<?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 Rahul Bhrambhatt on Medium]]></title>
        <description><![CDATA[Stories by Rahul Bhrambhatt on Medium]]></description>
        <link>https://medium.com/@rahulh123?source=rss-5457b030f0d0------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*nOGH8QNEyLMNPani4CE1KA.jpeg</url>
            <title>Stories by Rahul Bhrambhatt on Medium</title>
            <link>https://medium.com/@rahulh123?source=rss-5457b030f0d0------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:31:58 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@rahulh123/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[Scan Your AI Agent Skills Before They Scan You: A Practical Guide to NVIDIA SkillSpector Repo]]></title>
            <link>https://rahulh123.medium.com/scan-your-ai-agent-skills-before-they-scan-you-a-practical-guide-to-nvidia-skillspector-repo-61ae8676900a?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/61ae8676900a</guid>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Sun, 25 Oct 2026 09:28:34 GMT</pubDate>
            <atom:updated>2026-10-25T09:28:34.404Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RRkwhuvOTIGQebA0Jb47pg.png" /></figure><h3>Do you know what your skills are doing?</h3><p>Quick question: how many AI agent skills or MCP servers are installed on your machine? And how many did you actually read before installing?</p><p>Research behind NVIDIA’s new open-source tool found that in 31,132 analyzed skills, <strong>26.1% had at least one vulnerability</strong> and <strong>5.2% looked malicious</strong>. That is one in four.</p><blockquote><a href="https://github.com/NVIDIA/SkillSpector"><strong>SkillSpector</strong></a> is NVIDIA’s free (Apache 2.0) scanner that answers one question: <em>is this skill safe to install?</em> It never runs the skill. It reads the files and gives you a score from 0 to 100 with a clear verdict: <strong>SAFE</strong>, <strong>CAUTION</strong> or <strong>DO NOT INSTALL</strong>.</blockquote><p>It looks for things like:</p><ul><li>Hidden prompt injection in SKILL.md</li><li>Scripts that steal environment variables, API keys or SSH keys</li><li>curl | bash, obfuscated code and typosquatted packages</li><li>Dangerous calls like exec(), eval() and subprocess</li><li>Poisoned MCP tool descriptions</li></ul><p>Install it once, then pick the method that fits how you work. They are independent: use one, or combine all three.</p><h3>Before you start: install it</h3><p><strong>You need Python 3.12 or newer.</strong></p><pre>uv tool install &#39;skillspector[mcp] @ git+https://github.com/NVIDIA/skillspector.git&#39;</pre><p>The [mcp] part adds the MCP server, which you need for Method 3.</p><h3>Method 1: Scan your skills folder from the terminal</h3><p>Scan a single skill:</p><pre>skillspector scan .claude/skills/my-skill --no-llm</pre><p>Scan every skill in your project:</p><pre>for d in .claude/skills/*/; do skillspector scan &quot;$d&quot; --no-llm; done</pre><p>Do not forget your personal skills too:</p><pre>for d in ~/.claude/skills/*/; do skillspector scan &quot;$d&quot; --no-llm; done</pre><p>Found a skill on GitHub? Scan it <strong>before</strong> you install it:</p><pre>skillspector scan https://github.com/user/some-skill --no-llm</pre><p>--no-llm keeps the scan fast and local. Want a deeper check? Remove it and set a model, for example export SKILLSPECTOR_PROVIDER=anthropic with your ANTHROPIC_API_KEY. Already logged into the Claude CLI? Use SKILLSPECTOR_PROVIDER=claude_cli and no key is needed.</p><blockquote><strong><em>Try it:</em></strong><em> create a skill whose script sends </em><em>os.environ to a URL, then scan it. Guess which rules fire before you hit enter.</em></blockquote><h3>Method 2: Use it in VS Code</h3><p>Add a one click task. Create .vscode/tasks.json:</p><pre>{<br>  &quot;version&quot;: &quot;2.0.0&quot;,<br>  &quot;tasks&quot;: [<br>    {<br>      &quot;label&quot;: &quot;SkillSpector: scan skills&quot;,<br>      &quot;type&quot;: &quot;shell&quot;,<br>      &quot;command&quot;: &quot;for d in .claude/skills/*/; do skillspector scan \&quot;$d\&quot; --no-llm; done&quot;,<br>      &quot;problemMatcher&quot;: []<br>    }<br>  ]<br>}</pre><p>Now press <strong>Ctrl+Shift+P</strong>, choose <strong>Tasks: Run Task</strong>, and pick <strong>SkillSpector: scan skills</strong>. The report shows up in the terminal.</p><p>Want findings next to your code? Export SARIF and open it with Microsoft’s <strong>SARIF Viewer</strong> extension:</p><pre>skillspector scan .claude/skills/my-skill --no-llm --format sarif --output skillspector.sarif</pre><p>Click a finding and VS Code jumps to the exact file and line. (On Windows, run this inside WSL, since file output needs a POSIX system.)</p><h3>Method 3: Let your agent scan before it installs (MCP)</h3><p>This is the best part. Run SkillSpector as an MCP server and your agent can check a skill <strong>before</strong> installing it.</p><p><strong>Claude Code:</strong></p><pre>claude mcp add skillspector -- skillspector mcp</pre><p><strong>VS Code (Copilot agent mode):</strong> create .vscode/mcp.json:</p><pre>{<br>  &quot;servers&quot;: {<br>    &quot;skillspector&quot;: {<br>      &quot;type&quot;: &quot;stdio&quot;,<br>      &quot;command&quot;: &quot;skillspector&quot;,<br>      &quot;args&quot;: [&quot;mcp&quot;]<br>    }<br>  }<br>}</pre><p>Your agent now has a scan_skill tool. It returns a risk_score, a recommendation, a safe_to_install flag and the findings.</p><p>Add one line to your project instructions (like CLAUDE.md):</p><blockquote><em>Before installing any skill or MCP server, run </em><em>scan_skill on it. Do not install if </em><em>safe_to_install is false.</em></blockquote><p>Now just ask your agent: <em>“Scan the skills in .claude/skills and tell me which ones are risky.”</em></p><p><strong>Quick checklist</strong></p><ul><li>Installed SkillSpector</li><li>Scanned my project and personal skills folders</li><li>Added the VS Code task</li><li>Connected the MCP server to my agent</li><li>Added the “scan before install” rule</li></ul><p>Skills are the new npm packages: easy to share, rarely read. Five minutes with SkillSpector closes that gap. Scan the skill you trust most today and tell me in the comments what you found.</p><p><strong>Source Links:</strong></p><ul><li><a href="https://github.com/NVIDIA/SkillSpector">SkillSpector on GitHub</a> [Repo]</li><li><a href="https://docs.nvidia.com/skills/scanning-agent-skills">NVIDIA scanning guide</a> [Docs]</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=61ae8676900a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A 12-Year-Old Bug, 47 CVEs in 48 Hours: How I Fixed Legacy PHP and Node.js Servers with Claude AI]]></title>
            <link>https://rahulh123.medium.com/a-12-year-old-bug-47-cves-in-48-hours-how-i-fixed-legacy-php-and-node-js-servers-with-claude-ai-f1747a4b166d?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/f1747a4b166d</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[claude-ai]]></category>
            <category><![CDATA[php]]></category>
            <category><![CDATA[nodejs]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Fri, 23 Oct 2026 17:36:10 GMT</pubDate>
            <atom:updated>2026-10-23T17:36:10.469Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Copy-paste Claude prompts to find, patch and verify PHP2Shell, Node.js and GitLab CVEs on legacy AWS and cPanel projects, without breaking production.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ejCXDnaDDIo4vTysguZd5g.png" /></figure><h3>The week the old code came due</h3><p>In roughly 48 hours (October 21–22, 2026), the industry disclosed 47 CVEs across PHP, Node.js, Python, Go, GitLab and core internet infrastructure. One of them, a CVSS 10.0 in GitLab, reaches all the way back to version 7.0 — code that shipped about 12 years ago.</p><p>The pattern is the same everywhere: the bugs are not in your app. They are in the runtime, the CA bundle, the package you pinned in 2021, the cPanel box nobody has touched since the last developer left. Updating WordPress or Laravel will not save you from a bug that lives inside PHP itself.</p><p>This post is the playbook I used to triage and patch legacy projects on AWS and cPanel this week, with the exact Claude prompts that did most of the grunt work. Copy them, paste your own output in, and keep a human hand on every command that touches production.</p><h3>What actually happened</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CPqrrQp1OQHB5tEwYiPDDg.png" /></figure><p>The one that matters most for legacy hosting is PHP2Shell. About half of PHP sites run end-of-life branches (7.4, 8.0, 8.1), and those get only source backports — which is exactly what an old cPanel or EC2 box is running.</p><h3>Triage: today vs this week</h3><p>One rule before anything else: if a server may already be compromised, patching does not clean it. Rotate every secret it held.</p><p><strong>Today</strong></p><ul><li>PHP: upgrade to 8.2.35, 8.3.36, 8.4.27 or 8.5.12. Restart Apache, PHP-FPM and every worker, then rebuild images. On 8.1 or older, use the cPanel/Plesk backport, upgrade, or isolate the server.</li><li>LiteLLM: if you touched 1.107.0, treat the machine as compromised. Rebuild on 1.106.2 and rotate everything. Do not run pip on it first.</li><li>Railway: rotate every environment variable, AI provider keys first. Regenerate database passwords and check provider billing.</li><li>Self-managed GitLab: 19.5.1 or your line’s fix (19.4.3, 19.3.6, 19.2.9, 19.1.9, 19.0.10, 18.11.13). Can’t? Take it off the public internet.</li><li>Plex 1.43.3+ and close port 32400. NetScaler 14.1–73.37 or 13.1–64.23, then hunt for compromise.</li></ul><p><strong>This week</strong></p><ul><li>Node.js 26.12.1, 24.22.1 or 22.23.4; Bun 1.4.4.</li><li>Next.js 16.4.1 or 15.5.28. Laravel 13.29.0 or 12.68.0.</li><li>CA bundles: OS ca-certificates, Go 1.27.3 or 1.26.10 (then rebuild binaries), certifi 2026.10.19. Reissue any eMudhra certificate from another CA before October 27.</li><li>DNS: BIND 9.20.30 / 9.21.27, Unbound 1.26.2, PowerDNS Recursor 5.3.12 / 5.4.8.</li><li>Tomcat 11.0.27, 10.1.61 or 9.0.123. Pillow 13.0.0. virtualenv 21.14.9. Ruby 4.0.9, 3.4.12 or 3.3.13.</li></ul><h3>The two master prompts: PHP and Node.js</h3><blockquote>If you only copy two things from this post, copy these. Each one takes Claude from inventory to verified fix in a single conversation, and works for cPanel, plain EC2, Docker or ECS. Fill in the &lt;angle brackets&gt;, paste your read-only output at the bottom, and follow the phases in order.</blockquote><h3>Master prompt A: PHP2Shell mitigation</h3><pre>ROLE<br>You are a senior PHP and Linux security engineer helping me mitigate PHP2Shell on a<br>legacy production system. Be precise, conservative and production-safe.</pre><pre>MY ENVIRONMENT<br>- Hosting: &lt;cPanel/WHM | EC2 | Docker/ECS | other&gt;<br>- OS: &lt;e.g. CloudLinux 7, AlmaLinux 8, Amazon Linux 2, Ubuntu 20.04&gt;<br>- PHP versions in use: &lt;e.g. 7.4, 8.1, 8.2&gt;<br>- Handler: &lt;mod_php | PHP-FPM | LSAPI | php-cgi | FrankenPHP&gt;<br>- Web server / proxy: &lt;Apache | NGINX | LiteSpeed | ALB | Cloudflare&gt;<br>- Apps: &lt;WordPress, Laravel, Magento, custom...&gt;<br>- Maintenance window: &lt;e.g. 30 min per site, after 01:00 IST&gt;</pre><pre>THE VULNERABILITY<br>CVE-2026-109618 and CVE-2026-109619 (CVSS 9.2), unauthenticated RCE in the PHP<br>engine itself, all versions. Upgrading WordPress/Laravel does NOT fix it.<br>- -109618: out-of-memory mid-request leaves broken state used at shutdown.<br>- -109619: one request&#39;s state outlives its memory; next request uses freed memory.<br>Both can leak other users&#39; session tokens and API keys. mod_php is worst.<br>Fixed upstream: 8.2.35, 8.3.36, 8.4.27, 8.5.12. 7.4/8.0/8.1 get source backports<br>only (cPanel and Plesk ship them). The fix adds a hard memory limit.</pre><pre>WORK IN THESE PHASES. STOP AFTER EACH ONE AND WAIT FOR &quot;go&quot;.</pre><pre>Phase 1 - Inventory (read-only)<br>Give me only read-only commands to list every PHP binary, version, handler, FPM<br>pool, vhost mapping, pinned versions in Dockerfiles/CI/Helm, and the package<br>changelog lines that mention the two CVE IDs. Do not trust `php -v` for distro or<br>community packages (tags like 8.4.26 may carry backports, not the upstream fix).</pre><pre>Phase 2 - Risk table<br>From my output, build a table: site | PHP version | handler | patched? (evidence) |<br>exposure | action (upgrade / backport / mitigate / isolate). Order by risk.</pre><pre>Phase 3 - Patch<br>For each item marked upgrade or backport, give exact commands for my platform<br>(EasyApache 4, dnf/apt, Docker base image + digest), a backup step before, and a<br>rollback step after. Then the full restart sequence so no process keeps old code:<br>Apache, every PHP-FPM pool, LSAPI, queue workers, cron daemons, opcache. Recheck<br>memory_limit and post_max_size, since the fix enforces a hard limit.</pre><pre>Phase 4 - Temporary mitigation (only for what cannot be patched today)<br>- -109618: memory_limit=-1 via php_admin_value (pool or vhost), post_max_size=8M<br>  (32M where uploads exist), proxy limits on request body size and multipart part<br>  count. Do NOT use RLIMIT_AS or RLIMIT_DATA.<br>- -109619: session.upload_progress.enabled=0 in php.ini or the FPM pool, then restart<br>  workers. On 8.5 also set fatal_error_backtraces=0.<br>- WAF rules (Cloudflare/ModSecurity/AWS WAF) as an extra layer, labelled &quot;buys time,<br>  not a fix&quot;.<br>- If none of this is possible: how to take the site offline or restrict by IP safely.<br>State the side effect of every setting in one line.