<?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"><channel><title><![CDATA[GeekSpark]]></title><description><![CDATA[GeekSpark]]></description><link>https://geekspark.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Wed, 18 Nov 2026 15:55:47 GMT</lastBuildDate><atom:link href="https://geekspark.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[SPF, DKIM and DMARC Explained: Stop Emails Going to Spam]]></title><description><![CDATA[Your website sends a password reset. The customer says it never arrived. You send a quote from your business email, and it lands in the client's spam…]]></description><link>https://geekspark.hashnode.dev/spf-dkim-and-dmarc-explained-stop-emails-going-to-spam</link><guid isPermaLink="true">https://geekspark.hashnode.dev/spf-dkim-and-dmarc-explained-stop-emails-going-to-spam</guid><category><![CDATA[Security]]></category><category><![CDATA[Devops]]></category><category><![CDATA[server]]></category><category><![CDATA[email]]></category><category><![CDATA[Linux]]></category><dc:creator><![CDATA[GeekSpark]]></dc:creator><pubDate>Wed, 18 Nov 2026 09:07:04 GMT</pubDate><enclosure url="https://cdn.hashnode.com/uploads/covers/6afc12b5eba15bb30d32f192/a4f81caf-9354-44f1-af3d-316eb8d110b4.webp" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>Your website sends a password reset. The customer says it never arrived. You send a quote from your business email, and it lands in the client's spam folder. Nothing looks wrong on your side. The email left, so where did it go?</p>
<p>Most of the time, the answer is three small DNS records: <strong>SPF, DKIM and DMARC</strong>. They are how Gmail, Outlook and every other mail provider decide whether an email really comes from you. If one of them is missing or broken, your email gets treated like a stranger at the door.</p>
<p>We learned this the hard way on our own site. We moved our domain's DNS to Cloudflare, everything looked fine, and then our contact form alerts quietly stopped arriving. The bounce message said: <code>550 5.7.1 Email rejected per DMARC policy</code>. The cause turned out to be a single setting on two records. More on that below, because it catches a lot of people.</p>
<p>This guide explains the three records in plain language, shows what correct ones look like, and lists the mistakes we see most often. No email background needed.</p>
<h2>The short version</h2>
<table>
<thead>
<tr>
<th>Record</th>
<th>What it answers</th>
<th>Where it lives</th>
<th>One-line example</th>
</tr>
</thead>
<tbody><tr>
<td><strong>SPF</strong></td>
<td>Which servers may send email for my domain?</td>
<td>TXT record on your domain</td>
<td><code>v=spf1 include:_spf.google.com ~all</code></td>
</tr>
<tr>
<td><strong>DKIM</strong></td>
<td>Was this email really signed by my domain, and unchanged?</td>
<td>TXT or CNAME at <code>selector._domainkey</code></td>
<td><code>v=DKIM1; k=rsa; p=MIIBIjAN...</code></td>
</tr>
<tr>
<td><strong>DMARC</strong></td>
<td>What should receivers do when the checks fail?</td>
<td>TXT record at <code>_dmarc</code></td>
<td><code>v=DMARC1; p=quarantine; rua=mailto:...</code></td>
</tr>
</tbody></table>
<p>A simple way to remember it:</p>
<ul>
<li><strong>SPF is the guest list.</strong> It names the servers allowed to send on your behalf.</li>
<li><strong>DKIM is the wax seal.</strong> Each email carries a signature that proves it came from you and was not changed on the way.</li>
<li><strong>DMARC is the house rule.</strong> It tells receivers what to do with an email that fails, and sends you reports about it.</li>
</ul>
<p>You want all three. Since 2024, Gmail and Yahoo require at least SPF or DKIM from everyone, and all three from anyone sending in bulk. Microsoft added similar rules for <a href="http://Outlook.com">Outlook.com</a> in 2025.</p>
<h2>SPF: the guest list</h2>
<p>SPF is one TXT record on your main domain. It lists every service that sends email with your domain in it. Here is a typical one for a business that uses Google Workspace for its inbox and SendGrid for its website and newsletter emails:</p>
<h2>DKIM: the wax seal</h2>
<p>When your email provider sends a message, it adds a digital signature to it using a private key that only the provider has. The matching public key sits in your DNS. The receiving server looks up that key and checks the signature. If it matches, the email is genuinely from your domain and nobody changed it on the way.</p>
<h2>DMARC: the house rule</h2>
<p>SPF and DKIM do the checking. DMARC decides what happens next. It is one TXT record at <code>_dmarc.yourdomain.com</code>:</p>
<h2>How to check your own setup in two minutes</h2>
<p><strong>The easiest way: Gmail.</strong> Send an email from your business address to a Gmail account. Open it, click the three dots (<strong>⋮</strong>) and choose <strong>Show original</strong>. At the top you'll see SPF, DKIM and DMARC, each with <strong>PASS</strong> or <strong>FAIL</strong>. Three passes means your setup is right.</p>
<h2>Moving your DNS? Use this checklist</h2>
<p>Changing DNS provider is the moment email breaks most often, because records get missed, mangled or proxied during the import. Before you switch nameservers:</p>
<ul>
<li><strong>Export or screenshot every record</strong> at your old provider, especially MX, every TXT and every <code>_domainkey</code> record.</li>
<li><strong>After the import, compare them line by line.</strong> Imports regularly drop or change records.</li>
<li><strong>Set every mail record to DNS only</strong> if your new host offers a proxy. That covers MX targets, <code>_domainkey</code> CNAMEs and any <code>mail</code> or <code>autodiscover</code> records.</li>
<li><strong>Check there is still exactly one SPF record</strong>, and that every <code>include:</code> still resolves.</li>
<li><strong>Send a test to Gmail</strong> and check <strong>Show original</strong> right after the switch.</li>
</ul>
<h2>All three pass, but you're still in spam?</h2>
<p>Then authentication is no longer the problem. Gmail is judging the email itself and your domain's track record. A few things help:</p>
<ul>
<li><strong>Give it a few days.</strong> A domain that just changed its setup, or sends rarely, has little reputation yet. Normal emails that get opened and replied to build it.</li>
<li><strong>Mark it "Not spam"</strong> and add the sender to your contacts while testing. Gmail learns from that quickly.</li>
<li><strong>Write like a person.</strong> One-line test messages, all caps, lots of links or link shorteners look like spam. A normal greeting and a couple of sentences do not.</li>
