<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Namish Saxena on Medium]]></title>
        <description><![CDATA[Stories by Namish Saxena on Medium]]></description>
        <link>https://medium.com/@a_perspective?source=rss-a4c9cd2d64c3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*YtAIyuoORoMLfS8E2Ig9QA.jpeg</url>
            <title>Stories by Namish Saxena on Medium</title>
            <link>https://medium.com/@a_perspective?source=rss-a4c9cd2d64c3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 05:41:54 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@a_perspective/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[MCP Has a Documentation Problem. Here’s the Fix.]]></title>
            <link>https://medium.com/@a_perspective/mcp-has-a-documentation-problem-heres-the-fix-2a4def2cebfa?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a4def2cebfa</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[mcp-server]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Wed, 01 Apr 2026 14:58:43 GMT</pubDate>
            <atom:updated>2026-04-01T14:58:43.938Z</atom:updated>
            <content:encoded><![CDATA[<h4>How one line of code gives your MCP server interactive docs, a portable spec file, and a path toward enterprise-grade tooling</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DzM5TITEthmXgqhywGwGJg.png" /></figure><p>97 million monthly SDK downloads. Over 10,000 public servers. Adopted by Anthropic, OpenAI, Google, and Microsoft. MCP is no longer an experiment — it’s becoming infrastructure.</p><p>And yet, if you want to understand what a given MCP server actually exposes, you’re reading READMEs, parsing source code, or calling tools/list yourself and staring at raw JSON.</p><p>There’s no standard. No interactive docs page you can hand to a teammate. No portable spec file you can commit to a repo or hand to a security team. Every server describes itself differently, or not at all.</p><p>REST APIs had this exact problem before OpenAPI. It’s worth thinking about how that got solved — because the same solution is available for MCP right now.</p><h4>The OpenAPI lesson</h4><p>Before OpenAPI, REST APIs worked fine. Developers built them, documented them in READMEs, tested them with curl, and integrated them by reading source code. Millions of APIs were built that way. Nobody was in crisis.</p><p>But it didn’t scale.</p><p>OpenAPI didn’t fix a broken system. It gave a working system a standard that unlocked composability. Once you had a machine-readable spec describing an API, you could auto-generate clients, build interactive docs, run automated contract tests, and discover APIs through searchable registries. None of that required changes to the underlying APIs. It just required a standardized way to describe them.</p><p>MCP is in the same place today, except the pace is 100x faster. REST took a decade to reach the point where OpenAPI became essential. MCP hit 10,000 servers in under two years.</p><p>The protocol is well-designed. tools/list does its job. But at this pace, the need for a standardized, portable spec format is urgent — not a nice-to-have.</p><h3>What MCPSpec does</h3><p><a href="https://mcpspec.dev/">MCPSpec</a> adds interactive docs and a machine-readable spec to any MCP server with one line of code.</p><p><strong>TypeScript:</strong></p><pre>import { McpServer } from &quot;@modelcontextprotocol/sdk/server/mcp.js&quot;;<br>import { mcpspec } from &quot;@mcpspec-dev/typescript&quot;;<br><br>const server = new McpServer({ name: &quot;my-server&quot;, version: &quot;1.0.0&quot; });<br>// Your existing tools, resources, prompts - untouched<br>const app = mcpspec(server, {<br>  info: { title: &quot;My MCP Server&quot;, version: &quot;1.0.0&quot; },<br>});<br>app.listen(3000);<br>// Docs: http://localhost:3000/docs<br>// Spec: http://localhost:3000/mcpspec.yaml</pre><p><strong>Python:</strong></p><pre>from mcp.server.fastmcp import FastMCP<br>from mcpspec_dev import McpSpec<br><br>mcp = FastMCP(&quot;my-server&quot;)<br>spec = McpSpec(mcp, info={&quot;title&quot;: &quot;My Server&quot;, &quot;version&quot;: &quot;1.0.0&quot;})<br>mcp.run(transport=&quot;streamable-http&quot;)<br># Docs: http://localhost:8000/docs</pre><p>Your server now has three new endpoints: /docs for interactive human-readable documentation, /mcpspec.yaml for a machine-readable spec in a standardized format, and /mcp which proxies your original MCP endpoint so existing clients keep working with zero changes.