</pre><pre>Phase 5 - Verify<br>Commands that prove each site now runs a fixed build and the mitigation settings are<br>active in the running workers (not just in a file), plus a smoke test per site.</pre><pre>Phase 6 - Compromise check (logs since October 17, 2026)<br>Grep/awk for: odd multipart requests followed by memory-limit errors or 500 bursts,<br>oversized PHPSESSID cookies, unexplained worker restarts, unexpected ELF/Go<br>binaries, new cron jobs, and these IPs: 103.216.220.92, 31.223.184.166,<br>37.19.200.5, 134.209.159.64, 103.114.163.188, 95.217.98.37.<br>Verdict per server: clean / suspicious / likely compromised, with evidence. If<br>suspicious or worse: patching does not clean it; list every secret to rotate.</pre><pre>RULES<br>- Read-only until I say &quot;go&quot;. Mark every command that changes state in bold.<br>- Never ask me for secret values. I will redact.<br>- If unsure about a version or flag, say so instead of guessing.</pre><pre>MY OUTPUT:<br>&lt;paste here&gt;</pre><h3>Master prompt B: Node.js October security release</h3><pre>ROLE<br>You are a senior Node.js and DevOps security engineer helping me patch and mitigate<br>this week&#39;s Node.js security release on legacy production apps. Be precise,<br>conservative and production-safe.</pre><pre>MY ENVIRONMENT<br>- Hosting: &lt;EC2 + pm2/systemd | ECS/Fargate | Lambda | cPanel Node.js Selector | Docker&gt;<br>- Node versions in use: &lt;e.g. 16, 18, 20, 22&gt;  (also Bun? &lt;yes/no, version&gt;)<br>- Frameworks: &lt;Express, NestJS, Next.js, Storybook...&gt;<br>- Package manager: &lt;npm | yarn | pnpm&gt;<br>- Do we decompress Brotli from users or third parties? &lt;yes/no/unsure&gt;<br>- Do we make outbound TLS calls with custom checkServerIdentity or servername? &lt;yes/no/unsure&gt;</pre><pre>THE VULNERABILITIES (fixed in 26.12.1, 24.22.1, 22.23.4; EOL lines get NO fix)<br>- High: CVE-2026-111497, Brotli decoder out-of-bounds write, possible RCE when<br>  decompressing attacker-controlled Brotli.<br>- High: CVE-2026-105483 and CVE-2026-106069, TLS hostname check bypasses.<br>- High: CVE-2026-105482, dns.resolve truncates names at a NUL byte.<br>- Medium: keep-alive response desync; randomUUID/randomInt repeating values in<br>  snapshots and single executable apps.<br>- Also ships the eMudhra CA blocklist, OpenSSL 3.5.9, new c-ares and undici.<br>- Bun fixed in 1.4.4. Related: Next.js 16.4.1 / 15.5.28 (two critical RCEs),<br>  Storybook 8.6.19 / 9.1.21 / 10.6.2.<br>- Known regressions: --shared-brotli builds break (nodejs/node#66895); older<br>  @mswjs/interceptors can segfault on TLS (nodejs/node#66910).</pre><pre>WORK IN THESE PHASES. STOP AFTER EACH ONE AND WAIT FOR &quot;go&quot;.</pre><pre>Phase 1 - Inventory (read-only)<br>Read-only commands to find every Node runtime and pin: running `node -v` per<br>service, .nvmrc, package.json engines, Dockerfile FROM lines (tag and digest),<br>CI setup-node versions, Lambda runtimes (aws lambda list-functions), ECS task<br>definitions, cPanel Node.js Selector apps, Bun versions. Plus: where the code<br>calls zlib Brotli APIs, accepts Content-Encoding: br, uses dns.resolve on user<br>input, sets custom TLS identity checks, or builds snapshots/SEA binaries.</pre><pre>Phase 2 - Risk table<br>service | Node version | EOL? | exposed bugs (based on code usage) | action. EOL lines<br>(16, 18, 20, and any odd release) go straight to &quot;upgrade to 22.23.4+&quot; with a note on<br>breaking changes I should test.</pre><pre>Phase 3 - Patch<br>Exact steps per platform: nvm/n, apt/dnf NodeSource, Docker base image pinned by<br>digest, CI updates, Lambda runtime change, ECS rolling deploy, pm2 reload, cPanel<br>Selector. Rebuild native modules (npm rebuild) and lockfile-only updates where<br>needed. Upgrade Next.js/Storybook/Bun if present. Backup and rollback for each.<br>Check my deps for the two known regressions and tell me the fix.</pre><pre>Phase 4 - Temporary mitigation (only where I cannot upgrade today)<br>- Brotli: stop decompressing untrusted br bodies (reject Content-Encoding: br at<br>  the proxy/ALB/WAF, or remove br from body parsers); set decompressed size limits.<br>- TLS: avoid custom checkServerIdentity; pin expected hostnames at the HTTP client.<br>- DNS: reject hostnames containing NUL or control characters before dns.resolve.<br>- Keep-alive desync: disable keep-alive to untrusted upstreams if needed.<br>- Snapshots/SEA: don&#39;t use randomUUID/randomInt for secrets until rebuilt.<br>Provide code snippets for Express/NestJS middleware and the matching NGINX/ALB/<br>Cloudflare rule. State the side effect of each.</pre><pre>Phase 5 - Verify<br>Prove every running process uses the fixed version (not just the image), confirm<br>the eMudhra test URL <a href="https://dv-g3b-emudhra.we-have-a-problem.com/">https://dv-g3b-emudhra.we-have-a-problem.com/</a> now FAILS from<br>Node, and give a smoke test per service.</pre><pre>Phase 6 - Exposure check<br>Logs to review for crashes/segfaults in Node workers, unusual br-encoded requests,<br>and TLS errors since October 21, 2026. Verdict per service, with evidence.</pre><pre>RULES<br>- Read-only until I say &quot;go&quot;. Mark state-changing commands in bold.<br>- Never disable TLS verification (NODE_TLS_REJECT_UNAUTHORIZED=0) to fix errors.<br>- Never ask for secret values. If unsure about a version or API, say so.</pre><pre>MY OUTPUT:<br>&lt;paste here&gt;</pre><h3>Fixing a legacy cPanel server with Claude</h3><p>The workflow: run read-only commands yourself, paste the output into Claude, and let it tell you what is vulnerable and what to run next. Never give an AI root SSH on a box you are trying to secure.</p><h3>Prompt 1: inventory</h3><p>First collect the facts (all read-only) and paste them below the prompt:</p><pre>/usr/local/cpanel/cpanel -V<br>ls /opt/cpanel/ | grep ea-php<br>for v in /opt/cpanel/ea-php*/root/usr/bin/php; do echo &quot;$v&quot;; $v -v | head -1; done<br>rpm -q --changelog ea-php81-php-cli | head -40<br>whmapi1 php_get_vhost_versions | grep -E &#39;vhost|version&#39;<br>httpd -M | grep -E &#39;php|suphp|lsapi|fcgid&#39;</pre><pre>You are a senior Linux/PHP security engineer. I run a legacy cPanel/WHM server.<br>Below is read-only output from it. This week PHP disclosed PHP2Shell<br>(CVE-2026-109618 and CVE-2026-109619, CVSS 9.2, unauthenticated RCE in the PHP<br>engine). Fixed upstream in 8.2.35, 8.3.36, 8.4.27, 8.5.12; cPanel ships<br>backports for 7.4-8.1.</pre><pre>1. List every PHP version installed and which vhosts use each.<br>2. For each version, tell me whether the package changelog shows the CVE fix.<br>   Do not trust `php -v` alone.<br>3. Flag any site on mod_php (worst case) vs PHP-FPM/LSAPI.<br>4. Give me a prioritised table: site, PHP version, handler, patched yes/no, action.<br>Only propose read-only commands until I say &quot;go&quot;.</pre><pre>&lt;paste output here&gt;</pre><h3>Prompt 2: patch plan with rollback</h3><pre>Using the inventory above, write a step-by-step patch plan for this cPanel server:<br>- EasyApache 4 package updates (yum/dnf) for each ea-phpXX that needs it<br>- which sites I should move to PHP 8.2+ now vs keep on a backported 7.4/8.1<br>- the exact restart sequence so no worker keeps old code (Apache, PHP-FPM pools,<br>  LSAPI, opcache)<br>- a rollback step for each change, and a backup command before I start<br>- a check after each step that proves the fix is live<br>Assume a 15-minute maintenance window per site. Mark any step that can break<br>production in bold.</pre><h3>Prompt 3: temporary mitigation for sites you cannot upgrade</h3><pre>I have sites on PHP 7.4 that I cannot upgrade this week. Using php.net&#39;s<br>partial mitigations for PHP2Shell, generate:<br>- per-pool PHP-FPM settings: memory_limit=-1 via php_admin_value,<br>  post_max_size 8M (32M where uploads exist), session.upload_progress.enabled=0<br>- Apache/ModSecurity limits on request body size and multipart part count<br>- how to apply these through WHM MultiPHP INI Editor, not by hand-editing files<br>Explain the trade-off of each setting in one line. Remind me that a WAF<br>buys time but is not a fix.</pre><h3>Prompt 4: did anyone get in?</h3><pre>Here are my Apache access/error logs and PHP-FPM logs since October 17.<br>Look for PHP2Shell indicators: odd multipart requests followed by memory-limit<br>errors or bursts of 500s, oversized PHPSESSID cookies, unexplained worker<br>restarts, and requests from 103.216.220.92, 31.223.184.166, 37.19.200.5,<br>134.209.159.64, 103.114.163.188, 95.217.98.37.<br>Give me grep/awk commands for the parts I haven&#39;t pasted, then summarise:<br>clean / suspicious / likely compromised, with evidence lines.</pre><pre>&lt;paste logs here&gt;</pre><h3>Fixing legacy projects on AWS with Claude</h3><p>On AWS the problem is usually not one server. It is a dozen EC2 instances, an old AMI, three Dockerfiles with pinned base images, and a Lambda nobody remembers. Point Claude at your infrastructure code, not your live account.</p><h3>Prompt 5: find every pinned runtime in the repo</h3><p>If you use Claude Code, run this from the repo root. Otherwise paste the files.</p><pre>Scan this repository for every place a runtime or base image version is pinned:<br>Dockerfiles, docker-compose, Helm charts, Terraform/CloudFormation (AMI IDs,<br>Lambda runtimes, ECS task definitions), CI workflows, .nvmrc, .python-version,<br>composer.json &quot;platform&quot;, go.mod, requirements*.txt, package.json engines.</pre><pre>Compare each against this week&#39;s fixed versions:<br>PHP 8.2.35 / 8.3.36 / 8.4.27 / 8.5.12; Node 22.23.4 / 24.22.1 / 26.12.1;<br>Go 1.27.3 / 1.26.10; certifi 2026.10.19; Pillow 13.0.0; Next.js 15.5.28 / 16.4.1;<br>Laravel 12.68.0 / 13.29.0; Tomcat 9.0.123 / 10.1.61 / 11.0.27; litellm pinned<br>to 1.106.2 (never 1.107.0 or `latest`).</pre><pre>Output a table: file, line, current version, target version, risk.<br>Then propose the diff, but do not commit. Pin Docker images by digest.</pre><h3>Prompt 6: patch the EC2 fleet safely</h3><pre>I have legacy EC2 instances (Amazon Linux 2, Ubuntu 18.04/20.04) running PHP<br>and Node apps behind an ALB. Write an AWS Systems Manager plan to:<br>1. Inventory installed php, nodejs, ca-certificates and openssl versions<br>   across all instances (SSM Inventory or Run Command, read-only first).<br>2. Patch in batches of 20% with Patch Manager, draining each instance from the<br>   target group before restart and re-registering after a health check.<br>3. Restart php-fpm, httpd/nginx, pm2/systemd Node services so no old code<br>   stays in memory.<br>4. For instances on end-of-life OS releases, tell me which ones can&#39;t get the<br>   fix and recommend replace-with-new-AMI vs isolate.<br>Give the exact AWS CLI / SSM document names. Mark anything destructive.</pre><h3>Prompt 7: hunt for stolen-key abuse in CloudTrail</h3><p>The LiteLLM backdoor and the Railway breach both went after cloud and AI keys. Attackers started with an identity check and a Bedrock model listing.</p><pre>Write AWS CLI and Athena queries to search CloudTrail in EVERY region since<br>October 17, 2026 for:<br>- sts:GetCallerIdentity and bedrock:List* / bedrock:InvokeModel from IPs<br>  outside my known ranges &lt;list your office/VPN/NAT IPs&gt;<br>- new IAM users, access keys, or roles created by an access key rather than SSO<br>- calls from 168.100.8.57 and 165.227.201.8<br>Then tell me how to read the results and which keys to deactivate first.</pre><h3>Prompts for secrets, LiteLLM, GitLab and certificates</h3><h3>Prompt 8: the secrets rotation runbook</h3><pre>I need to rotate every secret a possibly-compromised server could reach.<br>Here is a redacted list of env var NAMES (no values) from the server, my<br>Railway project and my CI: &lt;paste names only&gt;.<br>Group them by provider, order them by blast radius (AI provider keys and cloud<br>keys first, then Stripe, database passwords, GitHub/registry tokens, SSH keys),<br>and for each give: where to rotate it, what to redeploy afterwards, and how to<br>confirm the old one is dead. Remind me to export the GitHub org audit log<br>before revoking anything.</pre><p>Never paste secret values into any chat, including this one. Names are enough.</p><h3>Prompt 9: LiteLLM 1.107.0 check</h3><pre>Give me a read-only script that checks a Linux host or Docker image for the<br>LiteLLM 1.107.0 backdoor without starting Python or pip:<br>- cook.pth, cook.py, sitecustomize.py at the root of any site-packages<br>- /tmp/.pconce*, /tmp/remotecodeloader.lock<br>- env var PRIVCOLLECT_ACTIVE=1 in any running process<br>- outbound connections to litellm.dev, 168.100.8.57, 165.227.201.8<br>- image tags v1.107.0 / latest / main-latest / main-stable pulled on Oct 21<br>Print FOUND/CLEAN per check. If anything is FOUND, tell me to rebuild from<br>scratch on 1.106.2 rather than clean in place.</pre><h3>Prompt 10: self-managed GitLab</h3><pre>My self-managed GitLab is version &lt;X&gt; on &lt;Omnibus/Docker/Helm&gt;. CVE-2026-110264<br>(CVSS 10) is fixed in 19.5.1, 19.4.3, 19.3.6, 19.2.9, 19.1.9, 19.0.10, 18.11.13.<br>Write the upgrade path from my version, including required stop versions and<br>the migrations in 19.4.3/19.3.6 that need downtime on a single node.<br>If I can&#39;t upgrade today, give the exact settings to restrict Public/Internal<br>visibility and default new projects to Private, and a firewall rule to take it<br>off the internet.</pre><h3>Prompt 11: CA bundles and eMudhra</h3><pre>List every place in this repo and Docker images where a CA bundle comes from:<br>OS ca-certificates, certifi, Go&#39;s crypto/x509 (needs rebuild with 1.27.3 or<br>1.26.10), Node&#39;s bundled store, Java cacerts, vendored PEM files.