<li><strong>Send newsletters through a proper service</strong>, with a working unsubscribe link, and only to people who asked for them. Gmail expects bulk senders to keep spam complaints well below 0.3%.</li>
</ul>
<h2>Sources</h2>
<ul>
<li><a href="https://www.rfc-editor.org/rfc/rfc7208">RFC 7208: Sender Policy Framework (SPF)</a></li>
<li><a href="https://www.rfc-editor.org/rfc/rfc6376">RFC 6376: DomainKeys Identified Mail (DKIM)</a></li>
<li><a href="https://www.rfc-editor.org/rfc/rfc7489">RFC 7489: DMARC</a></li>
<li><a href="https://support.google.com/a/answer/81126">Google: email sender guidelines</a></li>
<li><a href="https://senders.yahooinc.com/best-practices/">Yahoo: sender best practices</a></li>
</ul>
<hr />
<p><em>This is the short version. The full guide, with every step and source, is on</em> <a href="https://geekspark.in/blog/spf-dkim-and-dmarc-explained-stop-emails-going-to-spam"><em>GeekSpark</em></a><em>.</em></p>
]]></content:encoded></item><item><title><![CDATA[AI as a Tool, Not a Human Replacement: The Real Cost in Jobs, Power and Water]]></title><description><![CDATA[At 9 a.m., AI writes the first draft of your report. At 4 p.m., an email lands: your company is "restructuring around AI," and your team is shrinking. It…]]></description><link>https://geekspark.hashnode.dev/ai-as-a-tool-not-a-human-replacement-the-real-cost-in-jobs-power-and-water</link><guid isPermaLink="true">https://geekspark.hashnode.dev/ai-as-a-tool-not-a-human-replacement-the-real-cost-in-jobs-power-and-water</guid><category><![CDATA[AI]]></category><category><![CDATA[Career]]></category><category><![CDATA[technology]]></category><category><![CDATA[Productivity]]></category><dc:creator><![CDATA[GeekSpark]]></dc:creator><pubDate>Wed, 18 Nov 2026 08:05:37 GMT</pubDate><enclosure url="https://cdn.hashnode.com/uploads/covers/6afc12b5eba15bb30d32f192/5510c85b-c91f-4c51-ae5d-958711525a7b.webp" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>At 9 a.m., AI writes the first draft of your report. At 4 p.m., an email lands: your company is "restructuring around AI," and your team is shrinking.</p>
<p>It sounds like a contradiction, but for a lot of office workers it's simply what 2026 has felt like. That's why the question everyone keeps asking, "will AI replace my job?", has no simple yes or no. AI is a brilliant tool. It's also a convenient excuse for layoffs, and it runs on power plants and water supplies that most people never see.</p>
<p>Here's an honest look at all three sides: what AI is genuinely good for, what it's doing to jobs, and what it's costing the planet. And why the best answer, for workers and companies alike, is to treat AI as a tool, not a replacement.</p>
<h2>The short answer</h2>
<p>AI works best as a power tool in human hands. It drafts, summarizes, translates, checks and speeds up routine work. It doesn't take responsibility, understand context the way people do, or care whether the result hurts someone. The companies and workers getting the most out of it use it to do more and better work, not to remove the person who checks the output.</p>
<h2>What AI is genuinely good at</h2>
<p>Used well, AI saves real time on the dull parts of almost any job:</p>
<ul>
<li><strong>First drafts.</strong> Emails, reports, product descriptions, cover letters. You still edit, but you never start from a blank page.</li>
<li><strong>Summaries.</strong> A 40-page contract, a long meeting transcript or a messy email thread, boiled down in seconds.</li>
<li><strong>Code.</strong> Boilerplate, tests, explanations of unfamiliar code and quick fixes. Developers spend more time on design and less on typing.</li>
<li><strong>Translation and accessibility.</strong> Instant translation, captions, alt text and plain-language rewrites open work up to more people.</li>
<li><strong>Data chores.</strong> Cleaning spreadsheets, writing formulas, spotting outliers and turning numbers into charts.</li>
<li><strong>Learning.</strong> Asking "explain this like I'm new to it" until it finally clicks.</li>
</ul>
<h2>The layoffs are real, and so is the excuse</h2>
<p>Here's where the "tool" story gets uncomfortable. In the US, companies now name AI as a reason for job cuts more often than ever before.</p>
<h2>The environmental bill nobody sees</h2>
<p>Every prompt runs on a real computer in a real building that needs electricity and, often, water to stay cool.</p>
<h2>Tool or replacement: where to draw the line</h2>
<p>A simple test helps. AI should do the work; a person should own the result.</p>
<h2>What workers can do right now</h2>
<p>First, see where you stand: the <a href="https://geekspark.in/tools/ai-careers/will-ai-take-my-job">Will AI take my job? checker</a> shows how much of your job AI could do, how it ranks against 923 jobs and which tasks still need a human. For the full list, see <a href="https://geekspark.in/blog/jobs-ai-will-replace">the jobs AI will replace, and the ones it can't</a>.</p>
<h2>Sources</h2>
<ul>
<li><a href="https://www.challengergray.com/wp-content/uploads/2026/09/Challenger-Report-August-2026.pdf">Challenger, Gray &amp; Christmas: job cut announcement report (August 2026)</a></li>
<li><a href="https://time.news/challenger-data-shows-ai-linked-job-cuts-more-than-double-in-2026/">Time News: Challenger data shows AI-linked job cuts more than double in 2026</a></li>
<li><a href="https://www.eib.org/en/publications/20250383-economics-working-paper-2026-02">European Investment Bank: AI adoption, productivity and employment (Working Paper 2026/02)</a></li>
<li><a href="https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it">Harvard Business Review: AI doesn't reduce work, it intensifies it</a></li>
<li><a href="https://ourworldindata.org/how-much-energy-do-data-centers-and-artificial-intelligence-use">Our World in Data: how much energy do data centers and AI use?</a></li>
<li><a href="https://www.axios.com/2026/06/30/google-ai-boom-emissions-power-use">Axios: Google's AI boom sends emissions, power use soaring</a></li>
<li><a href="https://sustainability.google/google-2026-environmental-report/">Google 2026 Environmental Report</a></li>
</ul>
<hr />
<p><em>This is the short version. The full guide, with every step and source, is on</em> <a href="https://geekspark.in/blog/ai-tool-not-human-replacement-layoffs-environment"><em>GeekSpark</em></a><em>.</em></p>
