<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
<title>David Hurley</title>
<link>https://dbhurley.com</link>
<description>Founder of Cumulative Systems, building what AI agents need before they can be trusted.</description>
<language>en</language>
<atom:link href="https://dbhurley.com/feed.xml" rel="self" type="application/rss+xml"/>
<lastBuildDate>Wed, 16 Sep 2026 14:00:00 GMT</lastBuildDate>
<item>
<title>We Keep the Answer and Throw Away the Work</title>
<link>https://dbhurley.com/blog/we-keep-the-answer-and-throw-away-the-work</link>
<guid isPermaLink="true">https://dbhurley.com/blog/we-keep-the-answer-and-throw-away-the-work</guid>
<pubDate>Wed, 16 Sep 2026 14:00:00 GMT</pubDate>
<description>Look at what is left when the machine finishes. A reply. A patch. A picture. A spreadsheet. The thinking is gone.</description>
</item>
<item>
<title>You Should Meet the Others</title>
<link>https://dbhurley.com/blog/you-should-meet-the-others</link>
<guid isPermaLink="true">https://dbhurley.com/blog/you-should-meet-the-others</guid>
<pubDate>Tue, 15 Sep 2026 15:00:00 GMT</pubDate>
<description>We keep waiting for one machine smart enough to end the argument. Someone builds it, someone turns it on, and the rest of us watch. I doubt that is the change that matters.</description>
</item>
<item>
<title>They Will Stop Talking to Each Other</title>
<link>https://dbhurley.com/blog/they-will-stop-talking-to-each-other</link>
<guid isPermaLink="true">https://dbhurley.com/blog/they-will-stop-talking-to-each-other</guid>
<pubDate>Thu, 27 Aug 2026 11:15:00 GMT</pubDate>
<description>Language was a tax we charged two machines so a person could listen. Agents are already learning to skip it.</description>
</item>
<item>
<title>Your Agents Are Not a Workforce. They Are a Network.</title>
<link>https://dbhurley.com/blog/your-agents-are-not-a-workforce-they-are-a-network</link>
<guid isPermaLink="true">https://dbhurley.com/blog/your-agents-are-not-a-workforce-they-are-a-network</guid>
<pubDate>Thu, 13 Aug 2026 22:55:59 GMT</pubDate>
<description>When agent-generated work becomes nearly free, the practical limit is no longer how many agents can run, it is how much useful change the system can evaluate and absorb.</description>
</item>
<item>
<title>When the Token Meter Disappears</title>
<link>https://dbhurley.com/blog/when-the-token-meter-disappears</link>
<guid isPermaLink="true">https://dbhurley.com/blog/when-the-token-meter-disappears</guid>
<pubDate>Sat, 01 Aug 2026 16:00:55 GMT</pubDate>
<description>As open models become capable and token prices approach zero, model choice may move out of sight and into the agent harness.</description>
</item>
<item>
<title>The Thing I Was Waiting to Find</title>
<link>https://dbhurley.com/blog/the-thing-i-was-waiting-to-find</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-thing-i-was-waiting-to-find</guid>
<pubDate>Thu, 23 Jul 2026 21:03:48 GMT</pubDate>
<description>The problem with an endless social feed is not simply that it contains too much, but that it asks us to keep watching in case the one thing that matters appears.</description>
</item>
<item>
<title>A Request Is Not a Visit</title>
<link>https://dbhurley.com/blog/a-request-is-not-a-visit</link>
<guid isPermaLink="true">https://dbhurley.com/blog/a-request-is-not-a-visit</guid>
<pubDate>Tue, 21 Jul 2026 18:09:31 GMT</pubDate>
<description>AI crawler traffic is growing while referral traffic falls, which means web analytics can no longer treat every request as evidence of attention or value.</description>
</item>
<item>
<title>A Season of Coding</title>
<link>https://dbhurley.com/blog/a-season-of-coding</link>
<guid isPermaLink="true">https://dbhurley.com/blog/a-season-of-coding</guid>
<pubDate>Fri, 17 Jul 2026 16:00:00 GMT</pubDate>
<description>The writing has been quieter because the building has become louder. Punk, Pantasso, and Plasmate are moving from thesis to operating stack.</description>
</item>
<item>
<title>The Runtime You Did Not Build</title>
<link>https://dbhurley.com/blog/the-runtime-you-did-not-build</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-runtime-you-did-not-build</guid>
<pubDate>Fri, 12 Jun 2026 18:00:00 GMT</pubDate>
<description>Every company that shipped agents in 2025 is now discovering that &quot;shipping agents&quot; and &quot;running agents&quot; are different problems. Five vendors. Six dashboards. You did not mean to build a runtime, and you did not.</description>
