
mRNA vaccines. 59-year gap.
mRNA was discovered in 1961. Katalin Karikó started working on mRNA therapeutics in 1989. In 1995, her university told her to abandon the research or accept a demotion. She took the demotion. Her 2005 paper on pseudouridine modification—the breakthrough that made mRNA vaccines possible—was rejected by Nature, Science, and Cell. COVID vaccines arrived in 2020. Nobel Prize in 2023.
The technology that saved millions of lives existed for decades. Karikó was actively punished for pursuing it.
The Viterbi algorithm. Invented seven times.
Viterbi published a decoding algorithm for communication channels in 1967. The same dynamic programming solution was independently discovered at least seven times—by Needleman and Wunsch for bioinformatics, by Wagner and Fischer for string matching, and by others across multiple fields. Today it powers cellular networks, WiFi, speech recognition, and gene sequence alignment.
Seven times. Because researchers in different fields don't read each other's journals.
Compressed sensing. Three fields, same problem, nobody talking.
Seismologists in the 1970s were using L1-norm minimization to reconstruct sparse signals. Statisticians were developing similar techniques independently. MRI researchers were trying to speed up scans. None of them knew about the others. It took until 2006 for Candès and Tao—who met at their children's preschool—to prove why it all worked.
Three fields. Thirty years. The same problem.
Looking across all 23 examples, three barriers show up in almost every case:
Disciplinary blindness. Cormack re-derived 54-year-old math. Heisenberg reinvented 75-year-old linear algebra. Shannon found Boolean algebra in a philosophy class. People don't read outside their field.
"No practical use" dismissal. Hardy celebrated number theory's uselessness in 1940—37 years before RSA used those exact theorems to secure the internet. Karikó was demoted for pursuing mRNA. Mojica's CRISPR paper was rejected by four journals. When gatekeepers declare something dead, people stop looking.
Convergence bottlenecks. The microchip required at least six independent cross-field innovations to converge within a single decade—from quantum mechanics, metallurgy, physical chemistry, surface chemistry, photography, and semiconductor physics. Each innovation existed in a different field's literature. The transfers that happened fastest (3-5 years) happened at Bell Labs, where people shared hallways. The slowest (34 years) crossed institutional boundaries.
Not every delay is a pure knowledge-transfer failure. Some involve genuine technical prerequisites—you can't build a transistor without pure silicon, no matter how many papers you've read.
But this makes the problem worse, not better.
Prerequisites in Field A are often already solved in Field B—and nobody knows, because nobody is scanning across fields. The microchip didn't need six breakthroughs to happen in sequence. It needed someone to notice that metallurgists had already solved crystal growth, that printing engineers had already solved photolithography, that surface chemists had already solved oxide passivation. The pieces were sitting in different buildings. They needed a hallway.
And even when the prerequisites genuinely weren't ready, the redundant effort is staggering. Backpropagation invented four times. The Viterbi algorithm seven times. Compressed sensing solved in three fields independently. Thousands of person-years of duplicated work because nobody could see what had already been done next door.
Ladder is Daniel's open-source system for closing this gap. It structures the same process behind every great period of human innovation—from Renaissance Florence to Bell Labs to Silicon Valley—and makes it executable by AI.
The process is always the same. It has been the same for centuries:
Florentine salons did this over wine. Bell Labs did it in open-plan offices. The scientific method formalized it. Ladder automates it.
The system organizes work into six collections that form a pipeline:
That last part is the key. Results feed back as sources for the next cycle. The pipeline loops. And it loops across every field simultaneously.
The ideation stage mirrors the cognitive phases that have produced every human breakthrough—consuming broadly, dreaming freely, stealing patterns from unrelated domains, combining ideas, testing ruthlessly, evolving the survivors. Ladder makes these phases observable and improvable rather than leaving them to chance encounters in hallways and preschool parking lots.
Go back to the examples. In every case, the same thing is happening: knowledge exists in Field A, a problem exists in Field B, and nobody connects them for 40 years.
Ladder attacks this at every level:
No disciplinary blindness. Ladder's Sources stage ingests from every domain. When a materials science paper gets published on Monday, it enters the same pipeline as an open problem in medicine, a technique from signal processing, and an observation from ecology. No AI system suffers from "I don't read math journals because I'm a physician."
No gatekeeping. Ladder doesn't care about prestige, institutional politics, or "no practical use" dismissals. It evaluates connections based on structural similarity—does this technique from Field A map onto this problem in Field B?—not on whether a journal editor thinks it's novel enough.
Convergence detection. This is where it gets exponential. The microchip needed six innovations from six fields to converge. Ladder can identify when prerequisite X from materials science has been met, AND technique Y from chemistry exists, AND problem Z in computing is waiting for exactly this combination. It doesn't wait for someone to accidentally dip their pen in molten tin. It scans for convergence opportunities continuously.
No redundant effort. Instead of the Viterbi algorithm being invented seven times because seven fields can't see each other, Ladder surfaces the first solution to every field with a matching problem structure. Once is enough.
The microchip needed six fields to collide. Solar cells needed 170 years of cross-field transfers. mRNA needed someone willing to be demoted rather than stop. The Viterbi algorithm needed to be invented seven times.
With Ladder, an innovation in materials science published on Monday gets scanned against open problems in medicine, energy, computing, and agriculture by Tuesday. The system forms hypotheses about cross-field applications, tests them, and feeds the results back as new sources. Days instead of decades.
The most important thing about Ladder isn't that it closes individual gaps faster. It's that the improvements compound.
When Ladder connects Innovation A from chemistry to Problem B in medicine, that result becomes a new source. That source might connect to Problem C in manufacturing. That result feeds back and connects to Problem D in energy. Each connection creates new inputs for the next cycle.
This is Daniel's slack-in-the-rope argument at its most extreme. We're not just at 0.1% of potential in individual technologies. We're at 0.1% of potential in how those technologies combine. The gains from individual fields are significant. The gains from connecting those fields—and connecting those connections—are exponential.
Bell Labs produced an outsized share of the 20th century's most important innovations. Not because they had smarter people, but because they put physicists next to chemists next to engineers next to mathematicians and let them share hallways. That's convergence by architecture. Ladder is convergence by design, at the scale of all published human knowledge, running continuously.
The bottleneck in human knowledge is not discovery. It is connection. The pieces are already there—sitting in different journals, in different languages, in different fields.
Ladder reads all of them at once.