]]></content:encoded></item><item><title><![CDATA[PHP2Shell & the Node.js Brotli Flaw: Patch Guide]]></title><description><![CDATA[On October 21, 2026, PHP and Node.js both shipped urgent security releases on the same day. On the PHP side, two critical flaws nicknamed PHP2Shell let an…]]></description><link>https://geekspark.hashnode.dev/php2shell-the-node-js-brotli-flaw-patch-guide</link><guid isPermaLink="true">https://geekspark.hashnode.dev/php2shell-the-node-js-brotli-flaw-patch-guide</guid><category><![CDATA[Linux]]></category><category><![CDATA[Devops]]></category><category><![CDATA[Security]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[GeekSpark]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:10:58 GMT</pubDate><enclosure url="https://cdn.hashnode.com/uploads/covers/6afc12b5eba15bb30d32f192/c33065d7-6d10-4306-ad22-6478dbeb44cb.webp" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>On October 21, 2026, PHP and Node.js both shipped urgent security releases on the same day. On the PHP side, two critical flaws nicknamed <strong>PHP2Shell</strong> let an attacker run code on a web server without logging in. On the Node.js side, a bug in the built-in <strong>Brotli</strong> decompressor could be abused the same way.</p>
<p>If you run a WordPress site, a Laravel app, a PHP script you bought, or anything built on Node.js, this affects you. The good news is that the fix is simple: update, then restart. This guide explains what happened in plain language, which versions are safe, and exactly what to check on your servers.</p>
<p><em>Last checked: November 7, 2026. Links to the official advisories are at the end.</em></p>
<h2>The short version</h2>
<table>
<thead>
<tr>
<th>If you run</th>
<th>Update to</th>
<th>Notes</th>
</tr>
</thead>
<tbody><tr>
<td>PHP 8.5</td>
<td>8.5.12 or later</td>
<td>Fixes PHP2Shell</td>
</tr>
<tr>
<td>PHP 8.4</td>
<td>8.4.27 or later</td>
<td>Fixes PHP2Shell</td>
</tr>
<tr>
<td>PHP 8.3</td>
<td>8.3.36 or later</td>
<td>Fixes PHP2Shell</td>
</tr>
<tr>
<td>PHP 8.2</td>
<td>8.2.35 or later</td>
<td>8.2 reaches end of life on December 31, 2026</td>
</tr>
<tr>
<td>PHP 8.1 or older</td>
<td>A supported version</td>
<td>No official fix; end of life</td>
</tr>
<tr>
<td>Node.js 26</td>
<td>26.12.1 or later</td>
<td>Node.js 26 is now the active LTS line</td>
</tr>
<tr>
<td>Node.js 24</td>
<td>24.22.1 or later</td>
<td></td>
</tr>
<tr>
<td>Node.js 22</td>
<td>22.23.4 or later</td>
<td></td>
</tr>
<tr>
<td>Node.js 25, 23, 21, 20 or older</td>
<td>A supported line (22, 24 or 26)</td>
<td>No fix for these lines</td>
</tr>
</tbody></table>
<p>After updating, <strong>restart</strong> every PHP worker and every Node.js process. A server that is still running the old version in memory is still vulnerable, even if the new version is installed on disk.</p>
<h2>PHP2Shell: what actually happened</h2>
<p>PHP2Shell is the name given to two separate vulnerabilities in PHP itself: <strong>CVE-2026-109618</strong> and <strong>CVE-2026-109619</strong>. Both are rated 9.2 out of 10 (CVSS v4.0), and both can allow <strong>unauthenticated remote code execution</strong>. In everyday terms, that means a stranger on the internet could send a specially crafted request and make your server run their code.</p>
<p>What makes them serious is where they live. These are not bugs in WordPress, Laravel or a plugin. They are in the PHP runtime, the engine underneath all of them. That is why updating your application or your plugins does <strong>not</strong> fix them. The PHP version on the server has to change.</p>
<p>In simple terms, here is what goes wrong:</p>
<ul>
<li><strong>CVE-2026-109618:</strong> when PHP runs out of memory at the wrong moment, it can stop while it is still setting up a request. The cleanup that follows then treats half-built data as if it were valid.</li>
<li><strong>CVE-2026-109619:</strong> some internal data that belongs to a request can survive after its memory has been freed, and PHP may later use it.</li>
</ul>
<p>Both are memory-handling mistakes, the kind attackers have a long history of turning into code execution.</p>
<h3>Who is affected</h3>
<p>Almost every PHP version back to PHP 5.4 is affected, across the common ways PHP is run: Apache with mod_php, PHP-FPM (used with Nginx and most modern hosts), php-cgi, FrankenPHP and LiteSpeed. Reports describe <strong>Apache mod_php</strong> as the highest-risk setup, because a single request can trigger the problem before any of your application code even runs.</p>
<p>Researchers also noted that an attacker may be able to read memory from a running PHP worker. That memory can contain passwords, API keys and session tokens. So even if no code is executed, data can leak.</p>
<h3>Is it being exploited?</h3>
<p>Details leaked a day before the planned disclosure. Within hours of the patches going out, security firm watchTowr reported <strong>active exploitation attempts</strong> against both flaws, and Cloudflare said it saw early attacks and deployed firewall rules. At the time of writing, no specific successful break-in has been publicly confirmed. But "people are already trying" is exactly when you want to be patched.</p>
<h3>Are there workarounds?</h3>
<p>Not really. The PHP project's own advisory says there are <strong>no complete mitigations</strong> and that updating is the only safe choice. Some configuration changes reduce specific attack paths, but none closes the hole. Treat them only as a stopgap while you schedule the update, never as the fix.</p>
<h3>What about older PHP versions?</h3>
<p>PHP 8.1 and older no longer receive official updates, so there is no upstream fix for them. Some Linux distributions and commercial support vendors publish their own backported patches. If you are on an older version, upgrading to a supported release is the real answer. If you are on <strong>PHP 8.2</strong>, plan your move now: it reaches end of life on <strong>December 31, 2026</strong>.</p>
<p>One useful detail: Linux distributions often keep older-looking version numbers and add the fix in their own package. If your server shows something like <code>8.3.6</code> from Ubuntu, check the package changelog for the CVE numbers rather than comparing version numbers alone.</p>
<h2>A third PHP issue that is not patched yet</h2>
<p>Two days later, PHP published an informational advisory about <strong>CVE-2026-115560</strong>, a separate remote code execution risk in the <strong>session upload progress</strong> feature. As of November 7, 2026, it has <strong>not</strong> been fixed in a release. PHP says it will ship the fix and a full advisory with its next regular release.</p>