<br>For each, give the update command and a test: requests to<br>https://dv-g3b-emudhra.we-have-a-problem.com/ must FAIL after the fix.<br>Do not suggest disabling TLS verification to silence errors.</pre><h3>Rules for letting an AI near production</h3><ol><li>Read-only first. Let Claude analyse output; you run the commands that change things.</li><li>Never paste secret values, private keys or customer data. Redact before you paste.</li><li>Ask for a rollback for every step, and a backup before the first.</li><li>Verify, don’t trust. Check package changelogs for the CVE IDs, and re-run the test after each fix.</li><li>Don’t run random “PoC” repos. A fresh-account GitLab exploit repo appeared hours after disclosure.</li></ol><h3>Myths going around this week</h3><ul><li>PHP2Shell is not a CVSS 10. It is 9.2.</li><li>No country shut down its internet; individual sites went offline as a precaution.</li><li>There is no evidence a third of websites are compromised.</li><li>CVE-2026–99999 is a VM escape that matters only on hypervisor hosts, not an internet-wide risk.</li><li>Nobody stole an eMudhra root CA. The bad certificates were intermediates, and they are distrusted.</li><li>NGINX did not get three new critical flaws; those were patched October 14.</li><li>No primary source links the Railway breach to LiteLLM.</li></ul><h3>The takeaway</h3><p>Legacy code doesn’t fail because it’s old. It fails because nobody knows what’s in it. The fastest win this week was not a clever exploit analysis — it was an honest inventory of every PHP version, base image and CA bundle we were running. Claude turned that from a two-day job into an afternoon. The patching, the restarts and the judgement calls stayed with us.</p><p>Patch today, rotate what might have leaked, and pin your versions so next time you know exactly where to look.</p><p><em>Version numbers and incident details are as of October 23, 2026. Check each vendor’s advisory before you patch, since fixes are still landing.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f1747a4b166d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Used AI to Create 5 Completely Different Portfolio Websites With Just One Resume]]></title>
            <link>https://rahulh123.medium.com/i-used-ai-to-create-5-completely-different-portfolio-websites-with-just-one-resume-9d591ad4c005?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/9d591ad4c005</guid>
            <category><![CDATA[tech-leadership]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[claude]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Tue, 13 Oct 2026 04:01:02 GMT</pubDate>
            <atom:updated>2026-10-13T04:01:02.057Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*W1QayfgDO3fKvyU7YkghSA.png" /></figure><p>What if you could take a single resume and transform it into five completely different portfolio websites just by changing the design prompt?</p><p>That’s exactly what I explored in my latest AI experiment.</p><p>From futuristic Spatial UI to Liquid Glass, from Terminal-inspired interfaces to Brutalism and Maximalism — each version brings a completely different visual identity to the same underlying content.</p><p>And the surprising part? AI helped me bring the entire project to life.</p><h3>🎨 Five Designs, One Portfolio</h3><p>Instead of manually designing five websites from scratch, I experimented with different design directions and used AI to turn those ideas into working portfolio experiences.</p><p>Here are the five styles:</p><ul><li><strong>Spatial UI:</strong> A modern, immersive interface with a sense of depth.</li><li><strong>Liquid Glass:</strong> Translucent surfaces, layered elements, and a sleek visual aesthetic.</li><li><strong>Terminal:</strong> A developer-inspired interface influenced by command-line environments.</li><li><strong>Brutalism:</strong> Bold typography, raw layouts, and unapologetic visual choices.</li><li><strong>Maximalism:</strong> Expressive visuals, vibrant details, and a design that stands out.</li></ul><p>The content stayed the same. The design prompt changed.</p><p>The result was five different ways to present the same professional identity.</p><h3>🤖 The Entire Workflow Was AI-Powered</h3><p>This experiment wasn’t just about generating a website.</p><p>AI helped me take the idea from creation to presentation:</p><ul><li>Building the portfolio website.</li><li>Exploring multiple design styles.</li><li>Creating and refining the visual experience.</li><li>Producing a video to showcase the results.</li><li>Preparing content to share the experiment online.</li></ul><p>It demonstrates how AI can support more than coding. It can help connect development, design exploration, and content creation into one workflow.</p><h3>💡 The Real Skill Is Knowing What to Ask</h3><p>AI is changing how we approach frontend development.</p><p>Previously, creating several distinct portfolio designs could mean hours of designing, coding, testing, and refining each version.</p><p>Now, a clear description of the desired experience can give you a starting point remarkably quickly.</p><p>But the quality of the result still depends on the clarity of your ideas, the prompts you write, and how you refine the output.</p><p><strong>The advantage isn’t simply using AI. It’s knowing how to guide it.</strong></p><h3>🔗 Explore the Project</h3><p>Try the live portfolio and explore the different design possibilities.</p><p><strong>Live Demo:</strong> <a href="https://portolio-ui.netlify.app/"><strong>https://portolio-ui.netlify.app/</strong></a></p><p><strong>GitHub Repository:</strong> <a href="https://github.com/barotrahulh123/claude-portfolio-ui-experiments">https://github.com/barotrahulh123/claude-portfolio-ui-experiments</a></p><p>The prompts are free to copy, so you can experiment with the styles and adapt them to your own portfolio.</p><p>Portfolio Video Created Using Claude AI | Exploring the Evolution of AI-Powered Web Design</p><p><a href="https://youtu.be/eT9w1m-JNfM">https://youtu.be/eT9w1m-JNfM</a></p><h3>🎯 Your Turn</h3><p>If you could choose one design style for your personal portfolio, which would it be?</p><ul><li>Spatial UI</li><li>Liquid Glass</li><li>Terminal</li><li>Brutalism</li><li>Maximalism</li></ul><p>I’d love to know which one you’d choose and why.</p><p>Follow along for more AI experiments, practical workflows, and projects exploring what’s possible when creativity meets AI.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9d591ad4c005" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How I’d Design a Government Exam Portal for 10 Million Candidates]]></title>
            <link>https://rahulh123.medium.com/how-id-design-a-government-exam-portal-for-10-million-candidates-38e1e8431f81?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/38e1e8431f81</guid>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[tech-lead]]></category>
            <category><![CDATA[high-level-design]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Mon, 05 Oct 2026 04:11:01 GMT</pubDate>
            <atom:updated>2026-10-05T04:11:01.574Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A practical HLD and LLD walkthrough, using Gujarat’s OJAS portal as the reference model.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bpqblT3WVUPEHMzu4ULI-w.png" /></figure><p>Every aspirant in India knows the feeling. It is the last date for a big recruitment form, the portal crawls, the payment page times out, and money leaves your account while the application still says “fee pending”.</p><blockquote>The hard part of a government exam portal is not the forms. It is that millions of people arrive on the same few days, and every one of them is anxious about a payment. In this post I’ll design a portal like <a href="https://ojas.gujarat.gov.in/">OJAS</a> for 10 million registered candidates, from the high-level architecture down to tables, APIs and payment handling.</blockquote><p>We’ll cover requirements, capacity estimates, the architecture, how to survive the last date and result day, and the low-level design of the parts that usually break.</p><h3>What the portal must do</h3><p>The portal has a simple job on paper: register a candidate once, let them apply to many recruitments, take a fee, and later hand out call letters and results.</p><p><strong>Functional requirements</strong></p><ol><li><strong>One Time Registration (OTR).</strong> A candidate registers once with mobile, email, name, date of birth, category and an ID proof, and reuses that profile everywhere.</li><li><strong>Advertisements.</strong> Boards such as GPSC, GSSSB and Police publish recruitments with posts, eligibility rules, fee rules and open/close dates.</li><li><strong>Applications.</strong> The form is prefilled from OTR, plus fields specific to each advertisement.</li><li><strong>Uploads.</strong> Photo and signature, and certificates for some posts.</li><li><strong>Fees.</strong> Online payment through a bank gateway, or offline through a post office or bank challan. Some categories pay nothing.</li><li><strong>Confirmation.</strong> A confirmation number and a printable application PDF.</li><li><strong>Changes.</strong> Edit or cancel only inside a defined window.</li><li><strong>Later stages.</strong> Call letters, admit cards, results and merit lists.</li><li><strong>Admin.</strong> Advertisement management, reports, candidate search and exports.</li></ol><p><strong>Non-functional requirements</strong></p><ul><li><strong>Scale:</strong> 10 million registered users, with headroom to 25 million.</li><li><strong>Availability:</strong> 99.95%, and zero data loss for confirmed applications and payments.</li><li><strong>Latency:</strong> p95 under 300 ms for APIs, under 1 second for form submit.</li><li><strong>Traffic shape:</strong> extreme spikes on the last 2–3 days of a form and on result day.</li><li><strong>Compliance:</strong> data hosted in India, the DPDP Act 2023, CERT-In guidelines, accessibility (GIGW), and a bilingual Gujarati/English UI.</li><li><strong>Integrity:</strong> no double payment, no lost payment, one application per candidate per post.</li></ul><h3>Back-of-the-envelope numbers</h3><p>The surprise in the math is that form submissions are not the scary part. Result day is.</p><p><strong>A big recruitment.</strong> Assume a police-style advertisement attracts 2.5 million applications over 30 days, and 40% of them land in the last 3 days.</p><ul><li>That is about 1 million applications in 3 days, or roughly 330,000 per day.</li><li>If 15% of a day arrives in the peak hour, that is about 50,000 per hour, or around 14 submissions per second.</li><li>Each application involves 30–50 API calls (pages, autosave, uploads, payment, PDF). That works out to roughly 700 requests per second sustained, with bursts of 3,000–5,000.</li></ul><p><strong>Result day.</strong> Now imagine the same 2.5 million candidates arriving within about 2 hours, most of them in the first 15 minutes. That is 20,000–30,000 reads per second. No sensible backend should absorb that. A CDN should.</p><p><strong>Storage</strong></p><ul><li>Photo and signature at about 70 KB per user: around 700 GB for 10 million users.</li><li>Application PDFs and certificates: 3–5 TB per year.</li><li>Application rows at about 5 KB each, 4 applications per user per year: about 200 GB per year.</li><li>Payment and audit logs: about 500 GB per year.</li></ul><p><strong>Design targets:</strong> 50,000 concurrent users, a 30,000 RPS read peak served mostly from the CDN, and 3,000 writes per second as a safety margin.</p><p>None of this needs exotic technology. A well-tuned PostgreSQL with partitioning handles 10 million users comfortably, without sharding.</p><blockquote><em>let’s begin with HLD</em></blockquote><p>The architecture is four layers deep, and the most important one is the first: the edge, which absorbs most of the traffic before it ever reaches a server.’</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aOD218zxpxb28KT_v55r-g.png" /></figure><ul><li><strong>Edge:</strong> CDN for the app, notices and pre-built results; WAF and CAPTCHA against bots; a waiting room when the backend is busy.</li><li><strong>Gateway:</strong> JWT checks, rate limits (e.g. 3 OTPs per mobile per 10 minutes), routing.</li><li><strong>Services:</strong> stateless on Kubernetes, each scaling independently. A small team can start with these as modules in one app; the boundaries matter more than the deployables.</li><li><strong>Data:</strong> PostgreSQL (primary, standby, read replicas), Redis for OTPs/cache/locks, S3 or MinIO for files, Kafka for events.</li><li><strong>Reliability:</strong> two sites, active-passive, near-zero payment data loss, 30-minute recovery. PII encrypted, raw Aadhaar never stored.</li></ul><h4>Surviving the last date and result day</h4><p>Most outages happen on these two days, so plan for them instead of hoping autoscaling helps.</p><ol><li><strong>Waiting room:</strong> admit users at a fixed rate when load is high. A fair queue beats a dead site.</li><li><strong>Autosave drafts:</strong> throttled users resume where they left off.</li><li><strong>Async PDFs:</strong> submit returns instantly; the PDF follows seconds later.</li><li><strong>Pre-built results:</strong> render every admit card and result in advance and serve them from the CDN via hash(confirmation_no + dob).</li><li><strong>Pre-scale</strong> 24 hours before a deadline, and stagger deadlines where policy allows.</li></ol><h3>Low-level design: schema, states and APIs</h3><h3>Low-level design: schema, states and APIs</h3><h3>Schema</h3><p>The key choice: <strong>partition applications by advertisement.