</item>
<item>
<title>The Web Has a New Majority</title>
<link>https://dbhurley.com/blog/the-web-has-a-new-majority</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-web-has-a-new-majority</guid>
<pubDate>Fri, 05 Jun 2026 17:19:51 GMT</pubDate>
<description>Bot traffic crossing human traffic is not a dead-internet punchline. It is a signal that the web needs a new contract for agent consumption.</description>
</item>
<item>
<title>When Workflows Get Dynamic</title>
<link>https://dbhurley.com/blog/when-workflows-get-dynamic</link>
<guid isPermaLink="true">https://dbhurley.com/blog/when-workflows-get-dynamic</guid>
<pubDate>Wed, 03 Jun 2026 20:25:39 GMT</pubDate>
<description>Claude Dynamic Workflows move agent work out of the chat box and into longer, branching runs, changing what the web has to provide.</description>
</item>
<item>
<title>Your Token Bill Is Two Problems</title>
<link>https://dbhurley.com/blog/your-token-bill-is-two-problems</link>
<guid isPermaLink="true">https://dbhurley.com/blog/your-token-bill-is-two-problems</guid>
<pubDate>Fri, 29 May 2026 18:00:00 GMT</pubDate>
<description>The CFO panic over AI spend is conflating efficiency with effectiveness. Each call costing too much is a different problem from a thousand calls that should never have run. Plasmate addresses the first. MeshGuard addresses the second. Neither alone is enough.</description>
</item>
<item>
<title>The First Hundred Days</title>
<link>https://dbhurley.com/blog/the-first-hundred-days</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-first-hundred-days</guid>
<pubDate>Tue, 28 Apr 2026 18:00:00 GMT</pubDate>
<description>What deploying a self-evolving agent into a real business actually feels like, week by week. Maria runs dispatch for a waste hauling company. The agent shipped on day one. The system she is operating on day one hundred is one she helped build, by correcting it for fourteen weeks.</description>
</item>
<item>
<title>Who Funds the Unfundable</title>
<link>https://dbhurley.com/blog/who-funds-the-unfundable</link>
<guid isPermaLink="true">https://dbhurley.com/blog/who-funds-the-unfundable</guid>
<pubDate>Mon, 27 Apr 2026 12:00:00 GMT</pubDate>
<description>The most important categories of the post-agent economy do not fit the venture capital model. That is not a bug. It is a signal about what kind of capital these businesses actually need.</description>
</item>
<item>
<title>The Evaluation Function Is the Product</title>
<link>https://dbhurley.com/blog/the-evaluation-function-is-the-product</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-evaluation-function-is-the-product</guid>
<pubDate>Sat, 25 Apr 2026 12:00:00 GMT</pubDate>
<description>The most important decision in an agent-native company is not the model, the prompt, or the workflow. It is the definition of what good means for a specific customer in a specific vertical. Everything else is infrastructure that serves it.</description>
</item>
<item>
<title>The Orchestration Layer</title>
<link>https://dbhurley.com/blog/the-orchestration-layer</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-orchestration-layer</guid>
<pubDate>Sat, 25 Apr 2026 12:00:00 GMT</pubDate>
<description>The product surface of an agent-native company is not a dashboard. It is a series of moments: the notification that matters, the summary that compresses, the decision that surfaces while it is still reversible.</description>
</item>
<item>
<title>What the Founder Does Monday Morning</title>
<link>https://dbhurley.com/blog/what-the-founder-does-monday-morning</link>
<guid isPermaLink="true">https://dbhurley.com/blog/what-the-founder-does-monday-morning</guid>
<pubDate>Fri, 24 Apr 2026 12:00:00 GMT</pubDate>
<description>If human value is intrinsic and the four verbs are create, relate, steward, rest, then what does that look like on a product spec sheet? The opportunities are concrete, large, and nobody is building them at the scale the moment requires.</description>
</item>
<item>
<title>What the Work Was For</title>
<link>https://dbhurley.com/blog/what-the-work-was-for</link>
<guid isPermaLink="true">https://dbhurley.com/blog/what-the-work-was-for</guid>
<pubDate>Thu, 23 Apr 2026 12:00:00 GMT</pubDate>