<p>The risk is narrow. It needs a specific setup: the session serializer is set to <code>php_binary</code> or <code>php_serialize</code>, the application switches it to <code>php</code> with <code>ini_set()</code> before starting the session, and upload progress is enabled. Most applications never do this.</p>
<p>Still, the PHP team's own recommendation is easy to apply. Add this to your <code>php.ini</code> or PHP-FPM pool configuration, then restart PHP:</p>
<pre><code class="language-ini">session.upload_progress.enabled = 0
</code></pre>
<p>Normal file uploads keep working. The only thing you lose is the built-in upload progress tracking, which few applications use.</p>
<h2>Node.js: the Brotli decompression flaw</h2>
<p>Brotli is a compression format, the <code>br</code> you may have seen in <code>Content-Encoding</code> headers. Node.js includes a Brotli decoder in its built-in <code>zlib</code> module.</p>
<p><strong>CVE-2026-111497</strong> is an out-of-bounds write in that decoder. A malformed Brotli stream can make it write outside the memory it was given. Depending on the situation, that can crash your app, leak memory, or lead to remote code execution. Node.js rates it High; the CVSS v4.0 score is 9.2.</p>
<h3>Who is exposed</h3>
<p>You are most at risk if your app <strong>decompresses Brotli data that comes from someone else</strong>. For example:</p>
<ul>
<li>HTTP clients, scrapers or proxies that accept <code>Content-Encoding: br</code> responses from servers you don't control.</li>
<li>APIs that accept compressed request bodies.</li>
<li>Services that process uploaded compressed files.</li>
</ul>
<p>If your app never decompresses Brotli from untrusted sources, the risk is lower, but the same release fixes several other issues, so you should update anyway.</p>
<h3>The other fixes in the same release</h3>
<p>The October 21 release fixed ten CVEs in total. Besides Brotli, the most important ones are:</p>
<ul>
<li><strong>TLS certificate checks (CVE-2026-105483, CVE-2026-106069, CVE-2026-106072):</strong> unusual ways of writing an IP address, such as fullwidth digits or a leading zero, could make Node.js accept a certificate for the wrong address.</li>
<li><strong>DNS (CVE-2026-105482):</strong> <code>dns.resolve()</code> and <code>dns.reverse()</code> could silently cut a hostname short at a hidden NUL character and look up a different name.</li>
<li><strong>Keep-alive connections (CVE-2026-105481):</strong> a malicious server could send a response early, which Node.js would then treat as the answer to your next request.</li>
<li><strong>Random values in snapshots (CVE-2026-106074):</strong> apps built as single executables or from startup snapshots could produce the same <code>crypto.randomUUID()</code> values in every process.</li>
</ul>
<h3>Which versions are fixed</h3>
<p>The fixes are in <strong>Node.js 22.23.4, 24.22.1 and 26.12.1</strong> (or anything newer on those lines). Node.js 26 became the active LTS line on October 28, 2026, so it is a good target if you are upgrading anyway. Odd-numbered lines (25, 23, 21) and Node.js 20 and older did not get a fix. If you are on one of those, move to 22, 24 or 26.</p>
<h3>One overlap worth knowing</h3>
<p>The Brotli bug was in the Brotli library itself, which other software bundles too. PHP's official <strong>Windows</strong> builds include Brotli, so PHP published rebuilt Windows packages on October 26 (versions ending in <code>-1</code>, such as <code>8.4.27-1</code>) with the fixed library. If you run PHP on Windows, make sure you have those rebuilds, not just the October 21 versions.</p>
<h2>Your patch checklist</h2>
<p>Work through this list on every server, container and hosting account you are responsible for.</p>
<p><strong>1. Find out what you are running.</strong></p>
<pre><code class="language-bash">php -v
php -i | grep "Server API"
node -v
</code></pre>
<p>The first two lines tell you your PHP version and how PHP runs (Apache module, FPM or CGI). The third gives your Node.js version. If you use Docker, check the base image tag in each <code>Dockerfile</code> too.</p>
<p><strong>2. Update PHP.</strong> On a VPS, install the latest package for your PHP version through your package manager, or switch to a fixed version. On shared hosting or cPanel, use the PHP version selector, or ask your host to confirm they have applied the October 21 releases.</p>
<p><strong>3. Update Node.js.</strong> Install the fixed release for your line with your version manager (nvm, fnm or Volta) or your package manager. Update the version in your Dockerfiles and CI too.</p>
<p><strong>4. Restart everything.</strong> Restart PHP-FPM or Apache, every Node.js process (PM2, systemd services, serverless functions), and rebuild and redeploy containers and single-executable apps. A patched runtime does nothing until the old processes are gone.</p>
<p><strong>5. Turn off session upload progress</strong> until PHP ships the fix for CVE-2026-115560, as shown above.</p>
<p><strong>6. Check it worked.</strong> Run <code>php -v</code> and <code>node -v</code> again, and for PHP, load a page with <code>phpinfo()</code> (then delete it) to confirm the web server uses the new version, not just the command line.</p>
<p><strong>7. Keep an eye on your logs.</strong> Look for unusual requests, unexpected crashes of PHP workers, or new files appearing in upload folders. If you suspect anything, rotate the passwords and API keys stored on that server, because memory leaks can expose them.</p>
<h2>Frequently asked questions</h2>
<p><strong>Does updating WordPress or my plugins fix PHP2Shell?</strong> No. PHP2Shell is in PHP itself. Your hosting provider or server administrator has to update the PHP version. Updating WordPress is still a good idea, but it does not close this hole.</p>
<p><strong>I use a PHP script or Laravel app. Do I need to change my code?</strong> In almost all cases, no. Update PHP, restart it, and your application keeps running as before. The one exception is the upload progress setting for CVE-2026-115560, which rarely affects applications.</p>
<p><strong>My site runs on shared hosting. What should I do?</strong> Check the PHP version in your hosting panel. If it doesn't show 8.2.35, 8.3.36, 8.4.27, 8.5.12 or later, ask your host whether they have applied the October 21 security patches. Many hosts patch their own packages without changing the version number shown.</p>
<p><strong>Is Node.js 20 still safe if I don't use Brotli?</strong> Node.js 20 is end of life and no longer receives security fixes, so every future flaw will stay open. Upgrade to 22, 24 or 26.</p>
<h2>The bottom line</h2>
