<?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>DEV Community: Siya Jain</title>
    <description>The latest articles on DEV Community by Siya Jain (@siya_eduonix).</description>
    <link>https://dev.to/siya_eduonix</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4065188%2F03eeb000-f839-4da6-99e6-97dc9710bf7c.jpg</url>
      <title>DEV Community: Siya Jain</title>
      <link>https://dev.to/siya_eduonix</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/siya_eduonix"/>
    <language>en</language>
    <item>
      <title>AI agents: when should a decision use a model, and when should it use code?</title>
      <dc:creator>Siya Jain</dc:creator>
      <pubDate>Tue, 27 Oct 2026 16:38:43 +0000</pubDate>
      <link>https://dev.to/siya_eduonix/ai-agents-when-should-a-decision-use-a-model-and-when-should-it-use-code-19cn</link>
      <guid>https://dev.to/siya_eduonix/ai-agents-when-should-a-decision-use-a-model-and-when-should-it-use-code-19cn</guid>
      <description>&lt;p&gt;A common AI architecture mistake is sending every step through a large language model. But an agent workflow often contains several different kinds of decisions:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Fixed rules:&lt;/strong&gt; Is the user authorized? Is this action within the spending limit? These should usually be enforced in application code.&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Open-ended reasoning:&lt;/strong&gt; What does the user mean? How should several pieces of information be summarized? A general-purpose language model may be useful here.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Bounded choices:&lt;/strong&gt; Which known category fits this request? Which of these approved workflows should handle it? A structured decision model may be worth testing.&lt;/p&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Consider a support-ticket agent. Code can enforce access controls. A language model can understand an unusual request and draft a response. A classifier or decision model can route common tickets to a queue. The model should not be the only safeguard: validate its output, define a fallback for uncertain results, and require human approval for consequential actions.&lt;/p&gt;

&lt;p&gt;This is the technical question behind Jev, a model designed for structured decisions. Eduonix’s Kickstarter campaign proposes a course on using Jev in application engineering, with tools such as Python and LangChain. The outline is an interesting curriculum proposal, but it doesn’t establish how well the finished lessons work. I’d look for sample code, a clear evaluation method, and a comparison against simple rules.&lt;/p&gt;

&lt;p&gt;For a different learning angle, a &lt;a href="https://www.udemy.com/course/build-an-ai-agent-from-scratch-with-langgraph/" rel="noopener noreferrer"&gt;Udemy course outline for building an agent with LangGraph&lt;/a&gt; describes constructing an agent loop in Python, then rebuilding it with a framework. Its listed topics include tool calling, memory, evaluation, and guardrails. That makes it a useful syllabus to compare against a course focused on structured decisions; it isn’t a like-for-like course comparison or a claim about either course’s teaching quality.&lt;/p&gt;

&lt;p&gt;A practical way to evaluate a specialized decision model is to:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Define a narrow task, such as routing support tickets.&lt;/li&gt;
&lt;li&gt;Create representative examples, including ambiguous cases.&lt;/li&gt;
&lt;li&gt;Compare the model with a simple code-based baseline.&lt;/li&gt;
&lt;li&gt;Measure accuracy, latency, cost, and low-confidence behavior.&lt;/li&gt;
&lt;li&gt;Decide what the application should do when the model is uncertain or wrong.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The &lt;a href="https://www.kickstarter.com/projects/eduonix/complete-jev-application-engineering-masterclass?ref=erv7b3" rel="noopener noreferrer"&gt;Jev campaign outline&lt;/a&gt; is the project description I reviewed—not a course I’ve completed. &lt;/p&gt;

&lt;p&gt;Where do you draw the boundary between model decisions and application code in your own agent systems?&lt;/p&gt;

</description>
      <category>agents</category>
      <category>ai</category>
      <category>architecture</category>
    </item>
    <item>
      <title>Use Google's AI tools without turning every task into an "AI project."</title>
      <dc:creator>Siya Jain</dc:creator>
      <pubDate>Mon, 07 Sep 2026 12:34:12 +0000</pubDate>
      <link>https://dev.to/siya_eduonix/use-googles-ai-tools-without-turning-every-task-into-an-ai-project-3cdn</link>
      <guid>https://dev.to/siya_eduonix/use-googles-ai-tools-without-turning-every-task-into-an-ai-project-3cdn</guid>
      <description>&lt;p&gt;I've been looking at how small teams can use Google's AI tools without turning every task into an "AI project."&lt;/p&gt;

&lt;p&gt;A few tools caught my attention:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;NotebookLM&lt;/strong&gt; — useful when you're dealing with a lot of internal documents or research.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Google Vids&lt;/strong&gt; — potentially useful when a small team needs to turn an idea, process or explanation into video content without building a complicated production workflow.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;MusicFX&lt;/strong&gt; — more of a creative experimentation tool, but interesting for anyone working with content.&lt;/p&gt;

&lt;p&gt;The bigger lesson for me is that AI adoption doesn't necessarily mean adding AI everywhere.&lt;/p&gt;