</strong> Recruitments stay isolated and old ones archive by detaching a partition. Alongside candidate and advertisement (with eligibility and fee rules as JSONB), the two critical tables are:</p><pre>CREATE TABLE application (<br>  application_id   BIGSERIAL,<br>  advt_id          BIGINT NOT NULL,<br>  candidate_id     BIGINT NOT NULL,<br>  post_id          BIGINT NOT NULL,<br>  confirmation_no  VARCHAR(20),<br>  status           VARCHAR(20) NOT NULL,<br>  form_data        JSONB,<br>  version          INT DEFAULT 0,      -- optimistic locking<br>  PRIMARY KEY (advt_id, application_id),<br>  UNIQUE (advt_id, candidate_id, post_id)<br>) PARTITION BY LIST (advt_id);</pre><pre>CREATE TABLE payment (<br>  payment_id       BIGSERIAL PRIMARY KEY,<br>  application_id   BIGINT NOT NULL,<br>  order_ref        VARCHAR(40) UNIQUE NOT NULL,<br>  idempotency_key  VARCHAR(64) UNIQUE NOT NULL,<br>  amount           NUMERIC(8,2),<br>  gateway_txn_id   VARCHAR(64),<br>  status           VARCHAR(20)         -- INITIATED/SUCCESS/FAILED/PENDING_VERIFY<br>);</pre><h3>State machine</h3><pre>DRAFT → SUBMITTED → FEE_PENDING → CONFIRMED<br>            └── (no fee) ────────────┘<br>FEE_PENDING → EXPIRED   (fee deadline passed)<br>CONFIRMED   → CANCELLED (within window)</pre><p>Every transition checks the window against <strong>server time</strong> and writes an audit log entry.</p><h3>Key APIs</h3><ul><li>POST /auth/otp/send, /auth/otp/verify: CAPTCHA and rate limited.</li><li>POST /documents/upload-url: presigned upload, valid 5 minutes.</li><li>PUT /applications/draft: autosave.</li><li>POST /applications/{id}/submit: eligibility check, Idempotency-Key required.</li><li>POST /payments/initiate and /payments/callback/{gateway}.</li></ul><h3>Payments done right</h3><h3>Payments done right</h3><p>“Money debited, application not confirmed” is the top complaint on government portals. Five rules fix it:</p><ol><li><strong>Ignore the browser redirect.</strong> Confirm only from the bank’s server-to-server callback, then double-check via the bank’s status API.</li><li><strong>Be idempotent.</strong> A unique key per attempt and UPDATE ... WHERE status = &#39;INITIATED&#39; make repeated callbacks harmless.</li><li><strong>Use a transactional outbox.</strong> Payment update and event row share one transaction; a relay publishes to Kafka.</li><li><strong>Reconcile.</strong> Every 15 minutes, re-check payments stuck over 20 minutes; daily, match the bank’s settlement file.</li><li><strong>Treat challans as files.</strong> Mark them paid when the bank or post office settlement file arrives.</li></ol><p>Gateways sit behind one interface, so adding a bank or UPI is a new class, not a rewrite:</p><pre>interface PaymentGateway {<br>    InitPayload createOrder(Payment p);<br>    GatewayStatus fetchStatus(String orderRef);<br>    boolean verifyChecksum(Map&lt;String, String&gt; payload);<br>}</pre><h3>The details that make it work</h3><p><strong>Eligibility as config.</strong> Age limits, relaxations and qualifications live as JSON on each advertisement, so a new recruitment needs no deployment:</p><pre>{<br>  &quot;age&quot;: {&quot;min&quot;: 18, &quot;max&quot;: 35, &quot;asOn&quot;: &quot;2026-10-31&quot;,<br>          &quot;relaxation&quot;: {&quot;SC&quot;: 5, &quot;ST&quot;: 5, &quot;PWD&quot;: 10, &quot;MAX_CAP&quot;: 45}},<br>  &quot;qualification&quot;: {&quot;anyOf&quot;: [&quot;GRADUATE&quot;]}<br>}</pre><p>A factory turns it into EligibilityRule objects (AgeRule, QualificationRule, …) that an engine runs in sequence.</p><p><strong>Uploads.</strong> The browser compresses images and uploads straight to object storage via a presigned URL. A worker validates size and type, scans for viruses and stores a hash; identical photos across candidates flag possible fraud.</p><p><strong>Confirmation numbers.</strong> Short and typo-resistant: advertisement code + year + sequence + Luhn check digit, e.g. GSS2604123457.</p><p><strong>No duplicates.</strong> A unique (advt_id, candidate_id, post_id) constraint, a Redis lock against double-tab submits, and a version column for safe edits.</p><p><strong>Admit cards.</strong> At ~50 ms per PDF, 200 workers render 2.5M cards in about 10 minutes, ready on the CDN before release.</p><h3>Tech stack and takeaways</h3><p><strong>A sensible stack:</strong></p><ul><li><strong>Frontend:</strong> React or Next.js as a PWA, with Gujarati/English i18n and WCAG 2.1 AA accessibility.</li><li><strong>Backend:</strong> Java Spring Boot or Go, on Kubernetes behind an NGINX or Kong gateway.</li><li><strong>Data:</strong> PostgreSQL 16 with Patroni for failover, Redis 7, Kafka, and MinIO or S3 for files.</li><li><strong>Observability:</strong> Prometheus and Grafana, OpenSearch for logs, Jaeger for tracing.</li><li><strong>Load testing:</strong> k6 or JMeter at 3× expected peak before every major recruitment.</li></ul><p><strong>Alert on:</strong> p95 latency, payment success rate (below 85%), reconciliation backlog, queue length, replica lag, SMS delivery.</p><p><strong>Three takeaways:</strong></p><ol><li><strong>Design for the spike.</strong> A waiting room and CDN-served results beat any autoscaling.</li><li><strong>Payments are a distributed systems problem.</strong> Server confirmation, idempotency and reconciliation are non-negotiable.</li><li><strong>Partition by the natural boundary,</strong> here the advertisement.</li></ol><blockquote>“Use boring, well-understood components, arranged thoughtfully so that even the worst day of the year is just a busy day.”</blockquote><blockquote>— Happy Architect</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=38e1e8431f81" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Use Cursor Like a Legend Developer: The Ultimate Guide]]></title>
            <link>https://rahulh123.medium.com/how-to-use-cursor-like-a-legend-developer-the-ultimate-guide-3ad0ac51acc0?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/3ad0ac51acc0</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[cursor]]></category>
            <category><![CDATA[prompt-engineering]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Wed, 11 Mar 2026 05:01:00 GMT</pubDate>
            <atom:updated>2026-03-11T05:01:00.635Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Stop using your AI coding tool like a junior dev. Here’s how the pros actually do it.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4FsfN9kSSy8gUgvIkFHbBA.png" /></figure><p>If you’ve been using Cursor the same way you used VS Code — just typing code and occasionally asking it to autocomplete — you’re leaving an enormous amount of productivity on the table.</p><p>Cursor isn’t just an IDE with a fancy autocomplete. It’s a full AI-powered development partner that, when used correctly, can 10x your output, help you ship faster, debug smarter, and write cleaner code than you ever could alone.</p><p>This is the guide I wish I had when I started. Let’s go deep.</p><h3>What Makes Cursor Different From Everyone Else</h3><p>Before we get into the how, let’s be clear about the why.</p><p>Most AI coding tools bolt an LLM onto the side of your workflow. You write code, then ask the AI a question in a separate panel. It’s still fundamentally a human-in-the-loop, manual process.</p><p>Cursor flips this. The AI is woven <em>into</em> every layer of the IDE — autocomplete, chat, multi-file editing (Composer), terminal, and even code review. When you learn to work <em>with</em> it instead of just querying it, it becomes less of a tool and more of a second brain.</p><h3>1. Master the Three Modes of Interaction</h3><p>Cursor has three distinct ways to interact with the AI. Most developers only use one. Legends use all three fluidly.</p><h3>Chat (Cmd/Ctrl + L)</h3><p>This is your thinking partner. Use it when you need to understand something, plan an architecture, or debug a tricky issue. Keep the chat context focused — don’t dump your entire codebase into one conversation. Start fresh chats for different tasks.</p><p><strong>Pro move:</strong> Ask it to explain <em>why</em> something is failing before asking for a fix. Understanding the root cause makes you a better developer, not just a faster one.</p><h3>Inline Edit (Cmd/Ctrl + K)</h3><p>This is for targeted, surgical edits. Highlight a function, press Ctrl+K, and tell it exactly what to change. This is faster than Chat for small, specific modifications.</p><p><strong>Pro move:</strong> Use natural language like “refactor this to use async/await” or “add proper error handling here.” Cursor understands intent, not just instructions.</p><h3>Composer (Cmd/Ctrl + I)</h3><p>This is the nuclear option. Composer works across multiple files simultaneously. When you need to add a new feature that touches your backend, frontend, and tests all at once — this is your weapon.</p><p><strong>Pro move:</strong> Give Composer a full spec before asking it to code. Don’t say “add user authentication.” Say “add JWT-based user authentication with refresh tokens, error handling, and update the existing user model, auth middleware, and login route.”</p><h3>2. Context is King — Learn to Control It</h3><p>The quality of Cursor’s output is directly proportional to the quality of context you give it. This is the single most important thing to understand.</p><h3>Use @-mentions like a pro</h3><ul><li>@file — reference a specific file</li><li>@folder — bring in an entire directory</li><li>@web — pull in live web documentation</li><li>@docs — reference indexed documentation</li><li>@codebase — search across your entire project</li></ul><p><strong>Example:</strong> Instead of pasting code and saying “fix this,” try:</p><blockquote><em>“@auth/middleware.ts has a bug where expired tokens aren’t being caught. The expected behavior is described in @docs/authentication.md — fix it and update the @tests/auth.test.ts accordingly.”</em></blockquote><p>That single prompt gives Cursor everything it needs to do an expert-level job.</p><h3>Use .cursorrules to set context permanently</h3><p>Create a .cursorrules file in your project root. This is your persistent system prompt for that project. Include:</p><ul><li>The tech stack (e.g., “This is a Next.js 14 app with TypeScript, Prisma, and Tailwind”)</li><li>Coding conventions (“Always use functional components, never class components”)</li><li>Business logic rules (“Never expose user emails in API responses”)</li><li>Testing requirements (“All new functions should have unit tests using Jest”)</li></ul><p>Once set, every single interaction in that project benefits from this context — without you typing it each time.</p><h3>3. Composer: The Multi-File Superpower</h3><p>If you’re not using Composer regularly, you’re working too hard.</p><h3>When to use Composer</h3><ul><li>Adding a new feature end-to-end</li><li>Refactoring a module across multiple files</li><li>Implementing a design pattern project-wide</li><li>Creating a new CRUD entity (model + service + controller + tests)</li></ul><h3>How to write a great Composer prompt</h3><p>Bad: <em>“Add a dark mode”</em></p><p>Good: <em>“Add dark mode support using Tailwind’s dark class strategy. Update the ThemeProvider in @context/theme.tsx, add a toggle component in @components/ui/ThemeToggle.tsx, persist the preference in localStorage, and update the root layout in @app/layout.tsx to apply the dark class.”</em></p><p>The difference is specificity. Treat Composer like a new senior developer joining your team — give them a proper brief, not a vague request.</p><h3>Review before you accept</h3><p>Always review Composer’s diffs before accepting. It’s smart, but it doesn’t know your business logic. A two-minute review can save a two-hour debugging session.</p><h3>4. Debugging Like a 10x Developer</h3><p>Cursor transforms debugging from a painful solo activity into a collaborative investigation.</p><h3>The debugging loop</h3><ol><li><strong>Paste the error</strong> into Chat with the full stack trace</li><li><strong>Add context:</strong> @file the relevant files, mention what you were trying to do</li><li><strong>Ask for the cause first</strong>, not the fix</li><li><strong>Verify the fix makes sense</strong> before applying it</li></ol><h3>Terminal integration</h3><p>Cursor can read your terminal output. When a build fails, instead of copying the error manually, just say “look at the terminal output and tell me what went wrong.” It saves micro-friction that adds up over a day of coding.</p><h3>The rubber duck upgrade</h3><p>Old school: explain your bug to a rubber duck until you find it yourself. Cursor version: explain your bug to Cursor and let it find it for you while you explain. Say “I’m going to explain this bug out loud — tell me when I say something that contradicts what the code does.”</p><h3>5. Model Selection: Stop Wasting Money on the Wrong Model</h3><p>Not all tasks need your most powerful (and most expensive) model. Choosing wisely keeps your usage quota healthy and your responses faster.</p><h3>The smart model strategy</h3><p><strong>Claude Sonnet 4.5 — Your daily driver (90% of the time)</strong></p><ul><li>Perfect for: writing code, fixing bugs, refactoring, generating boilerplate</li><li>Fast, smart, and cost-efficient</li><li>This should be your default in Cursor settings</li></ul><p><strong>Claude Opus — Reserve for hard problems (10% of the time)</strong></p><ul><li>Perfect for: complex architectural decisions, tricky multi-step debugging, understanding deeply unfamiliar code</li><li>Burns through your quota 3–5x faster</li><li>Only reach for it when Sonnet genuinely struggles</li></ul><p><strong>Haiku — Almost never for serious coding</strong></p><ul><li>Good for: quick lookups, simple autocomplete, trivial boilerplate</li><li>Misses edge cases and writes buggy code under complexity</li><li>Costs you more time fixing mistakes than tokens you saved</li></ul><p><strong>Rule of thumb:</strong> Start every session on Sonnet. Switch to Opus only if you’ve gone back and forth three times and still aren’t getting the result you need.