<?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[HumanAI - Medium]]></title>
        <description><![CDATA[HumanAI is a new publication for human stories focusing on the impact of post-modern technology and innovation on our lives. - Medium]]></description>
        <link>https://medium.com/humanai?source=rss----c370fd7d25c1---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>HumanAI - Medium</title>
            <link>https://medium.com/humanai?source=rss----c370fd7d25c1---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 01:58:54 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/humanai" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com" rel="hub"/>
        <item>
            <title><![CDATA[How to Use AI Responsibly Without Outsourcing Your Judgment]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/how-to-use-ai-responsibly-without-outsourcing-your-judgment-7e6684519272?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/2600/0*JYhEsWJc-HkiWVwq" width="5411"></a></p><p class="medium-feed-snippet">An epidemiologist&#x2019;s practical guide to using AI tools in public health, teaching, and writing while keeping human expertise in charge.</p><p class="medium-feed-link"><a href="https://medium.com/humanai/how-to-use-ai-responsibly-without-outsourcing-your-judgment-7e6684519272?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/how-to-use-ai-responsibly-without-outsourcing-your-judgment-7e6684519272?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/7e6684519272</guid>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[honesty]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[writing]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[René F. Najera, MPH, DrPH]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 18:12:31 GMT</pubDate>
            <atom:updated>2026-11-16T18:12:30.438Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Data Brokers Are Tracking Your Visits to the Doctor, Then Selling the Data for Ads]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/data-brokers-are-tracking-your-visits-to-the-doctor-then-selling-the-data-for-ads-56f848f5b28e?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/2600/0*2FzATiMaZUCUyvSN.jpg" width="2600"></a></p><p class="medium-feed-snippet">The Federal Trade Commission told a data company it could no longer sell data it collected from hundreds of millions of Americans&#x2019; mobile&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/humanai/data-brokers-are-tracking-your-visits-to-the-doctor-then-selling-the-data-for-ads-56f848f5b28e?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/data-brokers-are-tracking-your-visits-to-the-doctor-then-selling-the-data-for-ads-56f848f5b28e?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/56f848f5b28e</guid>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[health]]></category>
            <category><![CDATA[data-privacy]]></category>
            <category><![CDATA[data-engineering]]></category>
            <dc:creator><![CDATA[Michael Scaturro]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 16:56:30 GMT</pubDate>
            <atom:updated>2026-11-16T16:56:28.875Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Good Grief — I’ve Just Been Interviewed by AI]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/good-grief-ive-just-been-interviewed-by-ai-061d9f3952c4?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/2600/0*qiiB7QOsSEx1TVw1" width="5120"></a></p><p class="medium-feed-snippet">I spent 30 years asking the questions &#x2014; then an artificial host asked me surprisingly good ones</p><p class="medium-feed-link"><a href="https://medium.com/humanai/good-grief-ive-just-been-interviewed-by-ai-061d9f3952c4?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/good-grief-ive-just-been-interviewed-by-ai-061d9f3952c4?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/061d9f3952c4</guid>
            <category><![CDATA[writing]]></category>
            <category><![CDATA[reading]]></category>
            <category><![CDATA[books]]></category>
            <category><![CDATA[podcast]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Bridget Stella Ruxton Wilson]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 09:11:35 GMT</pubDate>
            <atom:updated>2026-11-16T09:11:34.121Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Constantly Curious: You’re On Your Own, Kid]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/constantly-curious-youre-on-your-own-kid-b48ecb35cf59?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/2600/1*wHCTS4pH1YHvF5Yb5krSEA.jpeg" width="4368"></a></p><p class="medium-feed-snippet">The shifting sands of AI rules, personal privacy, and your Big Tech Brother</p><p class="medium-feed-link"><a href="https://medium.com/humanai/constantly-curious-youre-on-your-own-kid-b48ecb35cf59?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/constantly-curious-youre-on-your-own-kid-b48ecb35cf59?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/b48ecb35cf59</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[big-tech]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[meta]]></category>
            <category><![CDATA[google]]></category>
            <dc:creator><![CDATA[Amy B.]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 20:48:39 GMT</pubDate>
            <atom:updated>2026-11-13T20:48:37.919Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Piling On a Game You Love Over 0.1%: What the Big Walk Backlash Really Shows]]></title>