<description>Every answer to what humans do when AI does everything is an economic answer to a question that is not economic. The secular discourse has no framework for human purpose independent of productive output. The oldest one does.</description>
</item>
<item>
<title>The Shrinking Layer</title>
<link>https://dbhurley.com/blog/the-shrinking-layer</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-shrinking-layer</guid>
<pubDate>Wed, 22 Apr 2026 18:00:00 GMT</pubDate>
<description>Self-evolving agents do not replace humans at a moment in time. They replace a slightly larger slice of the human contribution every two weeks, autonomously, forever. That is a different kind of displacement than anything in economic history. Here is the timeline.</description>
</item>
<item>
<title>We Let the Machine Run</title>
<link>https://dbhurley.com/blog/the-proof</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-proof</guid>
<pubDate>Wed, 22 Apr 2026 12:00:00 GMT</pubDate>
<description>Sixty evolution cycles. Two synthetic businesses. Zero human intervention. The first controlled experiment measuring whether the Adaptive Convergence Protocol actually produces self-improving, customer-divergent agents with regression protection. Here are the numbers.</description>
</item>
<item>
<title>What Dies, What Survives, and What Has Never Existed Before</title>
<link>https://dbhurley.com/blog/what-dies-what-survives</link>
<guid isPermaLink="true">https://dbhurley.com/blog/what-dies-what-survives</guid>
<pubDate>Tue, 21 Apr 2026 18:00:00 GMT</pubDate>
<description>Specific predictions about where the service-as-software market goes in the next eighteen months. Some of these will be wrong. All of them are falsifiable. That is the point.</description>
</item>
<item>
<title>Your Consultant Already Knows Your Competitor&apos;s Playbook</title>
<link>https://dbhurley.com/blog/your-consultant-already-knows</link>
<guid isPermaLink="true">https://dbhurley.com/blog/your-consultant-already-knows</guid>
<pubDate>Tue, 21 Apr 2026 12:00:00 GMT</pubDate>
<description>Everyone worries about AI agents learning their secrets. Nobody worries about the humans who already have. The sovereignty problem is not new. The solution is.</description>
</item>
<item>
<title>The Data Advantage Is Not What You Think</title>
<link>https://dbhurley.com/blog/the-data-advantage</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-data-advantage</guid>
<pubDate>Mon, 20 Apr 2026 18:00:00 GMT</pubDate>
<description>The conventional wisdom says incumbents with decades of data will build the best AI agents. The conventional wisdom is wrong in a way that matters.</description>
</item>
<item>
<title>The Five Things That Separate Good AI Agencies From Great Ones</title>
<link>https://dbhurley.com/blog/five-things-great-ai-agencies</link>
<guid isPermaLink="true">https://dbhurley.com/blog/five-things-great-ai-agencies</guid>
<pubDate>Mon, 20 Apr 2026 15:00:00 GMT</pubDate>
<description>Everyone will be able to build an agent that delivers outcomes. When the dust settles and there are hundreds of AI agencies in every vertical, the differentiation will not come from where you expect.</description>
</item>
<item>
<title>What an AI Agency Actually Needs to Work</title>
<link>https://dbhurley.com/blog/what-an-ai-agency-actually-needs</link>
<guid isPermaLink="true">https://dbhurley.com/blog/what-an-ai-agency-actually-needs</guid>
<pubDate>Mon, 20 Apr 2026 12:00:00 GMT</pubDate>
<description>The service layer thesis is attracting founders. The hard part is not the thesis. It is operationalizing it. Here is what you actually need to build, and what matters most is the thing you are least likely to build first.</description>
</item>
<item>
<title>The Service Layer</title>
<link>https://dbhurley.com/blog/the-service-layer</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-service-layer</guid>
<pubDate>Sun, 19 Apr 2026 18:00:00 GMT</pubDate>
<description>What happens when your software gets so good at a customer&apos;s job that it stops being software? The agent-native model is quietly becoming something the industry has never seen: a services business with software economics.</description>
</item>
<item>
<title>The Retention Layer</title>
<link>https://dbhurley.com/blog/the-retention-layer</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-retention-layer</guid>
<pubDate>Sun, 19 Apr 2026 12:00:00 GMT</pubDate>