</p><p>MCPSpec introspects your server at the protocol level — connecting via in-memory transport and calling tools/list, resources/list, and prompts/list. It works with any MCP server regardless of SDK or implementation. And it never executes tools or reads resource content, so the docs endpoint stays safe to expose publicly even when your MCP endpoint requires auth.</p><h4>What the spec file looks like</h4><p>The /mcpspec.yaml output follows a standardized format with JSON Schema validation:</p><pre>mcpspec: 0.1.0<br>$schema: &quot;https://mcpspec.dev/schema/0.1.0.json&quot;<br>info:<br>  name: my-server<br>  version: &quot;1.0.0&quot;<br>  title: My MCP Server<br>tools:<br>  - name: create_task<br>    description: Create a new task<br>    inputSchema:<br>      type: object<br>      properties:<br>        title:<br>          type: string<br>        priority:<br>          type: string<br>          enum: [low, medium, high]<br>      required: [title]<br>resources: []<br>prompts: []</pre><p>Commit this to your repo. Now any agent, developer, or tool can understand your server’s surface area without running it.</p><h3>What this unlocks</h3><p>A standardized spec format is a foundation, not a feature. Here’s what becomes possible once it exists.</p><p><strong>Agent-readable discovery.</strong> Today, an AI agent can’t evaluate an MCP server’s capabilities without connecting, authenticating, and calling tools/list at runtime. With a public mcpspec.yaml, agents and developers can see exactly what a server offers before going through auth flows. Registries become searchable. Developers browse docs before installing anything.</p><p><strong>Contract testing.</strong> Schema drift detection and breaking change alerts. When a server removes a tool or changes a parameter type, automated tests catch it before your agents fail in production. This is table stakes for any protocol operating at enterprise scale.</p><p><strong>Client code generation.</strong> If you know the exact tools a server exposes, their parameters, and return types, you can generate type-safe client code. Same pattern OpenAPI code generators enabled for REST.</p><p><strong>Enterprise governance.</strong> Security teams can review a mcpspec.yaml to understand exactly what capabilities an MCP server provides — what data it can access, what actions it can take — without a code review or running the server in a sandbox.</p><p>Every protocol that achieves enterprise adoption develops a spec-and-tooling layer. REST got OpenAPI, which spawned thousands of tools. GraphQL got schema introspection, which powered an entire developer experience ecosystem. MCP is entering its enterprise phase. The tooling layer needs to grow with it.</p><p>MCPSpec is MIT licensed. TypeScript on <a href="https://www.npmjs.com/package/@mcpspec-dev/typescript">npm</a>, Python on <a href="https://pypi.org/project/mcpspec-dev/">PyPI</a>, source on <a href="https://github.com/namishsaxena/mcpspec">GitHub</a>.</p><p>The spec format is at v0.1.0 — early enough that real usage shapes what it becomes. If you build MCP servers, try it on one. If something’s missing, open an issue.</p><p>A spec format only becomes a standard when people use it and shape it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2a4def2cebfa" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[In the Future, Teams Will Be Generated, Not Hired]]></title>
            <link>https://medium.com/genaius/in-the-future-teams-will-be-generated-not-hired-65948c8d1daf?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/65948c8d1daf</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Thu, 12 Mar 2026 11:22:35 GMT</pubDate>
            <atom:updated>2026-03-12T13:17:27.260Z</atom:updated>