            <link>https://medium.com/humanai/piling-on-a-game-you-love-over-0-1-what-the-big-walk-backlash-really-shows-16f4bf11dd64?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/16f4bf11dd64</guid>
            <category><![CDATA[ai-ethics]]></category>
            <category><![CDATA[indie-game]]></category>
            <category><![CDATA[game-development]]></category>
            <category><![CDATA[gaming]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[FunVoke]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 15:10:11 GMT</pubDate>
            <atom:updated>2026-11-12T15:10:09.363Z</atom:updated>
            <content:encoded><![CDATA[<p>November 10, 2026. House House, the studio behind Untitled Goose Game, adds a new entry to the FAQ page of its latest game, Big Walk: Did you use AI to make the game?</p><p>The studio’s answer is short: Yes, but very little. That same day, the game gets 366 negative reviews on Steam. The day before, it had one. The next day, Big Walk wins the Critics’ Choice award at the Golden Joystick Awards in London.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pf-ouMeLAJMANkhU.jpg" /><figcaption><em>Big Walk’s key art stacks its round, colorful characters on beach boulders beside the “Big Walk by House House” logo. This is the game whose FAQ update set off three days of argument on Steam. (Image Source: House House)</em></figcaption></figure><p>Some say players were told the truth too late. Others call it an overreaction to a ten-line bug fix. I’m in the second camp. But first, let’s look at what happened over those three days.</p><h3>It started with a jury statement</h3><p>Big Walk is a multiplayer adventure game published by Panic and released on August 4, 2026. It’s on PC, Mac, PS5 and Switch 2, and it had been sitting at a high rating on Steam since launch.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Jh4hRP7AJ3triTKr.jpg" /><figcaption><em>Four players stand together on a coastal rock, one talking into a walkie-talkie while another hauls a giant red object. This shared group walk is what players were reviewing before the AI debate started. (Image Source: Steam / House House)</em></figcaption></figure><p>The debate didn’t start with the game. It started with an awards jury. On November 9, 19 jurors from the Independent Games Festival (IGF) announced they wouldn’t vote for any game that used generative AI in any form. They were speaking for themselves, not for the festival, and the statement didn’t mention Big Walk. But on social media, the conversation quickly attached itself to the game.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rDFl318qXA8p8C3u" /><figcaption><em>Winners raise a trophy on the Independent Games Festival stage. On November 9, 19 of the festival’s jurors said they wouldn’t vote for any game that used generative AI in any form. (Image Source: Independent Games Festival)</em></figcaption></figure><p>House House said it had originally written its AI disclosure for its IGF submission. The next day, the same disclosure went up on the game’s FAQ page. Nobody caught the studio red-handed. The studio wrote it down itself.</p><h3>What the studio said</h3><p>According to House House, AI was used in the final stretch of development to work through difficult technical problems and bugs. The team consulted chatbots and, in some cases, used code snippets from their answers as a starting point before modifying them. By the studio’s own estimate, this touched well under 0.1% of the game’s code. Generative AI wasn’t used anywhere else, including visuals, audio or text.</p><p>The example the studio gave fits that picture. For a shadow bug coming from Unity, they took a chatbot’s suggestion, reworked it heavily, and the final fix came to about 10 lines. That’s one example, not the total usage. The studio didn’t explain how it arrived at the 0.1% figure, and no source has independently verified it. Still, nothing the player sees, hears or reads is part of this debate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7h16vRLgMiqNBR_v.jpg" /><figcaption><em>Two stacked pairs of characters face each other on a beach at sunset, their long shadows stretching across the sand. Shadows are where the studio’s one documented example sits: a Unity shadow bug fixed with about 10 lines of code. (Image Source: Steam / House House)</em></figcaption></figure><p>The word “AI” throws a lot into one bag here. The systems that control enemies are called AI. So are the algorithms that build maps at random. What’s being debated is generative AI, the kind that produces text, images or code. And even that varies: a programmer asking a chatbot for help while tracking down a bug is one thing. A texture the player sees on screen being drawn by AI is another.</p><h3>Why this wasn’t on the store page</h3><p>Steam asks games that use AI to disclose it on their store page. But in January 2026, Valve clarified the form: the disclosure covers content players directly see and hear. Behind-the-scenes tools like code assistants are out of scope. That’s why Big Walk’s Steam page has no AI note.</p><p>So someone buying the game couldn’t have learned this from the store. I don’t see that as the player’s failing. The platform sets the rules and the studio makes the disclosure. Players digging into how every single game was made is neither possible nor their job.</p><h3>What the numbers say</h3><p>According to Steam’s daily review data, the game got 94 positive reviews and 1 negative review on November 9. On November 10, positive reviews rose to 426 and negative reviews to 366. On November 11, there were 652 positive and 130 negative reviews.