&lt;p&gt;A better approach is:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Find repetitive/expensive task → test one AI tool → measure the result → keep it only if it actually improves the workflow.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;I'm currently learning these tools through a structured &lt;a href="https://tinyurl.com/4455avby" rel="noopener noreferrer"&gt;Google AI resource&lt;/a&gt; covering NotebookLM, Vids, MusicFX and others.&lt;/p&gt;

&lt;p&gt;If anyone here has experimented with these tools, I'd be interested to hear which ones actually saved you time rather than just being interesting to try.&lt;/p&gt;

</description>
      <category>productivity</category>
    </item>
    <item>
      <title>Don't Use the Same AI Tool for Every Problem</title>
      <dc:creator>Siya Jain</dc:creator>
      <pubDate>Mon, 07 Sep 2026 12:31:25 +0000</pubDate>
      <link>https://dev.to/siya_eduonix/dont-use-the-same-ai-tool-for-every-problem-2k9g</link>
      <guid>https://dev.to/siya_eduonix/dont-use-the-same-ai-tool-for-every-problem-2k9g</guid>
      <description>&lt;p&gt;One thing I've noticed while exploring AI tools is that &lt;strong&gt;tool selection matters almost as much as prompting&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;A chatbot might be great for brainstorming, but it isn't necessarily the best choice when you're working with a large collection of source documents, creating a presentation, or experimenting with audio.&lt;/p&gt;

&lt;p&gt;Here is a simple way to think about Google's AI ecosystem:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Task&lt;/th&gt;
&lt;th&gt;Tool worth exploring&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Working with research/document sources&lt;/td&gt;
&lt;td&gt;NotebookLM&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Creating structured video content&lt;/td&gt;
&lt;td&gt;Google Vids&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Generative music experimentation&lt;/td&gt;
&lt;td&gt;MusicFX&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The interesting part is the workflow rather than the individual products.&lt;/p&gt;

&lt;p&gt;For example, if I were researching a technical topic, I could:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Collect reliable source material.&lt;/li&gt;
&lt;li&gt;Organize it in NotebookLM.&lt;/li&gt;
&lt;li&gt;Ask questions to identify the important concepts.&lt;/li&gt;
&lt;li&gt;Turn the resulting explanation into a structured presentation.&lt;/li&gt;
&lt;li&gt;Experiment with different ways of communicating the same information.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;That is a much more useful AI skill than simply knowing a list of tools.&lt;/p&gt;

&lt;p&gt;For anyone wanting to explore this workflow further, I've been using a &lt;strong&gt;&lt;a href="https://tinyurl.com/4455avby" rel="noopener noreferrer"&gt;Google AI learning resource&lt;/a&gt;&lt;/strong&gt; that covers NotebookLM, Vids, MusicFX and additional Google AI tools.&lt;/p&gt;

</description>
      <category>agents</category>
    </item>
    <item>
      <title>Is AI Actually Saving Time? Try Measuring the Workflow</title>
      <dc:creator>Siya Jain</dc:creator>
      <pubDate>Mon, 07 Sep 2026 12:29:20 +0000</pubDate>
      <link>https://dev.to/siya_eduonix/is-ai-actually-saving-time-try-measuring-the-workflow-6aj</link>
      <guid>https://dev.to/siya_eduonix/is-ai-actually-saving-time-try-measuring-the-workflow-6aj</guid>
      <description>&lt;p&gt;One mistake with AI productivity is measuring the speed of generating an output while ignoring the time spent fixing it.&lt;/p&gt;

&lt;p&gt;A better experiment is to measure the &lt;strong&gt;whole workflow&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;For example, take a research task:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Without AI&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Research → read documents → take notes → organize information → create summary&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;With AI assistance&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Collect sources → analyze with AI → verify important points → organize → create summary&lt;/p&gt;

&lt;p&gt;The second workflow isn't automatically better.&lt;/p&gt;

&lt;p&gt;The real question is:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Did the AI reduce the total amount of useful work required?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;I've been experimenting with Google's AI tools from this perspective.&lt;/p&gt;

&lt;p&gt;NotebookLM is interesting for source-based research, Google Vids for turning information into video, and MusicFX for creative experimentation.&lt;/p&gt;

&lt;p&gt;The tools are different, but the principle is the same: &lt;strong&gt;start with the problem, then choose the tool.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;I'm also working through a structured &lt;strong&gt;&lt;a href="https://tinyurl.com/4455avby" rel="noopener noreferrer"&gt;Google AI learning resource&lt;/a&gt;&lt;/strong&gt; covering these tools. I'm finding the hands-on approach more useful than simply collecting AI tool names.&lt;/p&gt;