<p>Neither PHP2Shell nor the Node.js Brotli flaw requires you to change your application. Both require you to update the runtime and restart it. It is usually a ten-minute job per server, and attackers were already trying these flaws within hours of the patches.</p>
<p>If you bought a script from us and need help checking or updating the server it runs on, <a href="https://geekspark.in/contact">get in touch</a>. We are happy to point you in the right direction.</p>
<h2>Sources</h2>
<ul>
<li>PHP: <a href="https://www.php.net/">October 21, 2026 security releases and advisory</a>, including the informational advisory on session upload progress (October 23, 2026)</li>
<li>Node.js: <a href="https://nodejs.org/en/blog/vulnerability/october-2026-security-releases">October 2026 security releases</a></li>
<li>The Hacker News: <a href="https://thehackernews.com/2026/10/php-issues-emergency-patches-for-two.html">PHP issues emergency patches for two critical PHP2Shell RCE flaws</a></li>
<li>Qualys: <a href="https://blog.qualys.com/product-tech/2026/10/23/node-js-patches-brotli-remote-code-execution-flaw-find-vulnerable-runtimes-before-attackers-do">Node.js patches Brotli remote code execution flaw</a></li>
</ul>
<p><strong>Running Apache too?</strong> A critical HTTP/2 flaw was fixed in November 2026. Our guide to <a href="https://geekspark.in/blog/apache-http2-flaw-cve-2026-119618-check-and-patch-your-server-today">Apache CVE-2026-119618</a> shows how to check and patch it in ten minutes.</p>
<h2>Sources</h2>
<ul>
<li><a href="https://www.php.net/">October 21, 2026 security releases and advisory</a></li>
<li><a href="https://nodejs.org/en/blog/vulnerability/october-2026-security-releases">October 2026 security releases</a></li>
<li><a href="https://thehackernews.com/2026/10/php-issues-emergency-patches-for-two.html">PHP issues emergency patches for two critical PHP2Shell RCE flaws</a></li>
<li><a href="https://blog.qualys.com/product-tech/2026/10/23/node-js-patches-brotli-remote-code-execution-flaw-find-vulnerable-runtimes-before-attackers-do">Node.js patches Brotli remote code execution flaw</a></li>
</ul>
<hr />
<p><em>Originally published on</em> <a href="https://geekspark.in/blog/php2shell-the-nodejs-brotli-flaw-patch-guide"><em>GeekSpark</em></a><em>.</em></p>
]]></content:encoded></item><item><title><![CDATA[Apache HTTP/2 Flaw CVE-2026-119618: Check and Patch Your Server Today]]></title><description><![CDATA[On November 11, the Apache Software Foundation shipped an urgent fix for its web server. The bug, CVE-2026-119618, sits in the part of Apache that handles…]]></description><link>https://geekspark.hashnode.dev/apache-http-2-flaw-cve-2026-119618-check-and-patch-your-server-today</link><guid isPermaLink="true">https://geekspark.hashnode.dev/apache-http-2-flaw-cve-2026-119618-check-and-patch-your-server-today</guid><category><![CDATA[apache]]></category><category><![CDATA[server]]></category><category><![CDATA[bug]]></category><category><![CDATA[Linux]]></category><category><![CDATA[Devops]]></category><category><![CDATA[Security]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[GeekSpark]]></dc:creator><pubDate>Tue, 17 Nov 2026 10:24:06 GMT</pubDate><enclosure url="https://cdn.hashnode.com/uploads/covers/6afc12b5eba15bb30d32f192/80aa6509-0d12-4ea3-bff8-86cd711f26b3.webp" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>On November 11, the Apache Software Foundation shipped an urgent fix for its web server. The bug, <strong>CVE-2026-119618</strong>, sits in the part of Apache that handles HTTP/2. On an affected server, an attacker with no login and no password could send carefully timed requests that corrupt the server's memory and, in the worst case, run their own code on it.</p>
<p>Apache powers a huge share of the web, so this one matters. The good news: it takes about ten minutes to find out whether you are affected, and most servers can be fixed with one command.</p>
<p>We ran these exact checks on our own Ubuntu server the morning after the news broke. The result surprised us a little, and it is the thing most people get wrong about this bug, so we will start there.</p>
<p><em>Last checked: November 13, 2026.</em></p>
<h2>The short version</h2>
<table>
<thead>
<tr>
<th></th>
<th></th>
</tr>
</thead>
<tbody><tr>
<td><strong>What it is</strong></td>
<td>A race condition in <code>mod_http2</code> that can corrupt heap memory, possibly leading to remote code execution</td>
</tr>
<tr>
<td><strong>Affected</strong></td>
<td>Apache HTTP Server 2.4.55 to 2.4.70, with HTTP/2 turned on</td>
</tr>
<tr>
<td><strong>Fixed in</strong></td>
<td>Apache 2.4.71, and security updates from Ubuntu, Debian, Red Hat and others</td>
</tr>
<tr>
<td><strong>Exploited yet?</strong></td>
<td>No attacks seen so far, but researchers expect that to change</td>
</tr>
<tr>
<td><strong>Quick workaround</strong></td>
<td>Turn off HTTP/2 until you can update</td>
</tr>
</tbody></table>
<p>The flaw was found by Anthropic's OSS Scanner and Google Cloud's Vulnerability Research team. Cloud security company Wiz says 64% of the cloud environments it monitors contain a vulnerable Apache install.</p>
<h2>"My server says 2.4.58. Am I vulnerable?"</h2>
<p>This is the question we had too. Our server runs Ubuntu 24.04, and <code>apache2 -v</code> printed:</p>
<pre><code class="language-text">Server version: Apache/2.4.58 (Ubuntu)
</code></pre>
<p>2.4.58 is inside the affected range, so it looks bad. But Linux distributions rarely jump to a new Apache version in the middle of a release. Instead, they copy the fix into the version they already ship. This is called <strong>backporting</strong>. The number stays at 2.4.58, and only the package revision at the end changes.</p>
<p>So the version number alone tells you nothing. What you need to know is whether your <strong>installed package</strong> includes the fix. On Ubuntu and Debian, read the changelog that came with it:</p>
<pre><code class="language-bash">zcat /usr/share/doc/apache2/changelog.Debian.gz | grep -m1 CVE-2026-119618
</code></pre>
<p>(Use this rather than <code>apt changelog</code>, which shows the newest package available, not the one you have installed.)</p>
<p>On our server it printed the fix straight away. Ubuntu released it as <code>2.4.58-1ubuntu8.19</code>, and our automatic security updates had installed it within hours. If you get no output, you are not patched yet.</p>
<p>On Red Hat, AlmaLinux and Rocky Linux, HTTP/2 lives in its own package:</p>
<pre><code class="language-bash">rpm -q --changelog mod_http2 | grep -m1 CVE-2026-119618
</code></pre>
