<?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 Praveen Sridhar on Medium]]></title>
        <description><![CDATA[Stories by Praveen Sridhar on Medium]]></description>
        <link>https://medium.com/@psbots?source=rss-820c969463aa------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*s1k90BoXaTr_f_K5.jpeg</url>
            <title>Stories by Praveen Sridhar on Medium</title>
            <link>https://medium.com/@psbots?source=rss-820c969463aa------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 23:32:51 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@psbots/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[The Arrival of LLM Agents]]></title>
            <link>https://medium.com/@psbots/the-arrival-of-llm-agents-85064ffc01db?source=rss-820c969463aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/85064ffc01db</guid>
            <dc:creator><![CDATA[Praveen Sridhar]]></dc:creator>
            <pubDate>Fri, 13 Dec 2024 05:43:46 GMT</pubDate>
            <atom:updated>2024-12-13T05:43:46.817Z</atom:updated>
            <content:encoded><![CDATA[<p>Agents are a natural progression of what is possible with Large Language Model(LLM) capabilities. What was thought of as another hyped term towards the end of last year, is now used almost daily by those who understand its potential - be it programmers using Cursor and Replit Agent or content creators using agentic workflows to research topics and generate the right content.</p><p><a href="https://llmagents-learning.org/f24">Berkeley’s LLM Agents course</a> has been extremely helpful to consolidate everything happening in the space with its amazing lineup of speakers from the industry - OpenAI, Deepmind, Anthropic, Meta, NVIDIA as well as academia.</p><p>It was very interesting to see the multitude of ways Agents can be defined, and the common pattern that emerges, sharing a few in the screenshots below :</p><ol><li>Interaction with the environment is a key component.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rP3IgEGTpiAW-Wlcs46wMA.png" /><figcaption><a href="https://rdi.berkeley.edu/llm-agents-mooc/slides/intro.pdf">https://rdi.berkeley.edu/llm-agents-mooc/slides/intro.pdf</a></figcaption></figure><p>2. Memory is another essential piece — both short-term as well as long-term</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u8XJDHSvHPWlW98ZVAsiNw.png" /><figcaption><a href="https://rdi.berkeley.edu/llm-agents-mooc/slides/llm_agent_history.pdf">https://rdi.berkeley.edu/llm-agents-mooc/slides/llm_agent_history.pdf</a></figcaption></figure><p>3. Conversation is an interesting pattern, where multiple agents critique and refine each other’s outputs. Autogen framework is almost entirely based off of this pattern</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2dTHAsjplCPaQTcspAVcrw.png" /><figcaption><a href="https://llmagents-learning.org/slides/autogen.pdf">https://llmagents-learning.org/slides/autogen.pdf</a></figcaption></figure><p>4. Specialised agents bring specific challenges</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CpTtsEDpx0J16GN0uL0eew.png" /><figcaption><a href="https://rdi.berkeley.edu/llm-agents-mooc/slides/neubig24softwareagents.pdf">https://rdi.berkeley.edu/llm-agents-mooc/slides/neubig24softwareagents.pdf</a></figcaption></figure><p>5. Tools and frameworks approach Agentic workflow development in different ways — Chains, Graphs, Tapes</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YeBb_kc2dd_aImWvrPnf5g.png" /><figcaption><a href="https://rdi.berkeley.edu/llm-agents-mooc/slides/agentworkflows.pdf">https://rdi.berkeley.edu/llm-agents-mooc/slides/agentworkflows.pdf</a></figcaption></figure><p>There is a lot going on in the LLM Agents space, a new framework is released almost every other day. We are yet to understand the right abstractions and tools to build agents — which makes it a very exciting area to work in!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=85064ffc01db" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A browser based code interpreter for Claude AI]]></title>
            <link>https://medium.com/@psbots/a-browser-based-code-interpreter-for-claude-ai-0d56f607789f?source=rss-820c969463aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/0d56f607789f</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[python]]></category>
            <dc:creator><![CDATA[Praveen Sridhar]]></dc:creator>
            <pubDate>Wed, 24 Jul 2024 08:55:47 GMT</pubDate>
            <atom:updated>2024-07-24T09:03:03.469Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vxBd-TFsWFDBM1qBih2fAQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CZpsWyz8GgfevKUSgbmgbg.png" /><figcaption>“Claude does not have the ability to run the code it generates yet.”</figcaption></figure><p>For the <a href="https://docs.anthropic.com/en/build-with-claude-contest/overview">Build with Claude contest</a> hosted by Anthropic, I thought of making a feature that Claude’s web version doesn’t currently have.</p><blockquote>“Claude does not have the ability to run the code it generates yet.”</blockquote><p>Can we do something about it?</p><p>The most straightforward way would be to make a terminal based system which can run the generated code, similar to the likes of <a href="https://github.com/OpenInterpreter/open-interpreter">OpenInterpreter</a><br>Another way would be to make a web app with a backend that can run the generated code and show the outputs. <a href="https://e2b.dev/">E2B</a> is a great tool to run code in sandboxed environments</p><p>However what I had in mind was a more esoteric approach — run the generated code within the browser. ChatGPT’s code interpreter mainly works with Python, so we need to find a way to make Python run on the browsers — which is exactly what PyScript is meant for.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*bkndIapzPIoe-WPc-iMYfw.jpeg" /></figure><blockquote><a href="https://pyscript.net/">PyScript</a> is an open source platform for Python in the browser.</blockquote><p>It enables you to run Python within the browser without any special installation and by default supports a lot of useful libraries like numpy, matplotlib, pandas etc.</p><p><a href="https://github.com/psbots/clauADA">ClauADA</a> or Claude ADA (Advanced Data Analysis) is a chrome extension that enables Code-Interpreter/Advanced Data Analysis mode for Claude by executing Python code within the browser. It can help make quick charts, do quick data analysis right in the browser without the need for installing Python and analysis libraries.</p><p>It generates code for the given query using Claude APIs and uses PyScript, Pyodide and WASM to run full-fledged Python code including data analysis libraries like Pandas, Matplotlib, numpy, nltk etc.</p><p>The generated code is shown in an editable interface, so you can make any changes if required before running the code.</p><p>Give it a spin here — <a href="https://github.com/psbots/clauADA">https://github.com/psbots/clauADA</a> .</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0d56f607789f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>