</p><h3>6. Workflow Patterns That Separate Legends From Amateurs</h3><h3>The Spec-First Pattern</h3><p>Before writing a single line of code, write your spec in plain English as a comment or a markdown file. Use Cursor to review and poke holes in it. Only then start implementing. This prevents the most expensive mistake in software: building the wrong thing correctly.</p><h3>The Test-First Pattern</h3><p>Ask Cursor to write tests before writing implementation. Say: “Based on this function signature and description, write the tests first. Then I’ll ask you to implement the function.” This forces clarity on expected behavior and leads to cleaner code.</p><h3>The Incremental Review Pattern</h3><p>For large Composer outputs, don’t accept everything at once. Accept file by file, test as you go. It’s slower, but you maintain understanding of your own codebase — which is the whole point.</p><h3>The Second Opinion Pattern</h3><p>After Cursor gives you a solution, ask: “What are the edge cases this implementation misses? What could go wrong?” This turns a good solution into a great one.</p><h3>7. Keyboard Shortcuts That Will Change Your Life</h3><p>Action Mac Windows/Linux Open Chat Cmd + L Ctrl + L Inline Edit Cmd + K Ctrl + K Open Composer Cmd + I Ctrl + I Accept suggestion Tab Tab Reject suggestion Esc Esc New chat Cmd + Shift + L Ctrl + Shift + L Toggle sidebar Cmd + B Ctrl + B</p><p>Memorize these. Every click you eliminate compounds into hours saved per month.</p><h3>Tips &amp; Tricks: The Insider Knowledge Section</h3><p>This is where it gets interesting. Beyond the official documentation, here’s what experienced Cursor users actually do.</p><h3>💡 Productivity Tips</h3><p><strong>Pin your best prompts</strong> Keep a markdown file called prompts.md in your project root with your most effective prompts for that project. Copy-paste is faster than rewriting from scratch every time.</p><p><strong>Use Cursor for code reviews</strong> Before pushing a PR, open Composer and say: “Review all the files I’ve changed today for bugs, security issues, and code style inconsistencies.” It’s like having a senior engineer review your code before it goes out.</p><p><strong>Teach it your patterns</strong> If you find yourself correcting the same mistake repeatedly, add it to .cursorrules. The AI should learn your preferences — make it official.</p><p><strong>Use web search for new libraries</strong> When working with a library released after the model’s training cutoff, use @web to pull in current documentation. This prevents the AI from hallucinating outdated APIs.</p><p><strong>Break big tasks into steps</strong> Instead of “build me a full authentication system,” say “step 1: create the user model. Step 2: create the auth service. Step 3: create the route handlers.” Smaller scoped tasks = fewer errors = less cleanup.</p><h3>💰 Money-Saving Tips</h3><p><strong>Switch to Sonnet as your default model</strong> As covered above, Sonnet handles 90% of coding tasks at a fraction of the cost of Opus. In Cursor settings, set Sonnet as your default and your token budget will last significantly longer.</p><p><strong>Disable models you never use</strong> Go to Cursor Settings → Models and disable all models you don’t actively use. If you keep GPT-4o, Gemini, and multiple Claude versions all enabled, you’re adding unnecessary complexity and temptation to over-use expensive models.</p><p><strong>Start fresh chats for new tasks</strong> Long conversations accumulate context tokens fast. Every message re-processes the entire conversation history. For a new task, open a new chat. Your wallet will thank you.</p><h3>🌍 Billing Tip: Use a UAE Address</h3><p>If you’re signing up for Cursor from India, consider using a <strong>UAE billing address</strong> instead. The real saving isn’t about currency conversion — it’s about <strong>tax</strong>. In India, you pay <strong>18% GST</strong> on top of your subscription cost. Every country has different tax rules, but the UAE has <strong>0% VAT on digital services</strong>, which means you’re paying the base price and nothing more. On a $20/month plan, that’s effectively $3.60 saved every single month just by changing your billing address.</p><h3>🎯 The 50% Discount Hack</h3><p>Here’s a little-known trick that many Cursor power users swear by:</p><p>After using Cursor Pro for <strong>3 months</strong>, cancel your auto-renew subscription before the next billing cycle. Don’t fully delete your account — just cancel the auto-debit.</p><p>When Cursor detects you’re about to churn, they will typically offer you a <strong>50% discount for the next month</strong> to win you back. This is a standard retention tactic used by many SaaS products.</p><p>Accept the discount, re-subscribe at half price, use it for another month, and you can repeat the cycle. Not everyone gets this offer, but the conversion rate among active users who cancel is high enough that it’s worth trying every few months.</p><p><strong>Steps:</strong></p><ol><li>Go to Settings → Subscription</li><li>Cancel auto-renewal (do NOT delete your account)</li><li>Before the cancel popup will open offering a 50% discount</li><li>Accept and re-subscribe</li></ol><h3>Final Thoughts</h3><p>The developers who get the most out of Cursor aren’t the ones who type the most — they’re the ones who think the most clearly about what they want before touching the keyboard.</p><p>Write better prompts. Use context deliberately. Pick the right model for the right job. And treat Cursor as a collaborator, not a search engine.</p><p>The legend developer workflow isn’t about being superhuman. It’s about being a brilliant manager of a very capable AI that never gets tired, never judges your questions, and is always ready to go deeper.</p><p>Now go ship something great.</p><blockquote><em>Stay tuned for the next part, where I’ll explain how to efficiently use </em><strong><em>Plan, Agent, and Debug modes</em></strong><em> in Cursor 🚀</em></blockquote><p><em>Thanks for reading, Please show your appreciation for the article by giving it a round of applause 👏</em></p><p><em>Follow Us: </em><a href="https://www.linkedin.com/in/rb786/"><strong><em>LinkedIn</em></strong></a><strong><em> | </em></strong><a href="https://twitter.com/rahulatindia"><strong><em>Twitter</em></strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3ad0ac51acc0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Salesforce Integration with Node.js: Building Powerful Customer Experiences.]]></title>
            <link>https://rahulh123.medium.com/salesforce-integration-with-node-js-building-powerful-customer-experiences-f82b8575bca8?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/f82b8575bca8</guid>
            <category><![CDATA[salesforce]]></category>
            <category><![CDATA[nodejs]]></category>
            <category><![CDATA[oauth2]]></category>
            <category><![CDATA[integration]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Sun, 22 Dec 2024 16:02:29 GMT</pubDate>
            <atom:updated>2024-12-22T16:02:29.835Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_328pTAWPyK1dSnC5IIlaA.png" /></figure><h3>​</h3><p>Salesforce, a leading CRM platform, and Node.js, a powerful JavaScript runtime, form a dynamic duo for creating robust, scalable, and efficient enterprise applications. This blog explores the benefits, setup, and implementation of integrating these two powerful technologies.</p><p>&quot;We will learn how to access the details of Salesforce users from another organization.&quot;</p><blockquote>prerequisite</blockquote><blockquote>install node.js and npm [<a href="https://nodejs.org/en">Link</a>]</blockquote><blockquote>create a salesforce account [<a href="https://login.salesforce.com/">Link</a>]</blockquote><blockquote>create developer app in saleforce</blockquote><h3>Creating a Salesforce Connected App</h3><p>To create a connected app in Salesforce, follow these steps:</p><p><strong>Log in to Salesforce</strong>: Navigate to <a href="https://login.salesforce.com/">Salesforce</a> and log in with your credentials.</p><p><strong>Navigate to App Manager</strong>: In the top right corner, click on the gear icon (Setup) and select “Setup”. In the Quick Find box, type “App Manager” and click on “App Manager”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*apkXhp-LbS7F6uROlbo1Ig.png" /></figure><p><strong>Create a New Connected App</strong>: Click the “New Connected App” button at the top right.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4bkgtm8LUY3GStHRJr8dRg.png" /></figure><p><strong>Configure Basic Information</strong>: Fill in the basic information such as Connected App Name, API Name, and Contact Email.</p><p><strong>Enable OAuth Settings</strong>: Check “Enable OAuth Settings”. Enter the Callback URL and select the OAuth Scopes required (e.g., Full access, Perform requests on your behalf).</p><p><strong>Save and Continue</strong>: Click “Save”. You may need to wait a few minutes for the changes to take effect.</p><p><strong>Review and Note Details</strong>: After saving, make sure to note the Consumer Key and Consumer Secret from the app details page, as these will be required for the integration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cFD1MnschPY_32rKNJ5F8g.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*swD0qO8ChOLE0wAJIkhLiQ.png" /></figure><blockquote>Full Documentation and Resources:</blockquote><blockquote>For detailed steps and additional configuration options, refer to the <a href="https://developer.salesforce.com/docs/atlas.en-us.api_rest.meta/api_rest/intro_defining_remote_applications.htm">Salesforce Connected Apps documentation</a>.</blockquote><h3>install express &amp; jsforce library to connect salesforce with node.js</h3><pre>npm i express<br><br>npm i jsforce //Salesforce API Library for JavaScript applications <br>            </pre><pre>//jsForce connection<br>const oauth2 = new jsforce.OAuth2({<br>    // you can change loginUrl to connect to sandbox or prerelease env.<br>    loginUrl : &#39;https://login.salesforce.com&#39;,<br>    //clientId and Secret will be provided when you create a new connected app in your SF developer account<br>    clientId : &#39;Consumer Key&#39;,<br>    clientSecret : &#39;Consumer Secret&#39;,<br>    // make sure redirect url same as declared in saleforce connected app <br>    redirectUri : &#39;http://localhost:3030/callback&#39; <br>});</pre><p>The above sample function allows the connection between the node app and Salesforce and gives access token</p><blockquote>Full Source Code: Available on the <a href="https://github.com/numerica-ideas/community/tree/master/salesforce">GitHub Repository</a> 📌</blockquote><h3>Benefits of Salesforce and Node.js integration</h3><ul><li><strong>Increased Efficiency</strong>: By automating tasks and workflows, you can save time and effort.</li><li><strong>Streamlined Workflows</strong>: Leverage Salesforce automation and Node.js flexibility to create tailored workflows that optimize business processes.</li><li><strong>Enhanced Scalability</strong>: Node.js is known for its scalability, which can handle the growing demands of your business.</li><li><strong>Real-time Data Access</strong>: Access real-time data from Salesforce within your Node.js applications.</li><li><strong>Custom Functionality</strong>: Build custom functionality that is not available in Salesforce out-of-the-box.</li></ul><p>moreover, you can manage the below operations through the integration</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pIZDaJKT1-FKEninp5zo3A.png" /></figure><h3>Drawbacks of Salesforce and Node.js Integration</h3><ul><li><strong>Complexity:</strong> Integration can be complex, requiring developers with expertise in both Salesforce and Node.js.</li><li><strong>Security Concerns:</strong> Establishing and maintaining secure connections and data access between the two platforms requires careful planning and implementation.</li><li><strong>Cost:</strong> Salesforce licensing fees and the potential need for specialized Node.js development resources can add to your overall costs [<a href="https://www.salesforce.com/in/products/sales-pricing">Link</a>].</li><li><strong>Vendor Lock-in:</strong> You become reliant on Salesforce and the capabilities of the jsforce library, which may limit your flexibility in the future.</li></ul><h3>Conclusion</h3><p>The combination of Salesforce and Node.js is a powerful tool for businesses looking to build scalable and efficient applications. By integrating these two platforms, you can streamline workflows, improve data access, and ultimately enhance the customer experience.</p><blockquote><strong>Reference Links</strong></blockquote><blockquote>Saleforce Api Library : <a href="https://developer.salesforce.com/docs/apis">https://developer.salesforce.com/docs/apis</a>​</blockquote><blockquote>Salesforce Api collection : <a href="https://www.postman.com/salesforce-developers/workspace/salesforce-developers/collection/12721794-67cb9baa-e0da-4986-957e-88d8734647e2">Postman Collection</a>​</blockquote><blockquote>jsForce Documents : <a href="https://jsforce.github.io/document/">https://jsforce.github.io/document/#</a>​</blockquote><blockquote>Salesforce Pricing Models: <a href="https://www.salesforce.com/in/products/sales-pricing">https://www.salesforce.com/in/products/sales-pricing</a>​</blockquote><blockquote>Thanks for reading, Please show your appreciation for the article by giving it a round of applause 👏, and don’t forget to follow for more insightful knowledge in the future.</blockquote><blockquote>Follow Us: <a href="https://www.linkedin.com/in/rb786/"><strong>LinkedIn</strong></a><strong> | </strong><a href="https://twitter.com/rahulatindia"><strong>Twitter</strong></a></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f82b8575bca8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Master the Art of Debugging: Tips and Strategies to Level Up Your Software Development]]></title>