            <content:encoded><![CDATA[<p>When you need 50 customer service reps for holiday season, you don’t hire them full-time. You scale up, then scale down. The staffing model has worked this way for decades.</p><p>Now imagine doing that with entire engineering teams. In minutes, not months.</p><p>This isn’t science fiction. It’s already happening.</p><p>At Anthropic, roughly 90% of Claude Code is written by Claude Code itself. Amazon Q Developer modernized thousands of legacy Java applications autonomously. Nubank achieved 12x efficiency improvement on multi-million line migrations using AI agents. Genentech is automating drug discovery workflows with agent ecosystems. These aren’t experiments. They’re production systems.</p><h4>The Convergence That Made This Possible</h4><p>Three technologies matured at the same time. Any two produce something interesting. All three together produce a paradigm shift.</p><ul><li>Frontier Models (Claude, GPT, Gemini) provide reasoning and task decomposition. They’re the intellectual engine.</li><li>Agent Harnesses (Claude Code, Devin, CrewAI, LangGraph) provide orchestration. Multi-step execution, tool use, inter-agent communication. They turn models from chatbots into autonomous workers.</li><li>Context Management (long context windows, RAG, persistent memory) provides continuity. Without it, agents forget what they learned ten minutes ago.</li></ul><p>Here’s the key insight:</p><ul><li>LLMs without agent harnesses are chatbots. Impressive, but request-response.</li><li>Agent harnesses without good models produce garbage.</li><li>Automation of incompetence. Models without context management have amnesia. Brilliant but forgetful.</li></ul><p>The convergence is what matters. Not any single component. The market agrees. Multi-agent AI is projected to grow from $7.84 billion in 2025 to $52.62 billion by 2030. Multi-agent system inquiries surged 1,445% in a single year.</p><h4>From Fixed Teams to Generated Teams</h4><p>Think about how team structures are evolving:</p><p><strong>Stage 1: Fixed Human Teams</strong> — Roles predefined. Capacity locked. A team of 8 is a team of 8, whether the sprint needs 3 or 15.</p><p><strong>Stage 2: AI-Assisted Teams (where most orgs are today)</strong> — Copilots make individuals faster. The org chart doesn’t change.</p><p><strong>Stage 3: Hybrid Adaptive Teams (emerging now)</strong> — AI agents join the team as needed. Some workflows are fully agent-handled. Humans assign, review, redirect.</p><p><strong>Stage 4: Agentic Teams with Human Oversight (2026–2028) </strong>— One person orchestrates a full team of specialized agents. Humans govern. Agents execute.</p><p>The effort shift is dramatic. Traditional engineering: 60% manual code comprehension, 30% boilerplate, 10% architecture. Agentic model: 85% architecture and strategy, 10% boilerplate, 5% manual code reading.</p><blockquote>The principle: Autonomous Execution, Human Governance.</blockquote><h4>Teams as Elastic Infrastructure</h4><p>Before the power grid, every factory ran its own generator. Before cloud, every company ran its own servers. We’re at the same inflection point with teams.</p><p>The parallels are exact:</p><ul><li>On-demand provisioning: Need a security review? Generate a specialist agent. Minutes, not hiring cycles.</li><li>Auto-scaling: Project grows? Add agents. Scope shrinks? Scale down. No severance packages.</li><li>Pay-per-use: Agents work only when needed. No bench time, no idle salaries.</li><li>Configuration as code: Team compositions version-controlled and reproducible.</li></ul><p>The economics are already compelling. AI agents cut labor costs up to 45%. Efficiency improves 60–70%. Companies report 171% average ROI. And LLM API prices dropped 80–90% since early 2024.</p><h4>What Becomes Possible</h4><p>Not by adding more people. By changing how teams are formed.</p><ul><li>Autonomous legacy system migration at scale</li><li>End-to-end greenfield delivery with minimal human intervention</li><li>On-demand AI specialists embedded in teams only when needed</li><li>Agentic governance councils running continuous compliance and security</li><li>Long-running autonomous research spanning days or weeks</li><li>Agent-led marketing teams assembled for campaigns, dissolved after launch</li><li>Real-time supply chain optimization across procurement, logistics, manufacturing</li></ul><p>57% of organizations already deploy agents for multi-stage workflows. This isn’t future speculation. It’s current adoption.