Edition 1 · Updated September 2026
The Kachyng Papers
A plain-English volume on agentic commerce: who an agent is, whether it does the same thing twice, and what the protocol stack still leaves unresolved. Free to read, no form. We add a chapter as each paper lands.
Contents
AI Memory Is a Lie
You are doing the work of believing it. This paper is about how that happened, who profited, and what real memory infrastructure would have to look like.
People are assigning a human quality to a marketing term. That is what is happening, every day, in millions of conversations between humans and the AI tools they use, in offices and homes and coffee shops, on phones and laptops and headsets, in every industry and in every language. Someone tells someone else that AI has memory. The someone else nods. They believe it. They have always known what memory is — they have been doing memory their entire life. They assume the AI is doing something like what they have been doing.
Read the chapter →
The Identity Layer for Agentic Commerce
Identity and delegation infrastructure for autonomous AI agents—IDX and KYA, the identity layer of Kachyng's Agentic Commerce Infrastructure
AI agents are about to transact and act on behalf of humans and enterprises at a scale the current identity stack was not built for. McKinsey projects $3–5 trillion in autonomous commerce globally by 2030 and calls the missing authorization layer "a critical authorization failure below the surface of agent adoption." Today's identity infrastructure — OAuth, SAML, Active Directory, service accounts — was designed for a human logging into a service, not for autonomous software reasoning, delegating, and transacting across systems. This paper covers the identity layer of Kachyng's Agentic Commerce Infrastructure: IDX (Identity Exchange) and KYA (Know Your Agent). Orchestration (AGX) and settlement (PRX) are addressed in companion papers.
Read the chapter →
Can You Trust an AI to Do the Same Thing Twice?
A Plain-English Guide to How AI Computes — and Why Its Answers Can Vary
Modern artificial intelligence depends on an enormous number of numerical calculations. These calculations are not generally performed with exact whole numbers. They use positive and negative fractional values that represent weights, activations, probabilities, similarities, and other learned relationships inside a neural network. A computer cannot store every such value with unlimited accuracy. It has a fixed number of binary storage positions, called bits, available to describe each number. Those bits must be divided among several jobs:
Read the chapter →
When AI Agents Go Shopping
A plain-English map of agent commerce, and the operating layer the protocol stack leaves unresolved
AI agents are beginning to transact: searching, carting, checking out, and paying on behalf of people and businesses. In under two years, more than a dozen protocols have appeared to standardize this, from MCP and A2A to UCP, ACP, AP2, TAP, Agent Pay, and x402. This paper maps all of them in plain English, in the order you would build them, through one running story: a customer, a sock shop, and its wool supplier. Three findings emerge from the map. First, retail agent commerce is being standardized quickly. Discovery, checkout, customer approval, and payment verification each now have credible, well- backed protocols, although rivals overlap and several have already changed shape after first contact with the market. Second, every protocol on the map assumes the business's internal logic is already agent-ready; none of them defines it. Third, enterprise commerce (contracts, purchase orders, approvals, invoices, payment terms) remains almost entirely unaddressed, and that is where the money is: global B2B payment volume runs at roughly $89 trillion a year against about $6 trillion for retail e-commerce. What is missing is an operating layer: the machinery that turns software built for humans into governed surfaces agents can safely operate, with identity, policy, approvals, limits, and evidence built in. The closing addendum describes how Kachyng's AGX and IDX address exactly that layer. The rest of the paper is the map that shows why it is needed.
Read the chapter →
We are writing 4 more. This page is the whole volume — keep the same link, and new chapters appear here as we publish them.
Take the volume with you
Every chapter above reads free on this page. If you want the printable PDF edition — all 4 papers in one file, formatted for the plane or the boardroom — tell us where to find you.