<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>NexcurAI - Writing</title>
    <link>https://nexcur.ai/blog/</link>
    <description>Essays on operating with Claude inside a consulting firm. Four pillars: applied AI, security, GEO, and the economics of shipping.</description>
    <language>en-us</language>
    <lastBuildDate>Wed, 22 Apr 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://nexcur.ai/feed.xml" rel="self" type="application/rss+xml" />
    <managingEditor>hello@nexcur.ai (NexcurAI Editorial)</managingEditor>
    <webMaster>hello@nexcur.ai (NexcurAI)</webMaster>
    <copyright>Copyright 2026 NexcurAI Inc. All rights reserved.</copyright>
    <image>
      <url>https://nexcur.ai/og-image.png</url>
      <title>NexcurAI - Writing</title>
      <link>https://nexcur.ai/blog/</link>
    </image>

    <item>
      <title>The self-critique loop: how we get Claude to find its own mistakes</title>
      <link>https://nexcur.ai/blog/self-critique-loop.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/self-critique-loop.html</guid>
      <pubDate>Wed, 22 Apr 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>Every handbook runs through a self-critique pass before a human sees it. Prompt patterns and real traces from our production pipeline, with the four critique checks that catch most of our errors.</description>
    </item>

    <item>
      <title>Threat modeling goes from 3 days to 3 hours when Claude drives STRIDE</title>
      <link>https://nexcur.ai/blog/threat-modeling-3-hours.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/threat-modeling-3-hours.html</guid>
      <pubDate>Tue, 21 Apr 2026 00:00:00 +0000</pubDate>
      <category>Security</category>
      <dc:creator>the operator</dc:creator>
      <description>STRIDE is a good framework but a slow one. With Claude driving the enumeration and a human driving the judgment, a workshop that used to take three days fits into one afternoon. A full trace.</description>
    </item>

    <item>
      <title>What tool use actually unlocks</title>
      <link>https://nexcur.ai/blog/tool-use-unlocks.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/tool-use-unlocks.html</guid>
      <pubDate>Sun, 19 Apr 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>Tool use is the feature that turns Claude from an essay writer into a collaborator that can touch real systems. Here is what it unlocks, what it does not, and the design discipline that keeps tool-using agents from becoming a liability.</description>
    </item>

    <item>
      <title>The 200k window is a context engineering problem, not a memory problem</title>
      <link>https://nexcur.ai/blog/200k-context-engineering.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/200k-context-engineering.html</guid>
      <pubDate>Sun, 19 Apr 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>The 200k context is not memory. It is a staging area you engineer, layer by layer. Retrieval patterns for long-running engagements and why bigger windows do not remove the need for careful assembly.</description>
    </item>

    <item>
      <title>How to build a topical map that ranks in Google AND gets cited by ChatGPT</title>
      <link>https://nexcur.ai/blog/topical-map-dual-optimization.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/topical-map-dual-optimization.html</guid>
      <pubDate>Sat, 18 Apr 2026 00:00:00 +0000</pubDate>
      <category>SEO and GEO</category>
      <dc:creator>the operator</dc:creator>
      <description>A page that ranks on Google is not the same document as one cited by an answer engine. The three-layer topical map (hub, spokes, citation surfaces) that serves both surfaces without compromising either.</description>
    </item>

    <item>
      <title>Why fixed-scope pricing is 10x better</title>
      <link>https://nexcur.ai/blog/pricing-fixed-scope-10x.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/pricing-fixed-scope-10x.html</guid>
      <pubDate>Fri, 17 Apr 2026 00:00:00 +0000</pubDate>
      <category>Economics of shipping</category>
      <dc:creator>the operator</dc:creator>
      <description>Fixed scope is not a concession to the client. It is an operating discipline that changes what gets built, how fast, and whether the engagement ends. The math and the four change-request clauses that make it survive contact with reality.</description>
    </item>

    <item>
      <title>The incident response runbook nobody wrote because nobody had time</title>
      <link>https://nexcur.ai/blog/incident-response-runbook.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/incident-response-runbook.html</guid>
      <pubDate>Thu, 16 Apr 2026 00:00:00 +0000</pubDate>
      <category>Security</category>
      <dc:creator>the operator</dc:creator>
      <description>Most teams do not have an IR runbook. The ones that do have a document written three years ago that references tools they no longer use. Template plus how Claude shortens the write-up without hiding decisions that matter.</description>
    </item>