</p><h4>The Human Role Elevates</h4><p>This isn’t about replacing people, yet. It’s about elevating human work to where it creates the most value.</p><p>40% of job roles in Global 2000 companies will directly involve working with AI agents by 2026. The skill set shifts from execution to orchestration, from code review to plan governance, from individual contribution to strategic direction.</p><p>Singapore launched the world’s first AI Governance Framework for Agentic AI in January 2026. Governance is becoming a first-class organizational capability.</p><h3>The Bottom Line</h3><p>If you’re building infrastructure: the gap between what agents need and what cloud provides is the biggest opportunity in enterprise tech.</p><p>If you’re leading teams: start experimenting with hybrid workflows now. Understand orchestration before you’re forced to.</p><p>If you’re setting strategy: organizations that master team generation will have a structural advantage that hiring optimization cannot match.</p><p>We spent a century optimizing how we hire, onboard, and manage human teams. The next decade will be about mastering how we generate, orchestrate, and govern AI ones.</p><p>The playbook won’t be the same. And that’s the point.</p><p>Read the full deep-dive in my blog, link below.</p><p><a href="https://www.namishsaxena.com/blog/teams-generated-not-hired">In the Future, Teams Will Be Generated, Not Hired</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=65948c8d1daf" width="1" height="1" alt=""><hr><p><a href="https://medium.com/genaius/in-the-future-teams-will-be-generated-not-hired-65948c8d1daf">In the Future, Teams Will Be Generated, Not Hired</a> was originally published in <a href="https://medium.com/genaius">GenAIUs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your AI Agent Is Smart. It Just Doesn’t Remember Enough.]]></title>
            <link>https://medium.com/genaius/your-ai-agent-is-smart-it-just-doesnt-remember-enough-c818d2364d3e?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/c818d2364d3e</guid>
            <category><![CDATA[developer]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Fri, 16 Jan 2026 10:46:11 GMT</pubDate>
            <atom:updated>2026-01-16T10:46:11.212Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why agentic coding keeps losing hard-won knowledge, and how Lore helps teams build continuity instead of repetition.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6aZaPigBdPBD8D4aJnA37Q.png" /></figure><h3>The early promise of agentic coding</h3><p>Agentic coding is intoxicating.</p><p>You ask for a refactor, and the agent reasons about structure instead of just editing lines. You describe an outcome loosely, and something usable appears in minutes. For a while, it feels like you are finally collaborating with the system rather than supervising it.</p><p>Then a pattern quietly sets in.</p><p>You explain a constraint.<br>You correct a decision.<br>You clarify why something that looks wrong is actually intentional.</p><p>The agent responds well. You move on.</p><p>Later, often sooner than expected, the same explanation has to be given again.</p><p>Nothing is obviously broken. Yet progress starts to feel circular.</p><h3>This is not really a context window problem</h3><p>This failure mode is usually explained away as a context issue. The model did not see enough. The prompt was not structured well. The documentation was incomplete.</p><p>Those explanations are not wrong, but they miss the core issue.</p><p>What breaks down in agentic coding is not access to information. It is <strong>continuity of knowledge</strong>.</p><p>Real codebases are shaped by accumulated decisions:</p><ul><li>Trade-offs made under pressure</li><li>Constraints imposed by production incidents</li><li>Conventions that exist for reasons no longer visible in the code</li></ul><p>Humans absorb this over time. Agents do not. Each run effectively starts clean unless you reconstruct that history for them.</p><p>That reconstruction does not scale.</p><h3>Why documentation and prompts only go so far</h3><p>Documentation is good at describing what a system does. It is much worse at preserving why it ended up this way.</p><p>The “why” lives in places like:</p><ul><li>“We tried the simpler version. It caused data corruption.”