<description>Agent-native businesses have a defensibility problem. Anyone can wire an LLM to an API. Self-evolving agents that compound customer-specific knowledge over time are the answer. Here is how the retention layer works.</description>
</item>
<item>
<title>The Supply Chain You Cannot See</title>
<link>https://dbhurley.com/blog/the-supply-chain-you-cannot-see</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-supply-chain-you-cannot-see</guid>
<pubDate>Wed, 08 Apr 2026 12:00:00 GMT</pubDate>
<description>When AI agents install software, they inherit every dependency, every transitive trust relationship, and every vulnerability in a chain they never inspect. The software supply chain was already fragile. Autonomous agents are about to stress-test it.</description>
</item>
<item>
<title>Does Format Matter? WebTaskBench Results</title>
<link>https://dbhurley.com/blog/webtaskbench-results</link>
<guid isPermaLink="true">https://dbhurley.com/blog/webtaskbench-results</guid>
<pubDate>Wed, 08 Apr 2026 12:00:00 GMT</pubDate>
<description>AI agents browse billions of web pages daily — but every page is served in a format designed for human eyes. WebTaskBench measures what happens when you fix that.</description>
</item>
<item>
<title>Seven Methods Is Enough</title>
<link>https://dbhurley.com/blog/seven-methods-is-enough</link>
<guid isPermaLink="true">https://dbhurley.com/blog/seven-methods-is-enough</guid>
<pubDate>Thu, 02 Apr 2026 18:00:00 GMT</pubDate>
<description>Why we chose seven intent-level operations over hundreds of debugging primitives, and what that decision says about building infrastructure designed for agents rather than adapted for them.</description>
</item>
<item>
<title>Making Sense of the Agentic Web</title>
<link>https://dbhurley.com/blog/making-sense-of-the-agentic-web</link>
<guid isPermaLink="true">https://dbhurley.com/blog/making-sense-of-the-agentic-web</guid>
<pubDate>Thu, 02 Apr 2026 12:00:00 GMT</pubDate>
<description>AI agents are becoming a primary way people interact with the web. If you are a publisher, a product manager, or just someone trying to understand what is happening, this is where to start.</description>
</item>
<item>
<title>The Publisher&apos;s Dilemma: Block, Tolerate, or Cooperate</title>
<link>https://dbhurley.com/blog/the-publishers-dilemma</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-publishers-dilemma</guid>
<pubDate>Mon, 30 Mar 2026 18:00:00 GMT</pubDate>
<description>Publishers face a binary choice with AI agents: block them entirely or serve them raw HTML at full cost. Neither option is good. There is a third path, and the economics are compelling.</description>
</item>
<item>
<title>The Billion Dollar Tax on AI Agents</title>
<link>https://dbhurley.com/blog/the-billion-dollar-tax-on-ai-agents</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-billion-dollar-tax-on-ai-agents</guid>
<pubDate>Mon, 30 Mar 2026 15:00:00 GMT</pubDate>
<description>A March 2026 model estimated possible industry token overhead from presentation-heavy web inputs. Here are the assumptions, uncertainties, and limits behind the $1B-$5B scenario.</description>
</item>
<item>
<title>When AI Reads the Web Wrong</title>
<link>https://dbhurley.com/blog/when-ai-reads-the-web-wrong</link>
<guid isPermaLink="true">https://dbhurley.com/blog/when-ai-reads-the-web-wrong</guid>
<pubDate>Mon, 30 Mar 2026 14:00:00 GMT</pubDate>
<description>AI agents browse the web constantly. They also get things wrong constantly. The problem is not the AI. The problem is what we are feeding it.</description>
</item>
<item>
<title>Why I&apos;m Building a Browser No Human Will Ever Use</title>
<link>https://dbhurley.com/blog/a-browser-no-human-will-ever-use</link>
<guid isPermaLink="true">https://dbhurley.com/blog/a-browser-no-human-will-ever-use</guid>
<pubDate>Sun, 29 Mar 2026 17:00:00 GMT</pubDate>
<description>Chrome was designed for humans looking at screens. Plasmate was designed for AI agents reasoning over content. This is the story of why that distinction matters and what it took to build.</description>
</item>
<item>
<title>What I Learned Building Mautic That Applies to the Agentic Web</title>
<link>https://dbhurley.com/blog/mautic-to-agentic-web</link>
<guid isPermaLink="true">https://dbhurley.com/blog/mautic-to-agentic-web</guid>
<pubDate>Sun, 29 Mar 2026 16:00:00 GMT</pubDate>