    <item>
      <title>Prompt caching changed our unit economics</title>
      <link>https://nexcur.ai/blog/prompt-caching-unit-economics.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/prompt-caching-unit-economics.html</guid>
      <pubDate>Wed, 15 Apr 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>When a 140 page handbook is generated against the same 20k token context forty times, the cache pays for the whole engagement. Break-even math with numbers from our own pipeline.</description>
    </item>

    <item>
      <title>The llms.txt discipline</title>
      <link>https://nexcur.ai/blog/llms-txt-discipline.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/llms-txt-discipline.html</guid>
      <pubDate>Tue, 14 Apr 2026 00:00:00 +0000</pubDate>
      <category>SEO and GEO</category>
      <dc:creator>the operator</dc:creator>
      <description>Most llms.txt files in the wild are either empty, sitemap clones, marketing copy, or stale. The six-section structure and quarterly review discipline we use to keep ours genuinely useful.</description>
    </item>

    <item>
      <title>OWASP LLM Top 10 as audit checks</title>
      <link>https://nexcur.ai/blog/owasp-llm-top-10-audit-checks.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/owasp-llm-top-10-audit-checks.html</guid>
      <pubDate>Sun, 12 Apr 2026 00:00:00 +0000</pubDate>
      <category>Security</category>
      <dc:creator>the operator</dc:creator>
      <description>The OWASP LLM Top 10 is a taxonomy, not a checklist. Two concrete audit checks per category, runnable by a security lead who does not need to become an AI researcher.</description>
    </item>

    <item>
      <title>The operator-to-senior ratio that works</title>
      <link>https://nexcur.ai/blog/operator-senior-ratio.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/operator-senior-ratio.html</guid>
      <pubDate>Sat, 11 Apr 2026 00:00:00 +0000</pubDate>
      <category>Economics of shipping</category>
      <dc:creator>the operator</dc:creator>
      <description>Traditional consulting firms are pyramids of juniors under a few partners. AI-native delivery inverts the ratio because Claude absorbs the junior work. The team structure that actually ships, and the failure mode to avoid.</description>
    </item>

    <item>
      <title>The handbook thesis, defended</title>
      <link>https://nexcur.ai/blog/handbook-thesis-defended.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/handbook-thesis-defended.html</guid>
      <pubDate>Fri, 10 Apr 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>Why we ship a long-form document at the end of every engagement, and why the slide deck is about to die as a consulting artifact. 38 minutes, with a companion sample you can read over lunch.</description>
    </item>

    <item>
      <title>Why we stopped writing one-shot prompts</title>
      <link>https://nexcur.ai/blog/stopped-writing-one-shot-prompts.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/stopped-writing-one-shot-prompts.html</guid>
      <pubDate>Wed, 08 Apr 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>Every deliverable runs through four stages now: draft, self-critique, human edit, eval. How we broke the pipeline apart and why one-shot prompts always underperform at engagement scale.</description>
    </item>

    <item>
      <title>Retainer vs sprint: picking the right shape</title>
      <link>https://nexcur.ai/blog/retainer-vs-sprint.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/retainer-vs-sprint.html</guid>
      <pubDate>Sun, 05 Apr 2026 00:00:00 +0000</pubDate>
      <category>Economics of shipping</category>
      <dc:creator>the operator</dc:creator>
      <description>A retainer and a sprint solve different client problems. Mixing them up costs both sides. The decision tree we use at the top of every sales conversation, with three concrete conversions we made last year.</description>
    </item>

    <item>
      <title>The pentest report is a literary form, and most of them are bad</title>
      <link>https://nexcur.ai/blog/pentest-report-literary-form.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/pentest-report-literary-form.html</guid>
      <pubDate>Fri, 03 Apr 2026 00:00:00 +0000</pubDate>
      <category>Security</category>
      <dc:creator>the operator</dc:creator>
      <description>Sixty percent of pentest reports we have inherited are unreadable to the executive and unactionable to the engineer. A taxonomy of failure modes and what Claude specifically helps with.</description>
    </item>

    <item>
      <title>We measured share of answer for 400 queries across 4 LLMs</title>
      <link>https://nexcur.ai/blog/share-of-answer-400-queries.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/share-of-answer-400-queries.html</guid>
      <pubDate>Sat, 28 Mar 2026 00:00:00 +0000</pubDate>
      <category>SEO and GEO</category>
      <dc:creator>the operator</dc:creator>
      <description>A structural audit of what actually gets cited in Claude, ChatGPT, Perplexity, and Gemini answers. Data, not theory. Includes the raw export.</description>
    </item>