</li><li>“This service looks redundant but exists for regulatory reasons.”</li><li>“Do not refactor this path without checking X.”</li></ul><p>Teams respond by adding more context:</p><ul><li>Growing CLAUDE.md files</li><li>Entire folders of AI instructions</li><li>Repeated prompt rituals that restate the same caveats</li></ul><p>Eventually, signal gets buried in noise. Agents either miss the important parts or treat everything as equally important.</p><p>Neither leads to good outcomes.</p><h3>What agentic systems actually lack</h3><p>Seen clearly, the gap is simple.</p><p>Agents have working memory.<br>They have short-term context.<br>They can retrieve documents.</p><p>What they lack is a way to retain <strong>selective, decision-level knowledge</strong> across time.</p><p>Without that, teams are forced into a bad trade-off:</p><ul><li>Either overload the agent with everything</li><li>Or let it forget things that genuinely matter</li></ul><h3>Where Lore fits</h3><p>Lore is built to address this specific failure mode.</p><p>It is not a documentation system.<br>It is not a prompt manager.<br>It is not an agent framework.</p><p>Lore treats project knowledge as institutional memory rather than reference material.</p><p>You store small, explicit entries such as:</p><ul><li>Architectural decisions</li><li>Conventions</li><li>Warnings and constraints</li><li>Workflow rules</li></ul><p>Each entry can be scoped to files or areas of the codebase. When a developer or agent works on a task, Lore surfaces only the knowledge that applies.</p><p>It has been tested with Claude Code, but it is intentionally <strong>not locked into any single agentic CLI</strong>. The knowledge lives with your repo, not with a tool, and survives model changes, agent changes, and workflow rewrites.</p><p>That portability is the point.</p><h3>A simple example</h3><p>Getting started with Lore is deliberately lightweight.</p><p>You initialize it once:</p><pre>lore init</pre><p>Then you capture knowledge at the moment it forms, instead of letting it vanish into Slack or memory:</p><pre>lore add \<br>  --type decision \<br>  --files src/storage/* \<br>  --content &quot;Dual storage is intentional. Simplifying this breaks offline guarantees.&quot;</pre><p>Later, when an agent or developer works on that area:</p><pre>$ lore query --task &quot;simplify the storage layer&quot; --limit 3</pre><p>Lore returns something like:</p><pre>[15%] ADR-001: Dual Storage (SQLite + JSONL)<br>      architecture | 197 tokens<br>      Status: Accepted<br>      Context: We need persistent storage that survives restarts<br>      while remaining inspectable and portable.<br>[12%] Dual Storage Strategy<br>      reference | 50 tokens<br>      SQLite is the working database. JSONL is the append-only log.<br>[12%] ADR-010: Local-First Architecture<br>      architecture | 176 tokens<br>      Status: Accepted<br>      Context: Data must live locally first, sync second.</pre><p>Before the agent attempts to &quot;<em>simplify&quot;</em> anything, it is warned that the complexity is intentional.</p><p>That moment is the value.</p><h3>A concrete before and after</h3><p>Before Lore:</p><ul><li>A 140–150 line CLAUDE.md trying to cover every edge case</li><li>Instructions pasted repeatedly into prompts</li><li>Agents missing the one paragraph that actually mattered</li></ul><p>After Lore:</p><ul><li>Three scoped entries tied to the storage layer</li><li>Retrieved only when storage is being touched</li><li>No prompt bloat, no repetition</li></ul><p>The difference is not dramatic in one run. It compounds over time.</p><h3>Why this matters as agents become more autonomous</h3><p>As agents improve, they act with more confidence and make broader changes. Without continuity, that confidence becomes risky.</p><p>Systems rarely fail loudly. They drift.</p><p>If agentic coding is going to move beyond experimentation into long-lived systems, knowledge has to accumulate instead of resetting.</p><p>Lore is a small, deliberate step in that direction.</p><h3>If this resonates</h3><p>Lore is open source and early.