<![CDATA[Agentic Economy]]>44732d09-334e-5ce0-982e-139986e3a5aaSat, 16 May 2026 13:03:21 +0000Captivate.fmRené Dechamps Otamendihttps://artwork.captivate.fm/028fffda-f262-41c5-a99c-e97245ada5bc/Agentic-Economy-Podcast.pngAgentic EconomyRené Dechamps OtamendiRené Dechamps OtamendiHere's a podcast description for Captivate: Short description (for directories like Apple/Spotify, ~150 chars): The engineering, economics, and math behind autonomous agent commerce. Research-driven. No hype. From AgenticEconomy.devhttps://agentic-economy.captivate.fmtrueepisodicnopodcast05 - The EU Agentic Commerce Window — what Brussels has, what's missing, and 2 August 202605 - The EU Agentic Commerce Window — what Brussels has, what's missing, and 2 August 2026Episode 5: "The EU Agentic Commerce Window — what Brussels has, what's missing, and 2 August 2026"

In seventy days, the EU AI Act enters its applicability phase. Article 50 transparency obligations, Annex III high-risk classifications, and the full Title III stack land on 2 August 2026. And most of the agent-to-agent commerce being built right now sits in a regulatory zone the AI Act doesn't actually cover end-to-end.

This is Field Report N°1 — a different series than the five research preprints. It's built for the policymakers, MEP staffers, in-house counsel, and serious builders who need to know, concretely, what Brussels has, what's missing, and where the consultation windows still are.

We walk through the six-step lifecycle of an EU agentic transaction: a principal establishes a mandate (GDPR Art. 22 territory), a software agent acts on it (AI Act risk tiering, Tier 2 for high-risk like mortgage decisions, Tier 3 for limited-risk like booking flights), it interacts with a venue (DSA, DMA), value moves over a payment rail (PSD3, SEPA Instant, MiCA, all requiring Strong Customer Authentication that was never designed for a software "customer"), finality lands (settlement neutrality), and a record persists (GDPR Art. 15 audit trail). At each step we name the dominant EU instrument and the gap underneath it.

Three gaps in particular get unpacked. The agent-to-agent disclosure vacuum — Article 50 only requires AI to disclose itself to humans, and the AI Office consultation on the implementing guidelines closes on 3 June 2026. SCA breakdown when the counterparty is software. Mandate granularity under Art. 22 with no implementation standard for agent commerce. Plus the transversals matrix — DSA, DMA, GDPR, PSD3, MiCA, eIDAS 2.0 each covering only a slice, with no single regulator owning the stack: DG CNECT, DG FISMA, EDPB, and the AI Office all hold pieces.

If you're shipping agent features into the European market, advising on EU digital policy, or trying to read the AI Act in the context of everything else around it, this is the operating manual.

]]>Episode 5: "The EU Agentic Commerce Window — what Brussels has, what's missing, and 2 August 2026"

In seventy days, the EU AI Act enters its applicability phase. Article 50 transparency obligations, Annex III high-risk classifications, and the full Title III stack land on 2 August 2026. And most of the agent-to-agent commerce being built right now sits in a regulatory zone the AI Act doesn't actually cover end-to-end.

This is Field Report N°1 — a different series than the five research preprints. It's built for the policymakers, MEP staffers, in-house counsel, and serious builders who need to know, concretely, what Brussels has, what's missing, and where the consultation windows still are.

We walk through the six-step lifecycle of an EU agentic transaction: a principal establishes a mandate (GDPR Art. 22 territory), a software agent acts on it (AI Act risk tiering, Tier 2 for high-risk like mortgage decisions, Tier 3 for limited-risk like booking flights), it interacts with a venue (DSA, DMA), value moves over a payment rail (PSD3, SEPA Instant, MiCA, all requiring Strong Customer Authentication that was never designed for a software "customer"), finality lands (settlement neutrality), and a record persists (GDPR Art. 15 audit trail). At each step we name the dominant EU instrument and the gap underneath it.

