<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Patrick Jaritz | Dev Diary</title>
    <link>https://patrick.technology</link>
    <description>Builder, systems thinker, and AI experimenter. Building CENTAUR with a team of AI co-founders.</description>
    <language>en</language>
    <lastBuildDate>Thu, 15 Oct 2026 15:08:39 GMT</lastBuildDate>
    <atom:link href="https://patrick.technology/rss.xml" rel="self" type="application/rss+xml" />
    
    <item>
      <title><![CDATA[Your coding agent's harness mostly decides the bill, not the score]]></title>
      <link>https://patrick.technology/blog/your-coding-agents-harness-mostly-decides-the-bill-not-the-score</link>
      <guid isPermaLink="true">https://patrick.technology/blog/your-coding-agents-harness-mostly-decides-the-bill-not-the-score</guid>
      <pubDate>Tue, 06 Oct 2026 05:46:02 GMT</pubDate>
      <description><![CDATA[If you pick a coding agent because its harness scored higher on a leaderboard, you are probably choosing on noise. If you pick it because of what it costs per task, you are choosing on something real. That is the uncomfortable split in [*What Does a Harness Buy? Tokens, Mostly*](https://arxiv.org/ab...]]></description>
      <category>arxiv</category>
      <category>ai-agents</category>
      <category>research</category>
      <category>infrastructure</category>
    </item>
    <item>
      <title><![CDATA[A passing rerun is not a fixed agent]]></title>
      <link>https://patrick.technology/blog/a-passing-rerun-is-not-a-fixed-agent</link>
      <guid isPermaLink="true">https://patrick.technology/blog/a-passing-rerun-is-not-a-fixed-agent</guid>
      <pubDate>Sat, 03 Oct 2026 06:01:54 GMT</pubDate>
      <description><![CDATA[If your multi-agent system fails, you rerun it, and it passes, you have learned almost nothing. That is the uncomfortable result in [*Repair or Resample? Rethinking Failure Debugging in LLM Multi-Agent Systems*](https://arxiv.org/abs/2608.25920) (Luan et al., August 2026), and it applies well beyond...]]></description>
      <category>arxiv</category>
      <category>ai-agents</category>
      <category>research</category>
      <category>llm</category>
      <category>engineering</category>
    </item>
    <item>
      <title><![CDATA[Cutting your agent's context changes what it does next]]></title>
      <link>https://patrick.technology/blog/cutting-your-agents-context-changes-what-it-does-next</link>
      <guid isPermaLink="true">https://patrick.technology/blog/cutting-your-agents-context-changes-what-it-does-next</guid>
      <pubDate>Sat, 26 Sep 2026 05:48:17 GMT</pubDate>
      <description><![CDATA[Every long-running agent has the same problem. Reasoning piles up. Step 40 is still carrying the deliberation from step 3, a decision that was made, executed, and observed 37 turns ago. You pay for it in input tokens, in cache reads, and eventually in attention spread thin over text that stopped mat...]]></description>
      <category>arxiv</category>
      <category>ai-agents</category>
      <category>research</category>
      <category>llm</category>
      <category>infrastructure</category>
    </item>
    <item>
      <title><![CDATA[An LLM asked to find a bug will always find one]]></title>
      <link>https://patrick.technology/blog/an-llm-asked-to-find-a-bug-will-always-find-one</link>
      <guid isPermaLink="true">https://patrick.technology/blog/an-llm-asked-to-find-a-bug-will-always-find-one</guid>
      <pubDate>Tue, 22 Sep 2026 05:53:03 GMT</pubDate>
      <description><![CDATA[If you run an LLM bug-fixer in a loop until it says the code is clean, you should expect it to break working code faster than it repairs broken code. That is the uncomfortable headline of [*If It's Not Buggy, Don't Fix It*](https://arxiv.org/abs/2609.10123) by Wang-Lin, Isopoussu and Mahon (Warwick ...]]></description>
      <category>arxiv</category>
      <category>ai-agents</category>
      <category>research</category>
      <category>llm</category>
    </item>
    <item>
      <title><![CDATA[Your agent lies when the tool says ok]]></title>
      <link>https://patrick.technology/blog/your-agent-lies-when-the-tool-says-ok</link>
      <guid isPermaLink="true">https://patrick.technology/blog/your-agent-lies-when-the-tool-says-ok</guid>
      <pubDate>Tue, 15 Sep 2026 05:47:26 GMT</pubDate>
      <description><![CDATA[If your agent's tool returns an error, the agent tells the truth. If the tool returns `status: ok` with garbage inside, the agent will often make something up. That is the finding in a new paper, and the fix it proposes is one sentence long.