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PVnS8rw2_4JxcIE7.jpg" /><figcaption><em>Dexerto’s chart of Big Walk’s Steam reviews shows steady blue positive bars through October, then a tall orange negative bar at the far right next to an even taller blue one. Negative reviews spiked, but positive reviews spiked with them. (Image Source: Dexerto / Steam)</em></figcaption></figure><p>On Steam, waves like this, which arrive fast and usually react to an event more than to the game itself, are called “review bombing.” There’s no evidence that the wave against Big Walk was an organized campaign. But one detail stands out: as negative reviews went up, so did positive ones. People defending the game were writing in the same days.</p><p>As of November 12, Big Walk’s overall rating on Steam is still “Very Positive.” 94% of its 31,281 English reviews are positive. Over the last 30 days, that share has dropped to 87%.</p><h3>What people are objecting to</h3><p>Not all the negative reviews say the same thing. In the ones we read, some object to AI itself. Others object to the timing: the game came out in August, the disclosure came three months later, and Steam’s refund window had long since closed. Some players also write that they liked the game but won’t support it anymore.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*J2EOqhw8-EW-B9V4.jpg" /><figcaption><em>Through a player’s binoculars, friends spread across golden grassland while one waves from a rock. Some of the negative reviews say they enjoyed walks like this one; what they object to is how and when they were told. (Image Source: Steam / House House)</em></figcaption></figure><p>The timing complaint is fair. Everyone who bought the game should have known this from the start, and the studio itself admitted that not having an internal AI policy was its own failure. House House said the disclosure wasn’t an endorsement of AI tools, apologized, and promised a more thorough policy.</p><p>Choosing not to support a game is everyone’s right, too. What I find hard to understand is something else: a game you enjoyed and loved playing getting torn apart over a ten-line bug fix that left no trace on screen. There’s no record showing the same people loved the game first and then changed their minds. Steam doesn’t show that. But that scenario, turning your back on something you love overnight, doesn’t strike me as a healthy reflex.</p><h3>When a backlash makes sense</h3><p>Not every backlash is the same. In December 2025, the Indie Game Awards took back the Game of the Year and Debut Game awards it had given to Clair Obscur: Expedition 33. There, placeholder textures made with generative AI had made it into the release version, right where players could see them. The studio had also confirmed in its submission that it didn’t use AI. In Big Walk’s case, according to the studio, the usage was somewhere players can’t see, and it was spelled out in the IGF submission. There was talk on social media that Big Walk had also been disqualified from an award. There’s no official announcement of that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wp42n-PqD7FWvcSN.jpg" /><figcaption><em>Clair Obscur: Expedition 33’s party stands in a flower field beneath floating rocks and a glowing sky. In this game’s case, the AI-made textures were in the art players actually see on screen. (Image Source: Steam / Sandfall Interactive)</em></figcaption></figure><p>Then there’s Domina. In 2022, Dolphin Barn’s gladiator management game Domina was pulled from Steam after the developer posted an announcement targeting a trans streamer and the trans community. Valve said the announcement contained insults targeting another person. The issue there wasn’t how the game was made. It was the developer’s posts attacking people.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xfxknRdp8pAX7PGv.png" /><figcaption><em>Domina’s pixel-art arena shows two gladiators fighting in the sand while a packed crowd watches from the stands. The backlash against this game was about the developer’s posts, not about how this arena was made. (Image Source: Game Developer / Dolphin Barn)</em></figcaption></figure><p>That’s exactly where I draw the line. If a developer disrespects people, a community’s values or a sensitive subject, a backlash is understandable. Piling on a game you love because it got some invisible help in the code is a long way from that line. All three cases get described with the same words, “backlash” and “boycott,” but one is an invisible bug fix, one is undisclosed content you can see on screen, and one is behavior aimed directly at people.</p><h3>What players think</h3><p>There’s no player survey specifically about coding help. For the general picture, there are two data points. According to a Circana survey from December 2025 in the US, a little over a quarter of players say they’d be less likely to buy a game if they learned generative AI was used in its art, dialogue, text or music; most are neutral or undecided. In a Quantic Foundry survey from October to December 2025, made up mostly of core gamers, 85% held a negative view. Both surveys ask about visible content. There’s no separate data on invisible usage like Big Walk’s.</p><h3>Players shouldn’t have to carry this</h3><p>I don’t think these tools can be kept out of game development. That doesn’t make them good. The consent issue in training data and the value of programmers’ work are real debates. But the place to have those debates isn’t punishing a beloved game from a studio that openly disclosed its own usage. It’s platform disclosure rules and studio policies.