</description>
      <category>discuss</category>
    </item>
    <item>
      <title>Don't Use the Same AI Tool for Every Problem</title>
      <dc:creator>Siya Jain</dc:creator>
      <pubDate>Mon, 07 Sep 2026 12:24:54 +0000</pubDate>
      <link>https://dev.to/siya_eduonix/dont-use-the-same-ai-tool-for-every-problem-g5b</link>
      <guid>https://dev.to/siya_eduonix/dont-use-the-same-ai-tool-for-every-problem-g5b</guid>
      <description>&lt;p&gt;One thing I've noticed while exploring AI tools is that tool selection matters almost as much as prompting.&lt;/p&gt;

&lt;p&gt;A chatbot might be great for brainstorming, but it isn't necessarily the best choice when you're working with a large collection of source documents, creating a presentation, or experimenting with audio.&lt;/p&gt;

&lt;p&gt;Here is a simple way to think about Google's AI ecosystem:&lt;/p&gt;

&lt;p&gt;Task    Tool worth exploring&lt;br&gt;
Working with research/document sources  NotebookLM&lt;br&gt;
Creating structured video content   Google Vids&lt;br&gt;
Generative music experimentation    MusicFX&lt;/p&gt;

&lt;p&gt;The interesting part is the workflow rather than the individual products.&lt;/p&gt;

&lt;p&gt;For example, if I were researching a technical topic, I could:&lt;/p&gt;

&lt;p&gt;Collect reliable source material.&lt;br&gt;
Organize it in NotebookLM.&lt;br&gt;
Ask questions to identify the important concepts.&lt;br&gt;
Turn the resulting explanation into a structured presentation.&lt;br&gt;
Experiment with different ways of communicating the same information.&lt;/p&gt;

&lt;p&gt;That is a much more useful AI skill than simply knowing a list of tools.&lt;/p&gt;

&lt;p&gt;For anyone wanting to explore this workflow further, I've been using a &lt;a href="https://tinyurl.com/4455avby" rel="noopener noreferrer"&gt;Google AI learning resource &lt;/a&gt;that covers NotebookLM, Vids, MusicFX and additional Google AI tools.&lt;/p&gt;

</description>
      <category>forums</category>
    </item>
    <item>
      <title>AI Agents: Where the Real Engineering Challenge Begins</title>
      <dc:creator>Siya Jain</dc:creator>
      <pubDate>Wed, 12 Aug 2026 12:44:38 +0000</pubDate>
      <link>https://dev.to/siya_eduonix/ai-agents-where-the-real-engineering-challenge-begins-2l8h</link>
      <guid>https://dev.to/siya_eduonix/ai-agents-where-the-real-engineering-challenge-begins-2l8h</guid>
      <description>&lt;p&gt;AI agents are getting a lot of attention, but building a basic one isn't necessarily the difficult part anymore. With today's LLM APIs and frameworks, developers can create agents that use tools and complete simple tasks relatively quickly.&lt;/p&gt;

&lt;p&gt;The harder question is: &lt;strong&gt;can an agent reliably complete a task without doing something unexpected?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;A normal chatbot usually follows:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;User → Prompt → LLM → Response&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;An agent adds another layer:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;Goal → Reason → Tool → Result → Decide → Repeat&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;For example, instead of simply answering &lt;em&gt;"Why is our API slow?"&lt;/em&gt;, an agent could check monitoring data, retrieve logs, compare recent deployments, and then explain the likely cause.&lt;/p&gt;

&lt;p&gt;While exploring these topics, I recently came across a resource covering generative AI, AI agents, machine learning, and data science. What I found useful was seeing these areas together rather than treating them as completely separate subjects.&lt;/p&gt;

&lt;p&gt;Tools are what make agents particularly useful. An agent might have functions such as &lt;code&gt;get_logs()&lt;/code&gt;,&lt;code&gt;search_database()&lt;/code&gt;, or &lt;code&gt;check_deployment()&lt;/code&gt;. But giving an agent more tools doesn't automatically make it better. Each tool should have a clear purpose, predictable output, and limited permissions.&lt;/p&gt;

&lt;p&gt;This becomes especially important when agents can perform real actions. Reading an order is relatively low risk. Changing customer information may require confirmation. Issuing a large refund or deleting data might require human approval.&lt;/p&gt;

&lt;p&gt;Another interesting question is whether we actually need multiple agents. A research agent, coding agent, reviewer agent, and manager agent can sound impressive, but every additional component adds complexity, latency, and potential failure points.&lt;/p&gt;

&lt;p&gt;Sometimes one well-designed agent with a few reliable tools is enough.&lt;/p&gt;

&lt;p&gt;I think this is where AI agent development is becoming less about clever prompts and more about traditional software engineering: validation, permissions, retries, logging, monitoring, and evaluation.&lt;/p&gt;

&lt;p&gt;The real question isn't &lt;strong&gt;"Can an AI agent do this?"&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;It's &lt;strong&gt;"Should it do this, how do we know it did it correctly, and what happens when it fails?"&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;I'm curious what other developers think: &lt;strong&gt;what's currently the biggest challenge with AI agents — reliability, security, tool use, or evaluation?&lt;/strong&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>machinelearning</category>
      <category>programming</category>
    </item>
  </channel>
</rss>