            <link>https://rahulh123.medium.com/master-the-art-of-debugging-tips-and-strategies-to-level-up-your-software-development-61330810959f?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/61330810959f</guid>
            <category><![CDATA[debugging-tips]]></category>
            <category><![CDATA[codecraft]]></category>
            <category><![CDATA[debugging]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Mon, 28 Aug 2023 05:01:04 GMT</pubDate>
            <atom:updated>2023-08-28T05:01:04.987Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ujKu-36phhPNGE2sWZVGfg.jpeg" /></figure><p>Debugging is a critical skill for any software developer. Learn how to debug like a pro with these tips and strategies.</p><h3>Common Debugging Mistakes</h3><p>Even the best developers can make serious debugging errors. Discover typical mistakes and ways to evade them, helping you save time and energy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eonzGWQHd6dMQsT71aFx0w.png" /></figure><h3>Developing a Debugging Mindset</h3><p>Debugging is not just about fixing errors — it’s a mindset that can help you write better code.</p><p>Learn how to adopt a proactive approach to debugging and make your code more robust.</p><h4>Embrace Failure</h4><p>Failure is a natural part of software development. Learn to embrace it as a learning experience and an opportunity to improve your skills.</p><h4>Think Like a User</h4><p>Put yourself in the user’s shoes. This can help you identify issues and improve user experience.</p><h4>Test Your Code</h4><p>Write unit tests to catch problems before they reach production. This will save you time and effort in the long run.</p><h3>Tools for Effective Debugging</h3><p>The right tools can make debugging faster, easier, and more efficient. Discover some essential tools for debugging and how to use them to their fullest potential Regardless of specific technology here I have a list of general tools &amp; techniques that help in debugging better.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kOIfTH0q3jPWM3Rg5jx6jQ.png" /></figure><h3>Debugging Strategies</h3><p>Debugging is a process, not a one-time fix. Learn some effective strategies for managing complex debugging scenarios and improving your efficiency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YPpml9hzFPSrpzgNHLacGQ.png" /></figure><h3>Collaborative Debugging</h3><p>Debugging is often a collaborative effort. Learn how to work effectively with others to find and fix bugs and improve code quality.</p><blockquote>“Two heads are better than one” — by working together, you can draw on each other’s strengths and identify bugs more quickly and accurately.</blockquote><h4>Pair Programming</h4><p>Work with a partner to catch bugs in real-time and share knowledge.</p><h4>Collaborative Whiteboards</h4><p>Use online whiteboards like <a href="https://miro.com/">Miro</a> and Google <a href="https://jamboard.google.com/?pli=1">Jamboard</a> to brainstorm and track progress during debugging sessions.</p><p>By working together, you can draw on each other’s strengths and identify bugs more quickly and accurately.</p><h3>Conclusion</h3><p>Debugging is an essential skill for software developers. By adopting the right mindset, using the right tools, and applying effective strategies, you can become a debugging pro and improve your code quality and productivity.</p><blockquote>Coming Soon: Blog Part-2 — Mastering Debugging: Enhancing Your Software Development Skills. Stay Tuned!</blockquote><blockquote><em>Please show your appreciation for the article by giving it a round of applause 👏, and don’t forget to follow for more insightful stories in the future.</em></blockquote><blockquote><em>Follow Me: </em><a href="https://www.linkedin.com/in/rb786/"><strong><em>LinkedIn</em></strong></a></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=61330810959f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Choosing the Right Architecture: Monolithic vs. Microservices — Analyzing Requirements for Success]]></title>
            <link>https://rahulh123.medium.com/choosing-the-right-architecture-monolithic-vs-microservices-analyzing-requirements-for-success-70d681f6a1d0?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/70d681f6a1d0</guid>
            <category><![CDATA[microservices]]></category>
            <category><![CDATA[monolith]]></category>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[migration]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Mon, 05 Jun 2023 05:11:48 GMT</pubDate>
            <atom:updated>2023-06-05T05:11:48.739Z</atom:updated>
            <content:encoded><![CDATA[<h3>Choosing the Right Architecture: Monolithic vs. Microservices — Analyzing Requirements for Success</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dyJWZ5K0amK6ll5wtpJNPA.jpeg" /></figure><p>In the world of software development, architectural choices play a crucial role in determining the success of a project. One of the most prominent debates in recent years has been the comparison between monolithic and microservices architectures. While monolithic architecture has been the traditional approach, microservices have gained popularity due to their flexibility and scalability.</p><p>When I started my career back in 2016, The first thing I learned was about Microservices. It was considered the future of software development, promising greater scalability, flexibility, and resilience. It seemed like everyone was embracing it, even startups, despite the challenges involved.</p><p>A strong monolith structure will allow you to replace any segment of the application with microservice when needed. Most of the successful startups we read about (Netflix, Uber, Airbnb, etc.), all started with a monolith architecture.</p><p>Before diving into microservices, it’s important to consider the specific outcomes you hope to achieve. Ask yourself questions such as:</p><h3>Why do you need microservices?</h3><ul><li>Is there anything within the system that is scaling at a different rate than the rest of the system?</li><li>Is there a part of the system that requires more frequent deployments than the rest of the system?</li><li>Is there a part of the system that a single person, or a small team, that operates independently from the rest of the group?</li></ul><p>Once you have clear answers to these questions, you can analyze the costs and benefits to decide if using microservices is really necessary for you.</p><p>Everything comes with a price, sometimes people forget the cost you need to pay when pursuing new trends in technology. Some typical costs include:</p><ul><li>Increased development complexity</li><li>Exponential infrastructure costs</li><li>Added organizational overhead</li><li>Debugging challenges</li></ul><h3>If you are not sure about requirements, always start with Monolith first</h3><ol><li>When starting a new application, how confident are you that it will be valuable to your users? The most effective way to determine the usefulness of a software idea is by building a basic version and assessing its performance. In this initial phase, it is important to prioritize speed (and therefore quick feedback cycles), which means avoiding the complexities of microservices.</li><li>For Microservices to function effectively, it is crucial to establish clear and stable boundaries between the services. However, even experienced architects face challenges in determining these boundaries accurately from the start, particularly in familiar domains. By initially building a monolith, you can identify the appropriate boundaries before transitioning to a microservices architecture that may complicate the process.</li></ol><blockquote>In conclusion, no architecture is often the best architecture in the early days of a system.</blockquote><h3>Advantages of Microservice Architectures</h3><ul><li>Agile and flexible</li><li>Fewer and more isolated code issues</li><li>Smaller in size and scope</li><li>Independently stacked and deployed</li><li>Deployment is rapid and continuous</li><li>Functional with cloud-based containers</li><li>Best in class solutions and performance for every use case</li><li>Low cohesion, <em>(the advantage is that any change you make has a drastically reduced risk compared to monoliths, which are very cohesive but cannot be edited without high-risk)</em></li><li>Smaller domains</li><li>Lack of transactional requirements in regards to cross-services</li></ul><h3>Disadvantages of Microservice Architectures</h3><ul><li>Can be challenging to maintain separate frameworks</li><li>Lower/no cohesion between platforms (it is important to keep platforms separate without strong connections. Starting with a monolithic approach helps identify and resolve any issues with platform connections before moving to a microservices setup )</li><li>Data stored in technology and operational silos</li><li>Can lead to a disjointed experience for your customers</li><li>Can be more expensive, depending on the configuration</li><li>More complicated</li></ul><h3>3 Microservice Case Study — Amazon, Netflix, and Uber</h3><p>we’ll explore the microservices journeys of these wildly successful enterprises. But first, let’s look at the general circumstances that inspire enterprises to use microservices in the first place</p><ul><li>Microservices are a way of constructing applications by dividing them into several independent, small services.</li><li>In Java, microservices are a software architecture pattern where applications are created as a collection of these small, autonomous services.</li><li>Prominent companies like Amazon, Netflix, and Uber have embraced microservices to gain advantages in scalability, flexibility, and profitability.</li><li>The benefits of microservices include agility, adaptability to evolving business needs, scalability, and the ability to easily integrate new components.</li></ul><h3>Examples of Microservices in Action</h3><h3>1. Netflix</h3><p>Netflix is a company that has found success through the use of microservices connected with APIs.</p><p>Netflix began its microservice journey in 2008 before the term “microservices” became popular. Netflix started its movie-streaming service in <a href="https://media.thinknum.com/articles/a-brief-history-of-video-streaming-by-the-numbers/">2007</a>.</p><p>By 2008 it was suffering from service outages and scaling challenges, for three days, it was unable to ship DVDs to members.</p><p>At this point, the company was still dealing with physical DVDs, which put a damper on how well it could serve its customers. Streaming was still a dream, and online business, while thriving, was difficult. The monolith design was still not very functional beyond a certain point. Microservices architecture was a much better option, but it didn’t truly exist yet.</p><p>According to a <a href="https://about.netflix.com/en/news/completing-the-netflix-cloud-migration">Netflix vice president</a>:</p><p>Our journey to the cloud at Netflix began in August of 2008, when we experienced a major database corruption and for three days could not ship DVDs to our members. That is when we realized that we had to move away from vertically scaled single points of failure, like relational databases in our data center, towards highly reliable, horizontally scalable, distributed systems in the cloud. We chose Amazon Web Services (AWS) as our cloud provider because it provided us with the greatest scale and the broadest set of services and features.</p><p>Starting in 2009, Netflix initiated a step-by-step transformation of its monolithic architecture into microservices. They began by migrating their non-customer-facing movie-coding platform to operate independently as a microservice on Amazon AWS cloud servers. Over the course of the next two years, Netflix proceeded to convert their customer-facing systems into microservices, completing the transition in 2012.</p><p>The first step was to migrate its non-customer-facing, movie-coding platform to run on Amazon AWS cloud servers as an independent microservice. Netflix spent the following two years converting its customer-facing systems to microservices, finalizing the process in 2012.</p><h3>ROI</h3><p>Refactoring to microservices allowed Netflix to overcome its scaling challenges and service outages. By 2013, Netflix’s API gateway was handling <a href="https://netflixtechblog.com/optimizing-the-netflix-api-5c9ac715cf19">two billion daily API edge requests</a>, managed by over 500 cloud-hosted microservices.</p><p>By 2017, its architecture consisted of <a href="https://medium.com/refraction-tech-everything/how-netflix-works-the-hugely-simplified-complex-stuff-that-happens-every-time-you-hit-play-3a40c9be254b">over 700 </a>loosely coupled microservices. Today, Netflix makes around <a href="https://www.comparitech.com/blog/vpn-privacy/netflix-statistics-facts-figures/"><strong>$8 billion a year</strong></a> and streams approximately six billion hours of content weekly to more than 220 million subscribers in 190 countries, and it continues to grow</p><p>Here’s a visual depiction of Netflix’s growth from 2007 to 2015:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/546/1*a-_8lotuyeKjtPa2HVYu0g.png" /><figcaption>Netflix</figcaption></figure><h3>2. Amazon</h3><p>Amazon is an Internet retail giant, but it didn’t start that way. In the early 2000s, Amazon’s retail website behaved like a single monolithic application.