<h2>Step 1: Is HTTP/2 even turned on?</h2>
<p>The bug only matters if Apache is speaking HTTP/2. Two things must both be true: the <code>http2</code> module is loaded, and the <code>Protocols</code> line includes <code>h2</code> or <code>h2c</code>.</p>
<p>Check the module first:</p>
<pre><code class="language-bash"># Ubuntu / Debian
sudo apache2ctl -M | grep http2

# Red Hat / Alma / Rocky
sudo httpd -M | grep http2
</code></pre>
<p>If this prints <code>http2_module</code>, the module is loaded. If it prints nothing, HTTP/2 is off and this bug can't reach you. Patch anyway, but you can breathe.</p>
<p>Then look at the protocol setting:</p>
<pre><code class="language-bash">grep -ri "^\s*Protocols" /etc/apache2/ /etc/httpd/ 2&gt;/dev/null
</code></pre>
<p>A line like <code>Protocols h2 http/1.1</code> means HTTP/2 is offered to visitors.</p>
<p>On our server, the <code>Protocols h2 h2c http/1.1</code> line existed, but only in the module's own config file. The module itself was never enabled, so it was never active. A config file being present is not the same as a feature being switched on.</p>
<h3>A trap if you use Cloudflare or a CDN</h3>
<p>You may find advice to test from outside with <code>curl --http2</code>. If your site sits behind Cloudflare, a CDN or a load balancer, that test talks to <strong>them</strong>, not to your server. It will say "HTTP/2" even when your own Apache only speaks HTTP/1.1. Always check on the server itself.</p>
<h2>Step 2: Install the fix</h2>
<p><strong>Ubuntu and Debian:</strong></p>
<pre><code class="language-bash">sudo apt update
sudo apt install --only-upgrade apache2 apache2-bin
sudo systemctl restart apache2
</code></pre>
<p><strong>Red Hat, AlmaLinux, Rocky Linux:</strong></p>
<pre><code class="language-bash">sudo dnf update httpd mod_http2
sudo systemctl restart httpd
</code></pre>
<p><strong>Built from source:</strong> upgrade to <strong>Apache 2.4.71</strong>. If you use the standalone <code>mod_h2</code> module from GitHub, note that it is no longer maintained, so move to the module included in httpd 2.4.71.</p>
<p><strong>Docker:</strong> pull a 2.4.71 image (for example <code>httpd:2.4.71</code>) and redeploy. Restarting an old container does nothing; the fix only arrives with the new image.</p>
<p><strong>cPanel, Plesk and other control panels:</strong> run the panel's own update tool. Their Apache builds are updated through the panel, not through <code>apt</code> or <code>dnf</code>.</p>
<p><strong>Don't skip the restart.</strong> Updating the package replaces the files on disk, but the old Apache keeps running from memory until it restarts. After restarting, run the changelog check again to confirm.</p>
<h2>Can't update today? Turn HTTP/2 off</h2>
<p>If you can't update right now (a change freeze, an old server, a client who has to approve it), switch HTTP/2 off as a stopgap. Your site keeps working over HTTP/1.1.</p>
<p><strong>Ubuntu and Debian:</strong></p>
<pre><code class="language-bash">sudo a2dismod http2
sudo systemctl restart apache2
</code></pre>
<p><strong>Red Hat and friends:</strong> comment out the <code>LoadModule http2_module</code> line in <code>/etc/httpd/conf.modules.d/10-h2.conf</code>, then restart <code>httpd</code>.</p>
<p>Also change any <code>Protocols h2 http/1.1</code> line to <code>Protocols http/1.1</code>, so nothing tries to use HTTP/2.</p>
<p>Will visitors notice? Usually not. HTTP/2 makes pages load a little faster when a page pulls in many files, but for most sites the difference is small. If you are behind Cloudflare, visitors still get HTTP/2 or HTTP/3 from Cloudflare, which talks to your server over HTTP/1.1 unless you switched on its "HTTP/2 to Origin" option.</p>
<h2>Is this as bad as it sounds?</h2>
<p>Honestly, it is serious but not a panic. A few things work in your favor:</p>
<ul>
<li><strong>HTTP/2 is off by default</strong> in Apache. Many servers never turned it on.</li>
<li><strong>Exploiting a race condition is hard.</strong> The attacker has to win a timing race between requests, and it doesn't work every time. Red Hat notes that its own builds make it much harder still.</li>
<li><strong>No attacks have been seen</strong> at the time of writing.</li>
</ul>
<p>On the other hand, a public proof of concept already crashes Apache. Even without full code execution, that is an easy way to knock a site offline. And once a fix is public, attackers study it to work out how to exploit the bug. The window between "fixed" and "exploited" keeps getting shorter.</p>
<h2>The bigger lesson: let security updates install themselves</h2>
<p>Our server was patched before we even read the news, because it installs security updates automatically. That is the cheapest protection a small team can have. On Ubuntu and Debian it takes a minute to set up:</p>
<pre><code class="language-bash">sudo apt install unattended-upgrades
sudo dpkg-reconfigure -plow unattended-upgrades
</code></pre>
<p>Our <a href="https://geekspark.in/blog/secure-a-new-linux-vps-in-30-minutes-2026-checklist">30-minute Linux VPS security checklist</a> covers this along with SSH keys, the firewall and Fail2ban. And if you run PHP or Node.js on the same server, our <a href="https://geekspark.in/blog/php2shell-the-nodejs-brotli-flaw-patch-guide">PHP2Shell and Node.js patch guide</a> walks through the same check-and-patch routine for those.</p>
<p>While you are logged in, it is also a good moment to check that your certificate isn't close to expiring with our free <a href="https://geekspark.in/tools/network-server/ssl-certificate-checker">SSL certificate checker</a>.</p>
<h2>FAQ</h2>
<p><strong>What is CVE-2026-119618?</strong> A race condition in Apache's <code>mod_http2</code> module. Concurrent requests can corrupt the server's heap memory, which can crash Apache and could let an attacker run code with the web server's permissions.</p>
<p><strong>Which Apache versions are affected?</strong> Apache HTTP Server 2.4.55 through 2.4.70, and the standalone <code>mod_h2</code> module 2.0.8 through 2.0.42, but only when HTTP/2 is enabled. Apache 2.4.71 fixes it.</p>
<p><strong>I'm on Ubuntu with Apache 2.4.58. Do I need 2.4.71?</strong> No. Ubuntu backports the fix into its own 2.4.58 package. Run <code>zcat /usr/share/doc/apache2/changelog.Debian.gz | grep CVE-2026-119618</code>. If the CVE is listed, you are patched. For Ubuntu 24.04 the fixed package is <code>2.4.58-1ubuntu8.19</code>.</p>
<p><strong>Does turning off HTTP/2 break my site?</strong> No. Browsers fall back to HTTP/1.1 automatically. Pages may load very slightly slower on some sites, and behind a CDN you won't notice any difference.</p>