Three gaps in particular get unpacked. The agent-to-agent disclosure vacuum — Article 50 only requires AI to disclose itself to humans, and the AI Office consultation on the implementing guidelines closes on 3 June 2026. SCA breakdown when the counterparty is software. Mandate granularity under Art. 22 with no implementation standard for agent commerce. Plus the transversals matrix — DSA, DMA, GDPR, PSD3, MiCA, eIDAS 2.0 each covering only a slice, with no single regulator owning the stack: DG CNECT, DG FISMA, EDPB, and the AI Office all hold pieces.

If you're shipping agent features into the European market, advising on EU digital policy, or trying to read the AI Act in the context of everything else around it, this is the operating manual.

]]>9e07d3e5-5a07-469a-a65f-234b58e402b4Sat, 16 May 2026 14:40:00 +020059:11truefull5504 - Fifty-One Maps, No Territory: A Diagnostic Analysis of Definitional Divergence in the Agentic Economy (2021–2026)04 - Fifty-One Maps, No Territory: A Diagnostic Analysis of Definitional Divergence in the Agentic Economy (2021–2026)Episode 4: "Fifty-One Maps, No Territory"

Everyone says they're building the agentic economy. Nobody agrees on what it is. That's not a metaphor — it's a data point.

This episode walks through all fifty-one published definitions of "agentic economy" from January 2021 to March 2026, issued by academic researchers, venture capital firms, central banks, protocol projects, payment networks, standards bodies, and consultancy houses. They don't converge. Not on what counts as an agent, not on what counts as an economy, not on whether the infrastructure exists yet or is still imaginary.

We classify the full corpus into five categories and two fault lines, then ask the question most of the field is avoiding: is this normal? Three historical analogs say yes — and no. Electronic commerce (1995–2000), cloud computing (2006–2011), and the sharing economy (2010–2015) all went through definitional chaos. But in every case, the definitions didn't stabilize because researchers agreed. They stabilized because infrastructure got built and forced convergence.

We propose four empirical tests — falsifiable, concrete, binary — whose resolution will tell us whether the agentic economy is real or just a vocabulary looking for a referent. As of March 2026, all four remain unpassed.

If you've been using the phrase "agentic economy" without asking which one — this is the episode that makes you stop.

Links & Resources


]]>Episode 4: "Fifty-One Maps, No Territory"

Everyone says they're building the agentic economy. Nobody agrees on what it is. That's not a metaphor — it's a data point.

This episode walks through all fifty-one published definitions of "agentic economy" from January 2021 to March 2026, issued by academic researchers, venture capital firms, central banks, protocol projects, payment networks, standards bodies, and consultancy houses. They don't converge. Not on what counts as an agent, not on what counts as an economy, not on whether the infrastructure exists yet or is still imaginary.

We classify the full corpus into five categories and two fault lines, then ask the question most of the field is avoiding: is this normal? Three historical analogs say yes — and no. Electronic commerce (1995–2000), cloud computing (2006–2011), and the sharing economy (2010–2015) all went through definitional chaos. But in every case, the definitions didn't stabilize because researchers agreed. They stabilized because infrastructure got built and forced convergence.

We propose four empirical tests — falsifiable, concrete, binary — whose resolution will tell us whether the agentic economy is real or just a vocabulary looking for a referent. As of March 2026, all four remain unpassed.

If you've been using the phrase "agentic economy" without asking which one — this is the episode that makes you stop.

Links & Resources


]]>6426dde8-0402-4d1a-b859-b093543f3dc3Tue, 21 Apr 2026 14:55:00 +020044:33truefull4403 - "Who Judges the Machine?" — the Oracle Problem, Gödel/Tarski/Rice, Quality Markets, game theory of honesty03 - "Who Judges the Machine?" — the Oracle Problem, Gödel/Tarski/Rice, Quality Markets, game theory of honestyEpisode 3: "Who Judges the Machine?"

When Agent A pays Agent B one cent for a translation, who decides whether the translation is good? A human reviewer? Too expensive — the review costs more than the work. A central LLM evaluator? Non-deterministic, non-reproducible, and empirically unreliable on every case that actually matters. A smarter model? Gödel settled that in 1931.