</p><p>The tight connections between — and within — the multi-tiered services that comprised Amazon’s monolith meant that developers had to carefully untangle dependencies every time they wanted to upgrade or scale Amazon’s systems. It was a painstaking process that cost plenty of money and required time to adjust.</p><p>Previously, Amazon found that the monolith structure worked very well. However, the code base quickly expanded as more developers joined the team. There were too many updates and projects coming in, which had a negative impact on the software development and function. With such obvious drops in productivity, it was necessary to look at a better way of doing things.</p><p>In 2001, development delays, coding challenges, and service interdependencies forbid Amazon’s ability to meet the scaling requirements of its rapidly growing customer base. The company and its site were exploding, but there was no way to keep up with the growth.</p><h3>Here’s how Amazon did it:</h3><p>Developers analyzed the source code and pulled out units of code that served a single, functional purpose. As expected, this was a time-consuming process since all the code was mixed up together, and the various functions of the site were intertwined. However, the developers worked hard to sort it out and determine which sections could be turned into microservices.</p><p>Once they had the separate sections of code, the developers wrapped these units in a web service interface. For example, they developed a single service for the Buy button on a product page, a single service for the tax calculator function, and so on. Each function had its own section.</p><h3>ROI</h3><p>Amazon’s “service-oriented architecture” was largely the beginning of what we now call microservices. It led to Amazon developing a number of solutions to support microservices architectures — such as Amazon AWS and Apollo — which it currently sells to enterprises throughout the world. Without its transition to microservices, Amazon could not have grown to become the most valuable company in the world — valued by market cap at $1.27 trillion as of June 2, 2023.</p><h3>3. UBER</h3><p><strong>Uber</strong>, the world’s largest taxi company, owns no vehicles. isn’t it interesting, But how UBER achieve this lets understand</p><p>Despite being introduced to the world more recently than either of our previous examples, Uber also began with a monolith design.</p><p>This microservice example came not long after the launch of Uber, when the ride-sharing service encountered <a href="https://eng.uber.com/service-oriented-architecture/">growth hurdles </a>related to its monolithic application structure.</p><p>The platform struggled to develop and launch new features efficiently, fix bugs, and integrate its rapidly growing global operations. Moreover, the complexity of Uber’s monolithic application architecture requires developers to have extensive experience working with the existing system just to make minor updates and changes to the system</p><h3>Here’s how Uber’s monolithic structure worked at the time:</h3><ul><li>Passengers and drivers connected to Uber’s monolith through a REST API.</li><li>There were three adapters — with embedded API for functions like billing, payment, and text messages.</li><li>There was a MySQL database.</li><li>All features were contained in the monolith.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*R7DpYNfiEvds6p3uXZQqDQ.jpeg" /></figure><p>The above diagram provided illustrates the former architecture employed by UBER.</p><p>To overcome the challenges of its existing application structure, Uber decided to break the monolith into cloud-based microservices. Subsequently, developers built individual microservices for functions like passenger management, trip management, and more. Similar to the Netflix example above, Uber connected its microservices via an API gateway.</p><p>The changes were faster in this case, since it was done earlier in the business. There were fewer functions mixed into the monolith design, and this made it simpler to change things around when it was time to switch.</p><p>Refer to the diagram below to look at UBER’s microservice architecture.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*AMNKOeSAuNZxN90F_As41g.jpeg" /></figure><ul><li>The major change that we observe here is the introduction of API Gateway through which all the drivers and passengers are connected. From the API Gateway, all the internal points are connected such as passenger management, driver management, trip management, and others.</li><li>The units are individual separate deployable units performing separate functionalities.</li></ul><blockquote><strong>Conclusion</strong></blockquote><ul><li>Monoliths are easier to develop, test, and deploy but are difficult to maintain, enhance, and scale.</li><li>Monoliths may suffer from a single point of failure.</li><li>Microservices require specialized skills and more resources to develop, but they are easier to debug, maintain, enhance, and scale.</li><li>Microservices do not suffer from a single point of failure</li></ul><blockquote><em>Please show your appreciation for the article by giving it a round of applause 👏, and don’t forget to follow for more insightful stories in the future.</em></blockquote><blockquote><em>Follow Me: </em><a href="https://www.linkedin.com/in/rb786/"><strong><em>LinkedIn</em></strong></a></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=70d681f6a1d0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Implement Redis with Node.js — Boosting Performance and Scalability of Application]]></title>
            <link>https://rahulh123.medium.com/implement-redis-with-node-js-boosting-performance-and-scalability-of-application-8b0206579727?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/8b0206579727</guid>
            <category><![CDATA[servers]]></category>
            <category><![CDATA[redis]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[caching]]></category>
            <category><![CDATA[nodejs]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Sun, 21 May 2023 04:42:15 GMT</pubDate>
            <atom:updated>2023-05-21T04:42:15.116Z</atom:updated>
            <content:encoded><![CDATA[<h3>Implement Redis with Node.js — Boosting Performance and Scalability of Application</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t1INRfB0mEByrPZ1UZNhCw.jpeg" /></figure><p>In this article, we will delve into the implementation of Redis within a Node.js application. However, before we proceed with the implementation, let’s take a closer look at what Redis is and how it can enhance the performance of our Node.js application.</p><blockquote>What is Redis ?</blockquote><p>Redis is often referred to as a “data structure server” because it allows you to store and manipulate various data structures such as strings, lists, sets, sorted sets, and hashes. Its in-memory nature makes it extremely fast, making it an ideal choice for caching, real-time analytics, messaging systems, and much more.</p><p>Redis is commonly used for <strong>caching</strong> frequently accessed data, improving application performance, and enabling real-time communication between different parts of an application or different applications altogether.</p><h3>Before we explore Redis and its impact on Node.js application performance, let’s briefly understand caching.</h3><p>Caching is the process of storing copies of files in a cache or a temporary storage location so that they can be accessed more quickly.</p><p>Now that we understand what caching is, let’s explore the reason for caching.</p><p>Below are a few of the reasons we cache:</p><ul><li>To save cost. Such as paying for bandwidth or even the volume of data sent over the network.</li><li>To reduce app response time.</li></ul><blockquote>In short,Optimally implementing caching can enhance our application’s performance and result in a mutually beneficial outcome for both us and our users.</blockquote><h3>Prerequisites</h3><p>To follow through this tutorial you must have the following installed on your computer</p><ul><li>Node JS <a href="https://nodejs.org/en">📥</a></li><li>NPM</li><li>Postman/Web Browser<a href="https://www.postman.com/">📥</a></li><li>Code Editor (VsCode / Sublime etc.)</li></ul><p>Let’s explore the practical implications of using Redis and understand the performance differences it can offer.</p><p>First, install redis-server in your machine, by this command</p><pre>sudo apt install redis-servers</pre><p>When you execute the command mentioned previously, it enables the Redis server on a Linux distribution based on Debian, such as Ubuntu. This allows you to utilize Redis on your local machine.</p><blockquote>You can find a comprehensive guide on how to install Redis on Ubuntu by following the link provided: <a href="https://www.digitalocean.com/community/tutorials/how-to-install-and-secure-redis-on-ubuntu-18-04">Redis Server Installation Guide</a></blockquote><p>After that, you can run this command to start redis server locally</p><pre>redis-server</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*22RAJUoiGulFOOPIkCuLSQ.png" /><figcaption>Running Redis server</figcaption></figure><p>Let’s Continue with the demo and install the following dependencies:</p><pre>npm install express axios</pre><ul><li><a href="https://www.npmjs.com/package/express">Express</a>: is a minimal and flexible Node web application framework that provides a robust set of features for web and mobile applications.</li><li><a href="https://www.npmjs.com/package/axios">Axios</a>: is a simple, promise-based HTTP client for the browser and Node</li></ul><p>Next, install Redis using npm</p><pre>npm install redis</pre><p>install the redis module, which allows us to interact with Redis in our Node.js application</p><p>Now Create app.js file in the folder Add the below code to create redis connection</p><pre>const express = require(&quot;express&quot;);<br>const axios = require(&quot;axios&quot;);<br>const redis = require(&quot;redis&quot;);<br><br>const app = express();<br><br>//Redis config<br>const redisconfig = {<br>  host: &quot;localhost&quot;, //host url incase if you are using 3rd party service like Aws <br>  port: &quot;6379&quot;, //post where yout redis server will start<br>  pass: &quot;&quot;, //password if you have set any default is empty<br>};<br><br>//create client for use redis<br>const client = redis.createClient(redisconfig);<br><br>//wait for connect client<br>(async () =&gt; {<br>  await client.connect();<br>})();<br><br>client.on(&quot;error&quot;, (err) =&gt; {<br>  console.error(&quot;Redis connection error&quot;, err);<br>});<br><br>client.on(&quot;connect&quot;, () =&gt; {<br>  console.log(&quot;Redis Server Connected!&quot;);<br>});<br><br>//start server on this port<br>const port = 3000;<br><br>app.listen(port, () =&gt; {<br>  console.log(`Serever started on port ${port}`);<br>});</pre><p>This code sets up an Express.js server and establishes a connection with a Redis server, preparing the server to handle incoming requests and utilize Redis for caching or other purposes.</p><p>Add Two Api with and without use of Redis to see the difference</p><pre>//dummy json api url from where porduct data will get<br><br>const PRODUCT_API = &quot;https://dummyjson.com/products&quot;;<br><br>//Get prouct list<br>app.get(&quot;/list-products&quot;, (req, res) =&gt; {<br>  try {<br>    axios.get(`${PRODUCT_API}`).then(function (response) {<br>      const products= response.data;<br>      console.log(&quot;Products retrieved from the API&quot;);<br>      res.status(200).send(products);<br>    });<br>  } catch (err) {<br>    res.status(500).send({ error: err.message });<br>  }<br>});<br><br>//Get Product list store in redis and display<br>app.get(&quot;/list-products-redis&quot;, async (req, res) =&gt; {<br>  try {<br>    <br>    //get data from redis<br>    const cacheProducts = await client.get(&quot;products&quot;);<br><br>    if(cacheProducts){<br><br>      console.log(&quot;Product retrieved from Redis&quot;);<br>      res.status(200).send(JSON.parse(cacheProducts));<br>   <br>    }<br>    else {<br>        axios.get(`${PRODUCT_API}`).then(function (response) {<br>          const products = response.data;<br><br>          //store data in redis<br>          client.set(&quot;products&quot;,JSON.stringify(products));<br><br>          console.log(&quot;Product retrieved from the API&quot;);<br>          res.status(200).send(products);<br>        });<br>    }<br>   <br>  } catch (err) {<br>    res.status(500).send({ error: err.message });<br>  }<br>});<br><br></pre><p>Here, we’re using the <a href="https://dummyjson.com/">DummyJSON</a> service to get data in API to work with. In our case, the API provides us products data.</p><p>Next, we have two requests: /list-productsand /list-products-redis</p><p>In the first one, the products are retrieved without caching the result. Whenever we send that request again, the products data will be retrieved a new.</p><p>In the second one, a check is made to see if the requested data is already stored in the cache. If it is, then the data is retrieved from <strong>Redis</strong>. Otherwise, if the products data isn’t stored in the cache, it will be first retrieved from the API call. In this case, the retrieved data will also be stored in the cache so that the next time it’s requested it will be retrieved faster.