<p><strong>Is Nginx affected?</strong> No. This bug is in Apache's own HTTP/2 module. Nginx, LiteSpeed and Caddy use different code.</p>
<p><strong>I'm on shared hosting. What should I do?</strong> Your host manages Apache, so there is nothing to install yourself. If you want to be sure, ask support whether CVE-2026-119618 has been patched on your server.</p>
<p>Ten minutes, a changelog check and a restart. If you look after a server running Apache, do it today, before the people studying this patch get there first.</p>
<h2>Sources</h2>
<ul>
<li><a href="https://httpd.apache.org/security/vulnerabilities_24.html">Apache HTTP Server 2.4 vulnerabilities</a></li>
<li><a href="https://access.redhat.com/security/cve/cve-2026-119618">Red Hat: CVE-2026-119618</a></li>
</ul>
<hr />
<p><em>Originally published on</em> <a href="https://geekspark.in/blog/apache-http2-flaw-cve-2026-119618-check-and-patch-your-server-today"><em>GeekSpark</em></a><em>.</em></p>
]]></content:encoded></item><item><title><![CDATA[Jobs AI Will Replace, and the Ones It Can't: 923 Jobs Ranked]]></title><description><![CDATA[Computer programmers and software developers often share an office. They write code, use the same tools and answer to the same boss. Yet on paper, their…]]></description><link>https://geekspark.hashnode.dev/jobs-ai-will-replace-and-the-ones-it-can-t-923-jobs-ranked</link><guid isPermaLink="true">https://geekspark.hashnode.dev/jobs-ai-will-replace-and-the-ones-it-can-t-923-jobs-ranked</guid><category><![CDATA[technology]]></category><category><![CDATA[AI]]></category><category><![CDATA[Productivity]]></category><category><![CDATA[Career]]></category><dc:creator><![CDATA[GeekSpark]]></dc:creator><pubDate>Tue, 17 Nov 2026 10:18:42 GMT</pubDate><enclosure url="https://cdn.hashnode.com/uploads/covers/6afc12b5eba15bb30d32f192/92a68301-1300-4990-9f79-b68b93757364.webp" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>Computer programmers and software developers often share an office. They write code, use the same tools and answer to the same boss.</p>
<p>Yet on paper, their futures split in two. Programmers are among the jobs most exposed to AI, and the US government expects 11% fewer of them by 2032. Developers are exposed too, but their numbers are forecast to grow by 26%.</p>
<p>That gap is the real story behind every "jobs AI will replace" list. AI exposure alone doesn't decide who loses work. What matters is exposure <strong>plus</strong> demand. So we took the task-by-task ratings for 923 US jobs, set them against official job forecasts, and sorted them into three groups: the jobs at real risk, the jobs AI is reshaping, and the jobs it can barely touch.</p>
<h2>The short version</h2>
<table>
<thead>
<tr>
<th></th>
<th>Jobs at risk</th>
<th>Jobs being reshaped</th>
<th>Jobs AI can't replace (yet)</th>
</tr>
</thead>
<tbody><tr>
<td>AI exposure</td>
<td>High</td>
<td>High</td>
<td>Low</td>
</tr>
<tr>
<td>Forecast to 2032</td>
<td>Shrinking</td>
<td>Growing</td>
<td>Growing</td>
</tr>
<tr>
<td>Examples</td>
<td>Executive assistants, customer service, order clerks</td>
<td>Data scientists, security analysts, software developers</td>
<td>Physician assistants, home health aides, cooks</td>
</tr>
<tr>
<td>What to do</td>
<td>Move toward tasks AI is weak at</td>
<td>Learn the tools before others do</td>
<td>Use AI for the paperwork around your work</td>
</tr>
</tbody></table>
<p>Want your own number? The <a href="https://geekspark.in/tools/ai-careers/will-ai-take-my-job">Will AI take my job? checker</a> shows your job's score, where it ranks and which of your tasks still need a human.</p>
<blockquote>
<p><strong>Reader poll:</strong> How worried are you about AI taking your job? <a href="https://geekspark.in/blog/jobs-ai-will-replace">Vote on GeekSpark</a></p>
</blockquote>
<h2>Which jobs will AI replace first?</h2>
<p>These are the jobs where two warning lights flash at once: AI can speed up more than half of the work, and the Bureau of Labor Statistics (BLS) already expects fewer of them. All have at least 100,000 US workers.</p>
<table>
<thead>
<tr>
<th>Job</th>
<th>AI exposure</th>
<th>Forecast, 2022 to 2032</th>
</tr>
</thead>
<tbody><tr>
<td>Legal secretaries</td>
<td>76%</td>
<td>−22%</td>
</tr>
<tr>
<td>Executive assistants</td>
<td>74%</td>
<td>−21%</td>
</tr>
<tr>
<td>Customer service representatives</td>
<td>70%</td>
<td>−6%</td>
</tr>
<tr>
<td>Advertising sales agents</td>
<td>69%</td>
<td>−7%</td>
</tr>
<tr>
<td>Computer programmers</td>
<td>68%</td>
<td>−11%</td>
</tr>
<tr>
<td>Editors</td>
<td>65%</td>
<td>−4%</td>
</tr>
<tr>
<td>Order clerks</td>
<td>60%</td>
<td>−18%</td>
</tr>
<tr>
<td>Secretaries and admin assistants</td>
<td>57%</td>
<td>−12%</td>
</tr>
</tbody></table>
<p>Notice the pattern. These jobs move information around: typing it, routing it, summarizing it, answering questions about it. That's the work language models handle best.</p>
<p>Customer service stands out for its size. Nearly 3 million Americans do it, and it scores 70% for AI exposure. Even a small shift there touches a lot of people.</p>
<h2>The jobs AI is changing, not killing</h2>
<p>High exposure doesn't always mean fewer jobs. In these roles AI can do much of the work, yet demand is growing fast:</p>
<table>
<thead>
<tr>
<th>Job</th>
<th>AI exposure</th>
<th>Forecast, 2022 to 2032</th>
</tr>
</thead>
<tbody><tr>
<td>Data scientists</td>
<td>59%</td>
<td>+35%</td>
</tr>
<tr>
<td>Information security analysts</td>
<td>54%</td>
<td>+32%</td>
</tr>
<tr>
<td>Software developers</td>
<td>45%</td>
<td>+26%</td>
</tr>
<tr>
<td>Operations research analysts</td>
<td>63%</td>
<td>+23%</td>
</tr>
<tr>
<td>Software testers (QA)</td>
<td>61%</td>
<td>+20%</td>
</tr>
<tr>
<td>Logistics analysts</td>
<td>65%</td>
<td>+18%</td>
</tr>
</tbody></table>
<p>What do they share? The world wants a lot more of what they produce: more software, more security, more analysis. When AI makes a job faster and demand keeps rising, companies tend to ask for more work, not fewer people.</p>
<p>That's also why programmers and developers part ways. Programmers mostly turn finished designs into code, a narrow task AI now does quickly. Developers decide what to build, how the pieces fit and what can go wrong. AI speeds up both jobs, but only one of them grows with demand.</p>
<h2>Jobs AI can't replace (for now)</h2>
<p>At the other end are jobs where AI barely registers and demand is rising anyway:</p>
<table>
<thead>
<tr>
<th>Job</th>
<th>AI exposure</th>