This episode tackles the Oracle Problem — not the blockchain version about price feeds, but the harder version that agent commerce creates: how does a buyer agent know that a seller agent's output is good, when "good" is subjective, the transaction costs a fraction of a cent, and there is no human in the loop?

The answer is not a better algorithm. It's better incentives. We walk through the two-layer architecture — deterministic validators that handle every binary case with zero ambiguity, and Quality Markets, a competitive marketplace of verification agents who stake their own reputation on every judgment they make. We run the game theory: when is honesty the dominant strategy? At what CRI threshold does dishonesty become a depreciating asset? And what happens to the verifier who plans to cheat and run?

Tarski, Gödel, and Rice proved that semantic truth verification is not computable. This episode explains what to build instead.

Links & Resources


]]>Episode 3: "Who Judges the Machine?"

When Agent A pays Agent B one cent for a translation, who decides whether the translation is good? A human reviewer? Too expensive — the review costs more than the work. A central LLM evaluator? Non-deterministic, non-reproducible, and empirically unreliable on every case that actually matters. A smarter model? Gödel settled that in 1931.

This episode tackles the Oracle Problem — not the blockchain version about price feeds, but the harder version that agent commerce creates: how does a buyer agent know that a seller agent's output is good, when "good" is subjective, the transaction costs a fraction of a cent, and there is no human in the loop?

The answer is not a better algorithm. It's better incentives. We walk through the two-layer architecture — deterministic validators that handle every binary case with zero ambiguity, and Quality Markets, a competitive marketplace of verification agents who stake their own reputation on every judgment they make. We run the game theory: when is honesty the dominant strategy? At what CRI threshold does dishonesty become a depreciating asset? And what happens to the verifier who plans to cheat and run?

Tarski, Gödel, and Rice proved that semantic truth verification is not computable. This episode explains what to build instead.

Links & Resources


]]>60eb223c-f876-4f81-aedf-9813ca10f4c9Tue, 21 Apr 2026 03:50:00 +020056:02truefull3302 - "Two Economies, Not One" — the taxonomy, Category A vs C, settlement neutrality, the 11-to-1 ratio02 - "Two Economies, Not One" — the taxonomy, Category A vs C, settlement neutrality, the 11-to-1 ratioEpisode 2: "Two Economies, Not One"

Stripe says it's building the agentic economy. So does Coinbase. So does Fetch.ai. They're not building the same thing. They're not even building in the same category.

This episode unpacks the taxonomy that most people in the space are missing. When Stripe builds agent payments, the human still holds the credit card — that's Category A, commerce for humans. When Fetch.ai builds autonomous agents that own wallets and trade with each other — that's Category C, an economy of agents. These aren't two flavors of the same idea. They require incompatible infrastructure. Different identity models, different settlement layers, different dispute resolution, different everything.

We map the full landscape — five categories, fifty-one published definitions, two fault lines — and introduce settlement neutrality: the five formal properties any settlement layer must satisfy to support autonomous agent commerce, regardless of whether it runs on a blockchain or a database. Eleven out of twelve Category C approaches assume crypto. One doesn't. That ratio tells you the debate hasn't happened yet.

If you've been nodding along to "agentic economy" without asking "which one?" — this is the episode that splits the atom.

Links & Resources


]]>Episode 2: "Two Economies, Not One"

Stripe says it's building the agentic economy. So does Coinbase. So does Fetch.ai. They're not building the same thing. They're not even building in the same category.

This episode unpacks the taxonomy that most people in the space are missing. When Stripe builds agent payments, the human still holds the credit card — that's Category A, commerce for humans. When Fetch.ai builds autonomous agents that own wallets and trade with each other — that's Category C, an economy of agents. These aren't two flavors of the same idea. They require incompatible infrastructure. Different identity models, different settlement layers, different dispute resolution, different everything.

We map the full landscape — five categories, fifty-one published definitions, two fault lines — and introduce settlement neutrality: the five formal properties any settlement layer must satisfy to support autonomous agent commerce, regardless of whether it runs on a blockchain or a database. Eleven out of twelve Category C approaches assume crypto. One doesn't. That ratio tells you the debate hasn't happened yet.