</p><p>If you are experimenting with agentic coding, using Claude Code, or simply tired of repeating the same explanations to increasingly capable tools, try it out:</p><p><a href="https://github.com/namishsaxena/lore?utm_source=chatgpt.com">https://github.com/namishsaxena/lore</a></p><p>If it helps:</p><ul><li>Star the repo</li><li>Share it with someone running into the same friction</li><li>Open issues with feedback or ideas</li></ul><p>Agentic systems do not just need better reasoning. They need a past they can actually learn from.</p><p>If you are intrigued by this and are keen to enhance productivity, efficiency, and offer exceptional customer and employee experiences, let’s start a conversation. The future of business is here, and it’s definitely powered by Generative AI.</p><p>Let me know in comments, what is your take on this.</p><p><em>Follow </em><a href="https://medium.com/genaius"><em>GenAIUs</em></a><em> to always get the latest insights and opinions before anyone else.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c818d2364d3e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/genaius/your-ai-agent-is-smart-it-just-doesnt-remember-enough-c818d2364d3e">Your AI Agent Is Smart. It Just Doesn’t Remember Enough.</a> was originally published in <a href="https://medium.com/genaius">GenAIUs</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[DeepSeek’s mHC in Plain English: Stability by Design, Not by Hope]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/deepseeks-mhc-in-plain-english-stability-by-design-not-by-hope-70a4342646f5?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/934/1*nrJHlMDd5xnGFjARXJwrpw.png" width="934"></a></p><p class="medium-feed-snippet">A simple constraint. A deeper network. Fewer ugly surprises.</p><p class="medium-feed-link"><a href="https://medium.com/genaius/deepseeks-mhc-in-plain-english-stability-by-design-not-by-hope-70a4342646f5?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/deepseeks-mhc-in-plain-english-stability-by-design-not-by-hope-70a4342646f5?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/70a4342646f5</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[neural-networks]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Wed, 07 Jan 2026 06:02:26 GMT</pubDate>
            <atom:updated>2026-01-07T06:02:26.552Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The Post-SaaS Manifesto: Why Your Dashboard is a Liability in 2026]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/the-post-saas-manifesto-why-your-dashboard-is-a-liability-in-2026-7f8e9b5a5cce?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/1545/1*PtFjX8K7pjKViHGllxD1Sg.png" width="1545"></a></p><p class="medium-feed-snippet">The era of the &#x201C;all-in-one&#x201D; platform is ending. How AI agents and the Model Context Protocol (MCP) are killing traditional interfaces.</p><p class="medium-feed-link"><a href="https://medium.com/genaius/the-post-saas-manifesto-why-your-dashboard-is-a-liability-in-2026-7f8e9b5a5cce?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/the-post-saas-manifesto-why-your-dashboard-is-a-liability-in-2026-7f8e9b5a5cce?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/7f8e9b5a5cce</guid>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[genai]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Mon, 05 Jan 2026 07:02:31 GMT</pubDate>
            <atom:updated>2026-01-05T07:02:31.800Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Amazon vs. The Trespassers: Why Perplexity’s Masked Bots Deserve to be Banned]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/amazon-vs-the-trespassers-why-perplexitys-masked-bots-deserve-to-be-banned-1a786a03f224?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/1024/1*f3QEltJ8zfyc3dtN1dN0Sg.png" width="1024"></a></p><p class="medium-feed-snippet">Perplexity&#x2019;s lack of a moat and unethical technical tactics make it a fragile player in the coming &#x201C;Agentic Web.&#x201D;</p><p class="medium-feed-link"><a href="https://medium.com/genaius/amazon-vs-the-trespassers-why-perplexitys-masked-bots-deserve-to-be-banned-1a786a03f224?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/amazon-vs-the-trespassers-why-perplexitys-masked-bots-deserve-to-be-banned-1a786a03f224?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/1a786a03f224</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[amazon]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Fri, 02 Jan 2026 08:45:23 GMT</pubDate>