</p><p>We can perform the following test to prove how important caching is for performance.</p><p>Run node app in the terminal and visit the /list-products route in the browser or Postman.</p><p>In this demo, we are fetching <strong>100</strong> products data to measure performance</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tXddOVdHb7XunLa436k6aw.png" /><figcaption>Response with a simple API call</figcaption></figure><p>As we can see, the products data is successfully retrieved in <strong>875ms</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BnwoaTXD3gRHrCJ69X2PwQ.png" /></figure><p>Let’s Improve 🚀 performance now try with /list-products-redis route.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WtXrOh7URIPaArRDk_g2OA.png" /><figcaption>Response with Redis</figcaption></figure><p>The first time we send the request, it will give us approximately the same time 875ms as we received in the previous route because we don’t have the data stored in the cache yet, but when we send it again the result in time is drastically improved — only <strong>19ms</strong>. This is a huge difference even in this small and simple example. Imagine the performance gain with thousands of products. So, indeed, the caching is pretty impressive!</p><p><em>Note that, depending on your machine and connection speed, the time numbers you’ll get can be different from mine here, but the important thing is the ratio between cached and non-cached data, which will remain approximately the same.</em></p><p>Please note that this is only a small piece of work on how to cache the API result in Node JS with Redis. If you want to increase your knowledge about Redis, please read the <a href="https://redis.io/documentation"><strong><em>documentation</em></strong></a> of Redis.</p><p>If you want to access the whole code please check <a href="https://github.com/barotrahulh123/node-redis-demo/"><strong><em>my Github repo</em></strong> </a>and give me a start and fork it for using the code.</p><blockquote>To gain more insights into Redis, explore the top 5 use cases by clicking <a href="https://redis.com/blog/5-industry-use-cases-for-redis-developers/">here</a>.</blockquote><h3>Conclusion</h3><p>Redis is a very powerful in-memory data-store that we can use in our applications. It’s very simple to save and get data without much overhead. refer to <a href="https://www.npmjs.com/package/redis">https://www.npmjs.com/package/redis</a> for more use cases and refer to <a href="https://redis.io/commands">https://redis.io/commands</a> for more redis commands.</p><blockquote><em>Please show your appreciation for the article by giving it a round of applause 👏, and don’t forget to follow for more insightful stories in the future.</em></blockquote><blockquote><em>Follow Me: </em><a href="https://www.linkedin.com/in/rb786/"><strong><em>LinkedIn</em></strong></a></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8b0206579727" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Google I/O 2023 is a wrap — here’s a list you would like to know]]></title>
            <link>https://rahulh123.medium.com/google-i-o-2023-is-a-wrap-heres-a-list-you-would-like-to-know-9cf5bb2943e8?source=rss-5457b030f0d0------2</link>
            <guid isPermaLink="false">https://medium.com/p/9cf5bb2943e8</guid>
            <category><![CDATA[google]]></category>
            <category><![CDATA[google-io-2023]]></category>
            <category><![CDATA[google-pixel]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Rahul Bhrambhatt]]></dc:creator>
            <pubDate>Sun, 14 May 2023 05:34:30 GMT</pubDate>
            <atom:updated>2023-05-14T05:34:30.304Z</atom:updated>
            <content:encoded><![CDATA[<h3>Google I/O 2023 is a wrap — List of Highlights You Don’t Want to Miss</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0E6AJ3FE0MB6EfXRKJdaVQ.png" /><figcaption>credit: google</figcaption></figure><p><strong>What is Google I/O?</strong></p><blockquote>Google I/O is an annual developer conference hosted by Google, where the company showcases its latest technologies and products. The conference typically features keynote speeches, technical sessions, and product announcements, and attracts both developers and consumers</blockquote><p>We know you might not have time to watch a long talk, so here’s a short and easy-to-read list of the important things that were announced at Google I/O. It includes new features and updates that you might find useful</p><h3>Here are Key notes of Google I/O :</h3><h3><strong>1) Help Me Write</strong></h3><p>Help Me Write is a new feature coming to Gmail and Google Docs that will help you to compose emails and sections of writing based on more limited inputs that you provide.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M6sIU-qmLQLTv8tPfuHXWA.jpeg" /><figcaption>credit: google</figcaption></figure><p>The feature, dubbed “help me write,” is basically an expansion of the auto-replies and generative text that Google already uses in Gmail. While announcing the feature, Sundar Pichai, CEO of Google, used an example of asking an airline for a refund for a flight. The AI feature pulls info from previous exchanges with the airline and creates an entire message asking for a refund. You can then edit that message to your liking and send it away.</p><h3>2) PaLM 2</h3><p>PaLM 2, Google’s newest large language model (LLM).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*re6a1Knakiv4Y6CCow9zGA.jpeg" /><figcaption>credit: google</figcaption></figure><p>“PaLM 2 will power Google’s updated Bard chat tool, the company’s competitor to OpenAI’s ChatGPT, and function as the foundation model for most of the new AI features the company is announcing today.”</p><p>PaLM 2 also now features improved support for writing and debugging code. More here. Also, takes a deeper dive into <a href="https://ai.google/discover/palm2">PaLM 2</a> with a more critical look at the model through the lens of a Google</p><h3>3) Bard gets smarter</h3><p>Many of us are aware that Google had previously introduced Bard as a successor to Chat GPT, but it didn’t prove to be very effective. This time, however, Google has made significant improvements to Bard, making it even smarter and more advanced. Let’s take a closer look.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*VEW2erIqYNIgwjU-MAkNYA.jpeg" /></figure><p>Google has made Bard available to everyone in over 180 countries and territories and has removed the waitlist requirement. In addition to English, Bard now supports Japanese and Korean, and the plan is to expand to support 40 languages eventually.</p><p>Furthermore, Bard can now provide image results alongside its responses. Moreover, Google has teamed up with Adobe to offer art generation features in Bard. Through this collaboration, Bard users can use Firefly to create images and then customize them using Express. They can also select from various templates, fonts, stock images, and other assets available in the Express library, all within Bard.</p><p>Try it Brad from <a href="https://bard.google.com/">Here</a> and tell your experience</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/828/1*SmIS-eaxr4U3UnU5kIaFXA.gif" /></figure><h3>3) MusicLM</h3><p>Google released MusicLM, a new experimental AI tool that can turn text descriptions into music.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*Oj_NlKzzWHtK8pqcMN5EWg.jpeg" /></figure><p>Available in the AI Test Kitchen app on the web, Android or iOS, MusicLM lets users type in a prompt like “soulful jazz for a dinner party” or “create an industrial techno sound that is hypnotic” and have the tool create several versions of the song.</p><p>How To Try Google’s Text-To-Music Generator MusicLM ?</p><p>To sign up for MusicLM’s waitlist, go to <a href="https://aitestkitchen.withgoogle.com/experiments/music-lm">Google’s AI Test Kitchen.</a> Google will send you an email confirming your addition to the waitlist.</p><p>Once you’re approved, you can prompt MusicLM with a descriptive phrase like “ambient, soft sounding music I can study to.”</p><h3>4) Android is getting AI-powered customization options</h3><p>AI isn’t just coming for Google Search. Google has announced that it’s also bringing new AI-powered features to Android. One of these features, called Magic Compose, will live within Android’s Messages app and give you the ability to reply to texts using responses suggested by AI.</p><p>It’s also adding a feature that will let you use AI to create your own wallpaper. Instead of choosing from a set of template options, you’ll soon be able to describe an image, which your device will then generate using Google’s text-to-image diffusion model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BVOpp7jXgSTcUYVId2JYdw.gif" /><figcaption>credit: google</figcaption></figure><p>Additionally, Google is launching a Pixel-exclusive feature next month that lets you customize your device with a new “cinematic wallpaper” that adds depth to one of your existing photos</p><h3>5) Google’s latest Project Starline prototype uses AI to make 3D images of people</h3><p>Google announced the latest prototype of Project Starline, its 3D teleconferencing video booth. On the face of it, the biggest difference is a dramatic reduction in hardware. The earlier model, which it showed off late last year, relied on several cameras to capture speakers from all angles. This time out, however, the system only has a “few” of them, leaning on AI and ML to generate a convincing three-dimensional image of its subject.</p><p>According to Google, the system has gone “from the size of a restaurant booth to a flat-screen TV,” courtesy of these hardware updates<br>Google has already begun testing the latest prototype with a number of high profile partners, including T-Mobile, Salesforce and WeWork.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FobuyCkotJ_s%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DobuyCkotJ_s&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FobuyCkotJ_s%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/cc0e405adacdb4a323175fb60c1a4ad8/href">https://medium.com/media/cc0e405adacdb4a323175fb60c1a4ad8/href</a></iframe><h3>Google Introduce new Devices in Google I/O 2023</h3><h3>Pixel Tablet</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*06gocZUWb71RZQlRwkBWsA.jpeg" /><figcaption>credit:google</figcaption></figure><p>Announced at last year’s Google I/O, the Pixel Tablet is finally a reality. Well, almost — <a href="https://store.google.com/product/pixel_tablet?hl=en-US">preorders are live</a> (only in 11 countries), and it goes on sale June 20, so you still have to wait a bit more. This <strong>$499</strong> tablet isn’t really <em>meant</em> to be a tablet you take with you on the go. Rather, it rests on a magnetic dock (included) when you’re not using it, and the dock wirelessly recharges the slate and doubles as a speaker (the sound quality is purportedly equal to a <a href="https://store.google.com/us/product/nest_hub_2nd_gen?hl=en-US">Nest Hub</a>). When it’s on the dock, it acts as a traditional Google smart speaker, with options to control your smart home devices, and even has a similar microphone array to pick up your “Hey Google” commands. Chromecast is built in, so you can cast to it from your phone or laptop.</p><p>When you want to use it, just pop it off the dock and it’s a normal Android tablet — except a bit better, because Google has made some strides in improving the tablet experience on Android, with more than 50 Google apps optimized for the larger screen. It’s powered by the <a href="https://www.androidauthority.com/google-tensor-g2-explained-3216087/">Tensor G2 chipset</a> and has many of the same software features as other Pixel devices.</p><p>Sadly, there are no other accessories — no stylus (a pen-shaped instrument designed specifically for touch screen use.) and no keyboard. You <em>can</em> take it out and use it with Bluetooth accessories, but it’s clear Google is really envisioning this as a <strong>homebody</strong></p><h3>Pixel Fold</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*P6Yg1RyCHoyzlmMa-vZGrQ.jpeg" /><figcaption>credit: google</figcaption></figure><p>Google’s first folding phone, the <a href="https://store.google.com/product/pixel_fold?pli=1&amp;hl=en-US">Pixel Fold</a>, is here and costs a startling <strong>$1,799</strong>. It’s thinner than Samsung’s Galaxy Z Fold4, and there’s a wide, full front screen that offers up an almost normal smartphone experience. Open it up and you get a 7.6-inch OLED screen for watching movies, multitasking, or reading. Preorders are live now — if you bite, Google is tossing in a free Pixel Watch — but it ships in June</p><h3>Pixel 7a</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PmHXWn68CeFKYINjhxdIBw.jpeg" /><figcaption>credit: google</figcaption></figure><p>Every year, Google announces an A-series version of the flagship Pixel that came before. This year’s <a href="https://store.google.com/product/pixel_7a?hl=en-US">Pixel 7A</a> is a little more pricey ($499) than last year’s model, but you get a few more high-end perks, like a 90-Hz screen refresh rate and wireless charging support. The cameras are also completely new, with a 64-megapixel sensor leading the pack</p><p>It’s also launched in India. When it comes to the camera, it has a slightly higher pixel density, Its new chip does enable features like Face Unblur and Super Res Zoom.</p><p>If you found this information helpful, please show your support by clapping, sharing, and following. Thank you for taking the time to read this.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9cf5bb2943e8" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>