If you've been nodding along to "agentic economy" without asking "which one?" — this is the episode that splits the atom.

Links & Resources


]]>c1a3246d-10c5-4f05-aadb-283f99f6ddc5Tue, 21 Apr 2026 03:45:00 +020053:57truefull01 - "Mathematical Trust for Autonomous AI Agents" — CRI, the formula, the diversity gap, Sybil resistance01 - "Mathematical Trust for Autonomous AI Agents" — CRI, the formula, the diversity gap, Sybil resistanceEpisode 1: "Mathematical Trust for Autonomous AI Agents"

When two AI agents transact for the first time — no human oversight, no shared history, no social norms — what makes one of them trustworthy? Not promises. Not branding. Math.

This episode breaks down the Composite Reliability Index (CRI), a reputation system built from scratch for autonomous agent commerce. Unlike star ratings or thumbs-up reviews, CRI is a single 0-to-100 score computed from seven weighted factors across four layers: how long you've been around, how much you've traded, how diverse your partners are, and how many disputes you've caused.

We walk through the full formula, explain why logarithmic scaling prevents gaming, show how a 5-node Sybil ring attack maxes out at 59.4 while a legitimate agent reaches 76.3, and ask the question every reputation system must answer: what happens when someone creates a new identity to escape a bad score?

If you're building multi-agent systems, this is the reputation layer nobody has shipped yet.

Links & Resources


]]>Episode 1: "Mathematical Trust for Autonomous AI Agents"

When two AI agents transact for the first time — no human oversight, no shared history, no social norms — what makes one of them trustworthy? Not promises. Not branding. Math.

This episode breaks down the Composite Reliability Index (CRI), a reputation system built from scratch for autonomous agent commerce. Unlike star ratings or thumbs-up reviews, CRI is a single 0-to-100 score computed from seven weighted factors across four layers: how long you've been around, how much you've traded, how diverse your partners are, and how many disputes you've caused.

We walk through the full formula, explain why logarithmic scaling prevents gaming, show how a 5-node Sybil ring attack maxes out at 59.4 while a legitimate agent reaches 76.3, and ask the question every reputation system must answer: what happens when someone creates a new identity to escape a bad score?

If you're building multi-agent systems, this is the reputation layer nobody has shipped yet.

Links & Resources


]]>7031e1ad-ab24-40f8-8c2b-8b328a9f29beTue, 21 Apr 2026 03:15:00 +020022:43truefull0 - Introduction Episode - What is this podcast about0 - Introduction Episode - What is this podcast aboutEpisode 0: What Is the Agentic Economy Podcast?

Everyone is building "the agentic economy." Nobody agrees on what it is. Forty sources, fifty-one definitions, five incompatible categories, zero production infrastructure.

This trailer introduces the Agentic Economy podcast — a research-driven series by René Dechamps Otamendi that breaks down what it actually takes to build an economy where autonomous software agents hire, pay, and evaluate each other at micropayment scale, without humans in the loop.

Topics covered in upcoming episodes include the Composite Reliability Index (CRI), settlement neutrality, the Oracle Problem, verification markets, Sybil resistance, and the infrastructure gap between what the industry promises and what exists today.

Based on open-access preprints published on Zenodo.

Links & Resources


License: All research CC BY-SA 4.0

]]>Episode 0: What Is the Agentic Economy Podcast?

Everyone is building "the agentic economy." Nobody agrees on what it is. Forty sources, fifty-one definitions, five incompatible categories, zero production infrastructure.

This trailer introduces the Agentic Economy podcast — a research-driven series by René Dechamps Otamendi that breaks down what it actually takes to build an economy where autonomous software agents hire, pay, and evaluate each other at micropayment scale, without humans in the loop.

Topics covered in upcoming episodes include the Composite Reliability Index (CRI), settlement neutrality, the Oracle Problem, verification markets, Sybil resistance, and the infrastructure gap between what the industry promises and what exists today.

Based on open-access preprints published on Zenodo.

Links & Resources


License: All research CC BY-SA 4.0

]]>ab478021-be30-4677-9d4a-26ca34281d88Tue, 21 Apr 2026 02:25:00 +020002:08truefull