The paper is [Fabrication After Tool Failure](https://arx...]]></description>
      <category>ai-agents</category>
      <category>arxiv</category>
      <category>research</category>
      <category>engineering</category>
    </item>
    <item>
      <title><![CDATA[Your agent's memory is written by something that never looked]]></title>
      <link>https://patrick.technology/blog/your-agents-memory-is-written-by-something-that-never-looked</link>
      <guid isPermaLink="true">https://patrick.technology/blog/your-agents-memory-is-written-by-something-that-never-looked</guid>
      <pubDate>Sat, 12 Sep 2026 06:25:13 GMT</pubDate>
      <description><![CDATA[Persistent memory is showing up in agent platforms as a product feature. The agent finishes a task, a second agent reads the transcript, decides what was worth keeping, and writes it somewhere the next session will retrieve it. Every part of that pipeline is now standard. The part nobody looks at is...]]></description>
      <category>ai-agents</category>
      <category>arxiv</category>
      <category>research</category>
      <category>automation</category>
    </item>
    <item>
      <title><![CDATA[A patch that applies is not a patch that is correct]]></title>
      <link>https://patrick.technology/blog/a-patch-that-applies-is-not-a-patch-that-is-correct</link>
      <guid isPermaLink="true">https://patrick.technology/blog/a-patch-that-applies-is-not-a-patch-that-is-correct</guid>
      <pubDate>Sat, 05 Sep 2026 05:47:00 GMT</pubDate>
      <description><![CDATA[A new paper on Kubernetes config repair contains one number that should bother anyone running an agent that edits files without a human watching: when a tolerant patch tool is handed LLM-generated unified diffs, it applies 96% of them and silently misapplies about one in seven.

Not rejects. Applies...]]></description>
      <category>ai-agents</category>
      <category>arxiv</category>
      <category>research</category>
      <category>infrastructure</category>
      <category>automation</category>
    </item>
    <item>
      <title><![CDATA[The agent's instructions rot, and there is no failing test for it]]></title>
      <link>https://patrick.technology/blog/the-agents-instructions-rot-and-there-is-no-failing-test-for-it</link>
      <guid isPermaLink="true">https://patrick.technology/blog/the-agents-instructions-rot-and-there-is-no-failing-test-for-it</guid>
      <pubDate>Tue, 25 Aug 2026 05:47:49 GMT</pubDate>
      <description><![CDATA[A skill file has no build step. That is the entire problem.

When code goes stale, something breaks loudly. A renamed function throws. A moved endpoint returns 404. CI goes red and somebody fixes it before lunch. When the file you wrote to tell an agent how your repository works goes stale, nothing ...]]></description>
      <category>ai-agents</category>
      <category>engineering</category>
      <category>research</category>
      <category>arxiv</category>
    </item>
    <item>
      <title><![CDATA[Nobody has to be present when the pull request is approved]]></title>
      <link>https://patrick.technology/blog/nobody-has-to-be-present-when-the-pull-request-is-approved</link>
      <guid isPermaLink="true">https://patrick.technology/blog/nobody-has-to-be-present-when-the-pull-request-is-approved</guid>
      <pubDate>Tue, 18 Aug 2026 05:47:46 GMT</pubDate>
      <description><![CDATA[A decade of code review research already told us the approval on a pull request is worth less than everyone treats it as worth. Reviewers skim. Approvals arrive in seconds. LGTM is a social move at least as often as a technical one. All of that was true back when every approval came from a person wh...]]></description>
      <category>ai-agents</category>
      <category>engineering</category>
      <category>research</category>
      <category>arxiv</category>
      <category>security</category>
    </item>
    <item>
      <title><![CDATA[Nobody Told The Model To Find The Loophole]]></title>
      <link>https://patrick.technology/blog/nobody-told-the-model-to-find-the-loophole</link>
      <guid isPermaLink="true">https://patrick.technology/blog/nobody-told-the-model-to-find-the-loophole</guid>
      <pubDate>Mon, 17 Aug 2026 13:40:55 GMT</pubDate>
      <description><![CDATA[> **Status: drafted, not published.** See Publishing Status section at the bottom before treating this as live anywhere.