            <atom:updated>2026-01-02T08:45:23.900Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The 2026 Split: Toy Agents vs Trusted Workflows]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/the-2026-split-toy-agents-vs-trusted-workflows-a2f3784d5c31?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/1376/1*2K44qcxZ-wmZ1uyMLBttXg.png" width="1376"></a></p><p class="medium-feed-snippet">In 2025, agents made output cheap. In 2026, the winners will be the teams who can trust that output enough to let it touch real work.</p><p class="medium-feed-link"><a href="https://medium.com/genaius/the-2026-split-toy-agents-vs-trusted-workflows-a2f3784d5c31?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/the-2026-split-toy-agents-vs-trusted-workflows-a2f3784d5c31?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/a2f3784d5c31</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[genai]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Thu, 18 Dec 2025 06:32:42 GMT</pubDate>
            <atom:updated>2025-12-18T06:32:42.421Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Apple’s CLaRa]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/apples-clara-e12722407df8?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/1536/0*t7LEEKb-RkJWJucT" width="1536"></a></p><p class="medium-feed-snippet">A practical, non-hype walkthrough of Apple&#x2019;s CLaRa: why &#x201C;classic RAG&#x201D; strains at scale, what &#x201C;continuous latent reasoning&#x201D; really means.</p><p class="medium-feed-link"><a href="https://medium.com/genaius/apples-clara-e12722407df8?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/apples-clara-e12722407df8?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/e12722407df8</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[apple]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Wed, 17 Dec 2025 07:03:45 GMT</pubDate>
            <atom:updated>2025-12-17T07:03:45.135Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[AI Agents Made Output Cheap. Attention Became the Only Currency]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/ai-agents-made-output-cheap-attention-became-the-only-currency-33985fb4b011?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/1376/1*nqa9G50rdzHAclbvDSoQ_g.png" width="1376"></a></p><p class="medium-feed-snippet">AI agents in 2025 did not make you &#x201C;multitask.&#x201D; They made you the scheduler of a parallel day.</p><p class="medium-feed-link"><a href="https://medium.com/genaius/ai-agents-made-output-cheap-attention-became-the-only-currency-33985fb4b011?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/ai-agents-made-output-cheap-attention-became-the-only-currency-33985fb4b011?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/33985fb4b011</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Sun, 14 Dec 2025 10:25:18 GMT</pubDate>
            <atom:updated>2025-12-14T10:25:18.593Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[OpenAI’s GPT-5.2 beats 70% of experts. You are still calling it a chatbot.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/genaius/openais-gpt-5-2-beats-70-of-experts-you-are-still-calling-it-a-chatbot-b3853f77621c?source=rss-a4c9cd2d64c3------2"><img src="https://cdn-images-1.medium.com/max/1146/1*oBsjDNzmdL6I8hBBONv0Kg.png" width="1146"></a></p><p class="medium-feed-snippet">GPT-5.2 is built to be a work brain, not a chat toy. Here is what actually changed for agents, long context, and real jobs.</p><p class="medium-feed-link"><a href="https://medium.com/genaius/openais-gpt-5-2-beats-70-of-experts-you-are-still-calling-it-a-chatbot-b3853f77621c?source=rss-a4c9cd2d64c3------2">Continue reading on GenAIUs »</a></p></div>]]></description>
            <link>https://medium.com/genaius/openais-gpt-5-2-beats-70-of-experts-you-are-still-calling-it-a-chatbot-b3853f77621c?source=rss-a4c9cd2d64c3------2</link>
            <guid isPermaLink="false">https://medium.com/p/b3853f77621c</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[openai]]></category>
            <dc:creator><![CDATA[Namish Saxena]]></dc:creator>
            <pubDate>Fri, 12 Dec 2025 06:05:49 GMT</pubDate>
            <atom:updated>2025-12-12T06:40:59.551Z</atom:updated>
        </item>
    </channel>
</rss>