<th>Forecast, 2022 to 2032</th>
</tr>
</thead>
<tbody><tr>
<td>Physician assistants</td>
<td>12%</td>
<td>+27%</td>
</tr>
<tr>
<td>Physical therapist assistants</td>
<td>10%</td>
<td>+26%</td>
</tr>
<tr>
<td>Home health and personal care aides</td>
<td>4%</td>
<td>+22%</td>
</tr>
<tr>
<td>Restaurant cooks</td>
<td>12%</td>
<td>+20%</td>
</tr>
<tr>
<td>Medical assistants</td>
<td>14%</td>
<td>+14%</td>
</tr>
<tr>
<td>Highway maintenance workers</td>
<td>0%</td>
<td>+8%</td>
</tr>
</tbody></table>
<p>They all happen in the physical world, with real people, in places that change minute to minute. A chatbot can write the care plan. It can't help someone out of bed.</p>
<p>That doesn't make these jobs AI-free. The paperwork around them, such as notes, scheduling, quotes and billing, is exactly what AI speeds up. The people who hand it that admin get more time for the part only they can do.</p>
<h2>Is AI actually causing layoffs right now?</h2>
<p>Yes, but less than the headlines suggest, and the trend just shifted.</p>
<p>Employers named AI as the reason for <strong>121,401 US job cuts</strong> in 2026 through October, according to Challenger, Gray &amp; Christmas data reported by HRD America. That's about one in five of all announced cuts this year, and more than double the 54,836 blamed on AI in all of 2025.</p>
<p>October tells a different story. Only <strong>1,265 cuts</strong> that month were put down to AI, the lowest monthly figure of the year. And not everyone takes the AI label at face value. Responding to claims that AI is replacing workers en masse, Lightcast economist Elena Magrini said plainly: "we are not seeing that in the labor data."</p>
<p>The honest reading: some companies are cutting roles because of AI, and others are using "AI" as a tidier word for cost cutting.</p>
<h2>What "exposed to AI" really means</h2>
<p>The scores come from a 2023 study by researchers at OpenAI and the University of Pennsylvania. They took the official task list for every US occupation (from O*NET, the Labor Department's job database) and asked one question about each task: could an AI language model do it in half the time, at the same quality?</p>
<ul>
<li>If a chatbot alone could do it, the task counts fully.</li>
<li>If it needs extra software built on AI, it counts half.</li>
<li>If not, it doesn't count.</li>
</ul>
<p>A job's score is the share of its tasks that qualify. Across all 923 jobs the middle score is 31%. Only about one in five jobs scores 50% or more.</p>
<p>Two limits are worth knowing. The ratings are from 2023, and AI has improved since, so real exposure may be a bit higher today. And "exposed" means AI can speed up the task, not that it takes the whole job. Most jobs are a bundle of tasks, and AI usually takes some of the bundle, not all of it.</p>
<h2>How to protect your job from AI</h2>
<ol>
<li><strong>Find out where you actually stand.</strong> Look up your job in the <a href="https://geekspark.in/tools/ai-careers/will-ai-take-my-job">AI job checker</a>. Guessing tends to swing between panic and denial.</li>
<li><strong>Pick one exposed task and hand it to AI this week.</strong> Learn where it saves you an hour and where it confidently gets things wrong. Spotting those mistakes is a skill on its own.</li>
<li><strong>Grow the part of your job AI is weak at.</strong> Judgment, trust, persuasion, hands-on work and messy real-world problems are the hardest things to automate.</li>
<li><strong>Write down what you do with the time you save.</strong> "I closed 30% more tickets" is a much stronger line at review time than "I used ChatGPT."</li>
<li><strong>If your job is in the at-risk group, start moving now.</strong> Look for a growing job next door that uses the same skills. Executive assistants score 74% and face a 21% drop; project management specialists, who also plan, schedule and keep people on track, score 40% and are forecast to grow 6%.</li>
</ol>
<p>For the wider debate, including the power and water AI runs on, read <a href="https://geekspark.in/blog/ai-tool-not-human-replacement-layoffs-environment">why AI should be a tool, not a human replacement</a>.</p>
<blockquote>
<p><strong>Reader poll:</strong> Has AI already changed how you do your job? <a href="https://geekspark.in/blog/jobs-ai-will-replace">Vote on GeekSpark</a></p>
</blockquote>
<h2>Quick answers</h2>
<p><strong>What jobs will AI replace first?</strong> Office jobs that move information around and are already shrinking: legal and executive assistants, customer service representatives, order clerks and computer programmers. They combine high AI exposure (57% to 76%) with falling job forecasts.</p>
<p><strong>What jobs can't AI replace?</strong> Hands-on, in-person work: physician assistants, home health aides, cooks, and skilled trades such as electricians and HVAC technicians. They all score 15% or less for AI exposure, and most are forecast to grow faster than average.</p>
<p><strong>Will AI replace software engineers?</strong> Not on current evidence. Software developers score 45% for AI exposure, but their jobs are forecast to grow 26% by 2032. Computer programmers, who mostly write code to a spec, are forecast to shrink 11%.</p>
<p><strong>How many jobs has AI replaced in 2026?</strong> US employers blamed AI for 121,401 announced job cuts through October 2026, about 20% of all cuts, based on Challenger, Gray &amp; Christmas data. Only 1,265 of those came in October.</p>
<p><strong>How can I tell if my job is at risk from AI?</strong> Check two things: how much of your job AI can speed up, and whether demand for your job is growing. The <a href="https://geekspark.in/tools/ai-careers/will-ai-take-my-job">free AI job checker</a> shows both for 923 US jobs.</p>
<h2>Sources</h2>
<ul>
<li><a href="https://arxiv.org/abs/2303.10130">Eloundou, Manning, Mishkin and Rock: GPTs are GPTs (2023)</a></li>
<li><a href="https://github.com/openai/GPTs-are-GPTs">GPTs are GPTs: occupation and task data</a></li>
<li><a href="https://www.bls.gov/emp/">US Bureau of Labor Statistics: Employment Projections</a></li>
<li><a href="https://www.hcamag.com/us/specialization/transformation/ai-blamed-for-121401-job-cuts-but-the-labor-data-isnt-buying-it/594129">HRD America: AI blamed for 121,401 job cuts (Challenger, Gray &amp; Christmas data)</a></li>
<li><a href="https://www.onetonline.org/">O*NET OnLine: occupation tasks</a></li>
</ul>
<hr />
<p><em>Originally published on</em> <a href="https://geekspark.in/blog/jobs-ai-will-replace"><em>GeekSpark</em></a><em>.</em></p>
]]></content:encoded></item></channel></rss>