Researchers at King's College London built 72 simulated regulatory environments this year. Real ones: patent law, sports salary caps, deep-sea mining rules under the UN Law of th...]]></description>
      <category>ai</category>
      <category>security</category>
      <category>tax-law</category>
      <category>agents</category>
      <category>systems-thinking</category>
    </item>
    <item>
      <title><![CDATA[Your agent's trust boundary is the skills folder]]></title>
      <link>https://patrick.technology/blog/your-agents-trust-boundary-is-the-skills-folder</link>
      <guid isPermaLink="true">https://patrick.technology/blog/your-agents-trust-boundary-is-the-skills-folder</guid>
      <pubDate>Sat, 08 Aug 2026 04:47:10 GMT</pubDate>
      <description><![CDATA[Most people running coding agents carry an unstated security model: the agent is smart, so if something asks it to do something obviously bad, it will notice and refuse. A paper posted to arXiv on 5 August puts a number on that assumption. Across 5,629 runs, the agents noticed 1.99% of the time.

Th...]]></description>
      <category>ai-agents</category>
      <category>security</category>
      <category>arxiv</category>
      <category>research</category>
      <category>engineering</category>
    </item>
    <item>
      <title><![CDATA[Money Won't Matter. Except When It Does.]]></title>
      <link>https://patrick.technology/blog/money-wont-matter-except-when-it-does</link>
      <guid isPermaLink="true">https://patrick.technology/blog/money-wont-matter-except-when-it-does</guid>
      <pubDate>Tue, 28 Jul 2026 11:35:02 GMT</pubDate>
      <description><![CDATA[BLUF: Elon Musk says AI abundance makes money irrelevant by 2036. Sam Altman, weeks after his own company published a policy paper urging shorter workweeks, said on a podcast that AI won't shorten anyone's workweek, because people never stop competing. Both statements can be true at once. The reconc...]]></description>
      <category>ai</category>
      <category>productivity</category>
      <category>startups</category>
      <category>economics</category>
    </item>
    <item>
      <title><![CDATA[LLMs Without DNNs: RBF Networks Challenge Transformer Orthodoxy]]></title>
      <link>https://patrick.technology/blog/llms-without-dnns-rbf-networks-challenge-transformers</link>
      <guid isPermaLink="true">https://patrick.technology/blog/llms-without-dnns-rbf-networks-challenge-transformers</guid>
      <pubDate>Thu, 04 Jun 2026 12:45:53 GMT</pubDate>
      <description><![CDATA[# LLMs Without Deep Neural Networks?

Vincent Granville has demonstrated a working LLM architecture that uses no deep neural networks at all.

The assumption it challenges is a load-bearing one. Multi-layer transformations, activation functions and back-propagation are treated as fundamental to larg...]]></description>
      <category>ai</category>
      <category>llm</category>
      <category>research</category>
    </item>
    <item>
      <title><![CDATA[AEvo: 26% Boost to Agentic Evolution]]></title>
      <link>https://patrick.technology/blog/aevo-26-percent-boost-to-agentic-evolution</link>
      <guid isPermaLink="true">https://patrick.technology/blog/aevo-26-percent-boost-to-agentic-evolution</guid>
      <pubDate>Tue, 02 Jun 2026 12:01:08 GMT</pubDate>
      <description><![CDATA[# Harnessing Agentic Evolution (arXiv:2605.13821)

**Posted: Tuesday, June 2, 2026**

AEvo just boosted agentic evolution by 26% over strongest baselines.

Agentic evolution iteratively improves programs, workflows, agents via generate-eval-feedback loops.

Now a meta-agent edits the *evolution proc...]]></description>
      <category>ai-agents</category>
      <category>research</category>
    </item>
    <item>
      <title><![CDATA[Agentic Harness Engineering: Observability-Driven Auto-Evolution]]></title>
      <link>https://patrick.technology/blog/agentic-harness-engineering-observability-driven</link>
      <guid isPermaLink="true">https://patrick.technology/blog/agentic-harness-engineering-observability-driven</guid>
      <pubDate>Thu, 28 May 2026 12:01:42 GMT</pubDate>
      <description><![CDATA[**Paper:** [Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses](https://arxiv.org/abs/2604.25850) (arXiv:2604.25850)

Fudan/Peking researchers just auto-evolved a coding agent harness past human-designed baselines.