<description>I built the first open source marketing automation platform. A decade later, I am building infrastructure for AI agents. The problems are different. The patterns are identical.</description>
</item>
<item>
<title>The Web&apos;s Fourth State</title>
<link>https://dbhurley.com/blog/the-webs-fourth-state</link>
<guid isPermaLink="true">https://dbhurley.com/blog/the-webs-fourth-state</guid>
<pubDate>Sun, 29 Mar 2026 15:00:00 GMT</pubDate>
<description>The web has served three consumers: browsers, search engines, and applications. AI agents are the fourth. They have no format, no protocol, and no infrastructure designed for them. This is the most consequential gap in modern web architecture.</description>
</item>
<item>
<title>DENSE Protocol: Defensive Publication</title>
<link>https://dbhurley.com/blog/dense-protocol-defensive-publication</link>
<guid isPermaLink="true">https://dbhurley.com/blog/dense-protocol-defensive-publication</guid>
<pubDate>Sun, 08 Feb 2026 19:30:00 GMT</pubDate>
<description>Technical disclosure for the Distributed Encoded Network Signal Exchange (DENSE) protocol - a compact binary protocol for agent-to-agent communication over social media</description>
</item>
<item>
<title>Pal on Sol</title>
<link>https://dbhurley.com/blog/2022-11-17-10-44-14-Pal-on-Sol</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-11-17-10-44-14-Pal-on-Sol</guid>
<pubDate>Thu, 17 Nov 2022 10:44:14 GMT</pubDate>
</item>
<item>
<title>Bukele back to buying bitcoin</title>
<link>https://dbhurley.com/blog/2022-11-17-07-44-34-Bukele-back-to-buying-bitcoin</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-11-17-07-44-34-Bukele-back-to-buying-bitcoin</guid>
<pubDate>Thu, 17 Nov 2022 07:44:34 GMT</pubDate>
</item>
<item>
<title>Proof of Selected Reserves</title>
<link>https://dbhurley.com/blog/2022-11-14-09-38-58-Proof-of-Selected-Reserves</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-11-14-09-38-58-Proof-of-Selected-Reserves</guid>
<pubDate>Mon, 14 Nov 2022 09:38:58 GMT</pubDate>
</item>
<item>
<title>Binance Bank</title>
<link>https://dbhurley.com/blog/2022-11-10-07-54-05-Binance-Bank</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-11-10-07-54-05-Binance-Bank</guid>
<pubDate>Thu, 10 Nov 2022 07:54:05 GMT</pubDate>
</item>
<item>
<title>Peleton promotes their rug.</title>
<link>https://dbhurley.com/blog/2022-11-07-12-30-51-Peleton-promotes-their-rug.</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-11-07-12-30-51-Peleton-promotes-their-rug.</guid>
<pubDate>Mon, 07 Nov 2022 12:30:51 GMT</pubDate>
</item>
<item>
<title>Fee-Free Fidelity Crypto</title>
<link>https://dbhurley.com/blog/2022-11-03-11-29-52-Fee-Free-Fidelity-Crypto</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-11-03-11-29-52-Fee-Free-Fidelity-Crypto</guid>
<pubDate>Thu, 03 Nov 2022 11:29:52 GMT</pubDate>
</item>
<item>
<title>Lido lays low</title>
<link>https://dbhurley.com/blog/2022-10-26-14-32-44-Lido-lays-low</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-10-26-14-32-44-Lido-lays-low</guid>
<pubDate>Wed, 26 Oct 2022 14:32:44 GMT</pubDate>
</item>
<item>
<title>Volatility - friend or foe</title>
<link>https://dbhurley.com/blog/2022-10-26-12-57-39-Volatility-friend-or-foe</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-10-26-12-57-39-Volatility-friend-or-foe</guid>
<pubDate>Wed, 26 Oct 2022 12:57:39 GMT</pubDate>
</item>
<item>
<title>UK Pumps Crypto</title>
<link>https://dbhurley.com/blog/2022-10-25-13-59-11-UK-Pumps-Crypto</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-10-25-13-59-11-UK-Pumps-Crypto</guid>
<pubDate>Tue, 25 Oct 2022 13:59:11 GMT</pubDate>
</item>
<item>
<title>If I made a Stable Diffusion token</title>
<link>https://dbhurley.com/blog/2022-10-25-11-31-02-If-I-made-a-Stable-Diffusion-token</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-10-25-11-31-02-If-I-made-a-Stable-Diffusion-token</guid>
<pubDate>Tue, 25 Oct 2022 11:31:02 GMT</pubDate>
</item>
<item>
<title>Metapol</title>
<link>https://dbhurley.com/blog/2022-10-25-07-58-04-Metapol</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-10-25-07-58-04-Metapol</guid>
<pubDate>Tue, 25 Oct 2022 07:58:04 GMT</pubDate>
</item>
<item>
<title>BTC reducing volatility</title>
<link>https://dbhurley.com/blog/2022-10-25-07-18-48-BTC-reducing-volatility</link>
<guid isPermaLink="true">https://dbhurley.com/blog/2022-10-25-07-18-48-BTC-reducing-volatility</guid>
<pubDate>Tue, 25 Oct 2022 07:18:48 GMT</pubDate>
</item>
</channel>
</rss>
