This means that AI enabled knowledge harvesting is immediately ready for deployment in your organization. The product associated with this deployment is a Claude Project. Included in the team oriented project licensed fee is six hours to modify the project system instructions to represent your critical knowledge area, and check-in midstream, when you need it, and to help you conduct an after action review to close out the project.
Before LT created 8 and 96, knowledge harvesting was not a scalable capability. Today it is. The world should know.
Entire energy sectors have walked away from incalculable amounts of value because they did not ask their experts to engage in knowledge harvesting. Before today, it was expensive, difficult to accomplish, and wildly mysterious. Today that’s no more.
When the subject matter expert is ready, the project site can be established in less than one hour. The experts, within just a half-hour, should feel at home and reveal new insights: knowledge they’ve never spoken or written about before.
Not because they're holding back, but because expertise becomes automatic and stops being sayable. 96 is the instrument that asks the questions that get it out.
The expert who shuts a line down on a feeling. The manager who knows who to actually call. Their most valuable knowledge is the most automatic, and automatic knowledge resists interviews, documentation, and exit checklists. It walks out the door at retirement, and the five-year onboarding clock starts over.
If the word knowledge makes you flinch, use a plainer one. What an expert knows is encoded in language. Structured questions decompose that language into structured information, inspectable, taggable, transmissible. You don't have to believe in tacit knowledge as a mystical category. You have to believe the right question surfaces what a flat question can't. That is testable.
Three decades of doing this by hand, one expert at a time. A method with a paper trail older than most of the field.
Building the AI version in public, dated, in the open. The work left footprints while it walked.
96: the same engine, running without the practitioner in the room. A competitor can copy a prompt. They cannot copy the arc that produced it.
A master's thesis at Auburn, EOR: Establish, Organize, Represent, directed by Charles A. Snyder, written under the influence of Doug Engelbart's unfinished work on augmenting human intellect. Its dialogue subsystem, Tutor, was a structured-questioning engine: the direct ancestor of the 96 questions. LearnerFirst (1992) turned it into shipping software. Thirty years of engagements followed.
When frontier models arrived, the engine already fit, because both a human expert and a model encode what they know as language, and structured questions are how you surface either one's implicit content. The trail:
A structured-questioning mechanism that has run for 37 years against the only mind that existed, the human one, and now runs against a second kind of mind as well.
It is being finished in public, and parts of it already ship. 8, the shipped slice, eight ways of knowing on a working grid, is live and priced today. 96 itself is in active build. We tell you exactly where it stands, because the honesty is the product.
A system prompt. You experience the simplicity.
Eight neurologically-distinct agents, an orchestrator that detects when to hand off, a decomposition pipeline (CATS + the 8×12 grid), and a question mapped to the neural system that stores the knowledge it reaches for.
You are sold the surface. The architecture is the provenance that makes the surface worth trusting. The door is simple; the house behind it is thirty-seven years deep.
We tested it. A competing frontier model was handed a generic description of 96 and asked to rebuild it. It derived the architecture cleanly, intake, parse, classify, matrix. Then it was asked for the part that makes the architecture work: the empirical calibration, which ninety-six questions, and why, and it could not produce it. Under operational pressure, the rebuild broke, every time.
The architecture is derivable. The calibration is not.
The text of a prompt copies in a keystroke. The thirty-seven years of knowing which question to ask, in which order, for which kind of knowledge, that is what you're buying, and it is what cannot be lifted.
A master practitioner's elicitation used to require the master in the room. The engagement ended when they left, and the cost never came down. 96 drops the marginal cost of that elicitation toward zero.
The claim is not "we scale", everything scales. The claim is that we scaled the one thing that used to live in a single practitioner's head and died with the engagement, and kept it expert-grade. The ninety-six neural-matched questions are the fidelity: drop them and you have another wrapper; keep them and the scaled output still carries a master's judgment.
Behind 96 sits the thing 96 is an instance of: one engine, expressed four times, Tutor (1989) → Builder-Player (the patent) → the Knowledge Harvesting steps → the 96 cells. A structured-questioning mechanism that makes implicit knowledge explicit, and runs against any knowledge-bearing system.
The honest limit, stated plainly: pointed at a human, the engine surfaces knowledge that existed and was hidden. Pointed at a model, it surfaces patterns, the residue of human cognition it was trained on. Humans hold knowledge; AI holds patterns; information is what moves between them. The mechanism is universal. What it returns is not the same in both cases, and we don't pretend otherwise.
Engelbart's vision was never the mouse or the window, those were the artifacts that got commercialized. The methodology and training meant to co-evolve with them never did. He called the goal collective IQ: raising how well people, together, solve harder problems faster. Knowledge Harvesting is a thirty-year contribution to that unfinished agenda. The 1989 thesis was written under his influence. The lineage is load-bearing, not decorative.