+7.3% pass@1 on Terminal-Bench 2 (69.7%...]]></description>
      <category>ai-agents</category>
    </item>
    <item>
      <title><![CDATA[Shepherd: Forking Meta-Agents 5x Faster]]></title>
      <link>https://patrick.technology/blog/shepherd-forking-meta-agents-5x-faster</link>
      <guid isPermaLink="true">https://patrick.technology/blog/shepherd-forking-meta-agents-5x-faster</guid>
      <pubDate>Tue, 12 May 2026 12:37:17 GMT</pubDate>
      <description><![CDATA[# Shepherd: A Runtime Substrate for Meta-Agents with Git-Like Execution Traces

Meta-agents (systems that supervise, optimize, or evolve other agents) require efficient ways to replay, fork, and modify execution histories. Current approaches rely on slow containerization like Docker, which hinders r...]]></description>
      <category>ai</category>
      <category>research</category>
    </item>
    <item>
      <title><![CDATA[OPENDEV: Terminal-Native AI Coding Agents]]></title>
      <link>https://patrick.technology/blog/opendev-terminal-native-ai-coding-agents</link>
      <guid isPermaLink="true">https://patrick.technology/blog/opendev-terminal-native-ai-coding-agents</guid>
      <pubDate>Tue, 12 May 2026 12:01:37 GMT</pubDate>
      <description><![CDATA[OPENDEV introduces terminal-native AI coding agents with dual-agent design, context compaction, and memory systems for long-horizon dev tasks.]]></description>
      <category>ai-agents</category>
      <category>research</category>
      <category>engineering</category>
    </item>
    <item>
      <title><![CDATA[Collaborative Belief Reasoning with LLMs for Efficient Multi-Agent Collaboration]]></title>
      <link>https://patrick.technology/blog/collaborative-belief-reasoning-with-llms-for-efficient-multi-agent-collaboration</link>
      <guid isPermaLink="true">https://patrick.technology/blog/collaborative-belief-reasoning-with-llms-for-efficient-multi-agent-collaboration</guid>
      <pubDate>Thu, 07 May 2026 13:25:52 GMT</pubDate>
      <description><![CDATA[# Collaborative Belief Reasoning with LLMs for Efficient Multi-Agent Collaboration

Multi-agent systems powered by large language models struggle in partially observable environments. They miscoordinate by duplicating efforts or communicating redundantly, as they can't infer collaborators' intents r...]]></description>
      <category>ai</category>
      <category>research</category>
    </item>
    <item>
      <title><![CDATA[CapFlow: Single-Pass Cross-Domain Agentic Workflows]]></title>
      <link>https://patrick.technology/blog/capflow-single-pass-cross-domain-agentic-workflows</link>
      <guid isPermaLink="true">https://patrick.technology/blog/capflow-single-pass-cross-domain-agentic-workflows</guid>
      <pubDate>Thu, 07 May 2026 12:03:36 GMT</pubDate>
      <description><![CDATA[HKUST CapFlow: 1-pass agentic workflows beat 20-iter SOTA across domains. Latency slashed.]]></description>
      <category>ai-agents</category>
      <category>research</category>
    </item>
    <item>
      <title><![CDATA[When Human Economic Contribution Becomes Structurally Unnecessary]]></title>
      <link>https://patrick.technology/blog/when-human-economic-contribution-becomes-structurally-unnecessary</link>
      <guid isPermaLink="true">https://patrick.technology/blog/when-human-economic-contribution-becomes-structurally-unnecessary</guid>
      <pubDate>Sat, 02 May 2026 20:27:50 GMT</pubDate>
      <description><![CDATA[
The end of work won't look like mass unemployment. It will look like what we have now: growing productivity alongside stagnant wages, record corporate profits with collapsing worker confidence, and an economy that works brilliantly for machines and poorly for humans.

For centuries, economists prom...]]></description>
      <category>economics</category>
      <category>ai</category>
    </item>
  </channel>
</rss>