    <item>
      <title>Billing hourly in 2026 is a moral hazard</title>
      <link>https://nexcur.ai/blog/hourly-billing-moral-hazard.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/hourly-billing-moral-hazard.html</guid>
      <pubDate>Sun, 22 Mar 2026 00:00:00 +0000</pubDate>
      <category>Economics of shipping</category>
      <dc:creator>the operator</dc:creator>
      <description>When an operator plus Claude can do a week of classical consulting work in a day, the hourly rate incentivizes slower delivery. How we price fixed-scope and what it took to make that work.</description>
    </item>

    <item>
      <title>Extended thinking is a research budget, not a default</title>
      <link>https://nexcur.ai/blog/extended-thinking-research-budget.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/extended-thinking-research-budget.html</guid>
      <pubDate>Sat, 14 Mar 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>Extended thinking is expensive compute pretending to be a setting. Treat it like any research budget: spend it where the return is visible, leave it off everywhere else.</description>
    </item>

    <item>
      <title>The IAM attack path every Series A has</title>
      <link>https://nexcur.ai/blog/iam-attack-path-series-a.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/iam-attack-path-series-a.html</guid>
      <pubDate>Sat, 07 Mar 2026 00:00:00 +0000</pubDate>
      <category>Security</category>
      <dc:creator>the operator</dc:creator>
      <description>Every Series A we have pentested in the last year shares the same root-to-production path. It is not subtle. It exists because nobody owns it. The shape, the three links in the chain, and how we close each one.</description>
    </item>

    <item>
      <title>Citations are the new backlinks</title>
      <link>https://nexcur.ai/blog/citations-new-backlinks.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/citations-new-backlinks.html</guid>
      <pubDate>Sat, 28 Feb 2026 00:00:00 +0000</pubDate>
      <category>SEO and GEO</category>
      <dc:creator>the operator</dc:creator>
      <description>For fifteen years the link graph was the currency of SEO. In the answer-engine era, the currency is the citation. The economics are not the same.</description>
    </item>

    <item>
      <title>The consulting firm P&amp;L that Claude rewrites</title>
      <link>https://nexcur.ai/blog/consulting-firm-pnl.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/consulting-firm-pnl.html</guid>
      <pubDate>Sat, 21 Feb 2026 00:00:00 +0000</pubDate>
      <category>Economics of shipping</category>
      <dc:creator>the operator</dc:creator>
      <description>A traditional firm clears 25 to 40 percent operating margin at senior-heavy utilization. An AI-native firm does not just do the same work with fewer people. Every line of the P&amp;L shifts, and the strategic question shifts with it.</description>
    </item>

    <item>
      <title>The templates that survive a model upgrade</title>
      <link>https://nexcur.ai/blog/templates-survive-model-upgrade.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/templates-survive-model-upgrade.html</guid>
      <pubDate>Sat, 14 Feb 2026 00:00:00 +0000</pubDate>
      <category>Applied AI</category>
      <dc:creator>the operator</dc:creator>
      <description>We have carried the same template library across six major model releases. About 80 percent of the templates survived every upgrade untouched. The 20 percent that broke all broke for the same two reasons.</description>
    </item>

    <item>
      <title>Claude is not a pentester. A pentester with Claude is.</title>
      <link>https://nexcur.ai/blog/claude-is-not-a-pentester.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/claude-is-not-a-pentester.html</guid>
      <pubDate>Sat, 07 Feb 2026 00:00:00 +0000</pubDate>
      <category>Security</category>
      <dc:creator>the operator</dc:creator>
      <description>We use Claude on every engagement. We also refuse to ship a finding Claude generated without a human operator validating it. Where each side of that line is, and why vendors who hide the line are selling marketing copy as security.</description>
    </item>

    <item>
      <title>Why we stopped chasing featured snippets</title>
      <link>https://nexcur.ai/blog/stopped-chasing-featured-snippets.html</link>
      <guid isPermaLink="true">https://nexcur.ai/blog/stopped-chasing-featured-snippets.html</guid>
      <pubDate>Sat, 31 Jan 2026 00:00:00 +0000</pubDate>
      <category>SEO and GEO</category>
      <dc:creator>the operator</dc:creator>
      <description>Position zero was the optimization target for a decade. For most categories we work in, chasing it is now a waste of budget. What the data said and where we moved the same effort.</description>
    </item>

  </channel>
</rss>