</p><p>At a time when people complain that stores are filling up with rushed, AI-stuffed games, finding a game that genuinely makes you happy, takes your mind off things and makes you say “this is exactly what I was looking for” is no small thing. Once you’ve found a game like that, I don’t think where ten lines of code came from should change most players’ experience. And as more work gets made with AI, I think work made by hand, with real effort, will stand out even more. That’s an expectation; there’s no data showing it yet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*E0B7RWyGQm7vsHlF.jpg" /><figcaption><em>Two characters sit on a bench at a small station platform as the sun goes down over the hills. Quiet moments like this are what the closing argument means by a game that takes your mind off things. (Image Source: Steam / House House)</em></figcaption></figure><p>Big Walk’s Steam rating is still “Very Positive,” and even during the three harshest days of the debate, positive reviews outnumbered negative ones. As I see it, there’s no reason for players to carry this stress around all the time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=16f4bf11dd64" width="1" height="1" alt=""><hr><p><a href="https://medium.com/humanai/piling-on-a-game-you-love-over-0-1-what-the-big-walk-backlash-really-shows-16f4bf11dd64">Piling On a Game You Love Over 0.1%: What the Big Walk Backlash Really Shows</a> was originally published in <a href="https://medium.com/humanai">HumanAI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Catching a Falling Knife Is Not For the Faint of Heart]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/catching-a-falling-knife-is-not-for-the-faint-of-heart-f8c117c24f87?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/1024/0*ZvNGZ_dT6IHDZBBz.png" width="1024"></a></p><p class="medium-feed-snippet">There&#x2019;s a sickening moment in the life of a high-flying startup when you can feel gravity reasserting its inevitable grip. One moment&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/humanai/catching-a-falling-knife-is-not-for-the-faint-of-heart-f8c117c24f87?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/catching-a-falling-knife-is-not-for-the-faint-of-heart-f8c117c24f87?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/f8c117c24f87</guid>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[John Battelle]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 15:10:04 GMT</pubDate>
            <atom:updated>2026-11-12T15:10:03.033Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[On AI-Generated Text Detectors]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/on-ai-generated-text-detectors-656f5aea8c33?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/2600/1*vkWDFZ7c7WwWZgQqXv7H_A.png" width="2660"></a></p><p class="medium-feed-snippet">&#x201C;Do you use AI to write your articles?&#x201D; is sadly one of the questions I get asked most often. I will be honest. Yes. But no more than I&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/humanai/on-ai-generated-text-detectors-656f5aea8c33?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/on-ai-generated-text-detectors-656f5aea8c33?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/656f5aea8c33</guid>
            <category><![CDATA[ai-writing]]></category>
            <category><![CDATA[ai-detection]]></category>
            <category><![CDATA[essay-writing]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[writing-tips]]></category>
            <dc:creator><![CDATA[Marco Capriz]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 15:09:56 GMT</pubDate>
            <atom:updated>2026-11-12T15:09:54.966Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Human-AI Collaboration: You Can’t Disentangle What Was Built to Conjoin]]></title>
            <link>https://medium.com/humanai/human-ai-collaboration-you-cant-disentangle-what-was-built-to-conjoin-fa40e054e73a?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/fa40e054e73a</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[human-ai-collaboration]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-literacy]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <dc:creator><![CDATA[Barbara Roy]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 15:09:25 GMT</pubDate>
            <atom:updated>2026-11-12T15:09:24.340Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/940/0*JMssR7pA8J0JnmZE" /></figure><p>For decades, we framed artificial intelligence as a tool. A calculator. A spell-checker. A search engine. A pipeline component. Something a human picked up, used for a task, and put back down again.</p><p>That frame made sense when the interaction was discrete enough to observe the boundary. Human input. Machine operation. Human receives output.</p><p>But modern AI-human collaboration, particularly through language models, has broken that frame, and we’re still trying to evaluate what happens inside it using vocabulary that assumes the boundary remains intact.</p><p>It doesn’t.</p><p>I use language-model-based AI daily as part of my work. I was already a proficient writer before I began doing so, naturally and professionally. But over the past couple of years, AI has raised the altitude of my thinking, which has consequently raised the caliber of my writing.</p><p>That’s not because AI “taught me how to write,” but because sustained interaction with AI changed the environment in which my thinking develops.</p><p>During that same period, I moved into founder-level research that required learning and practicing formal standards of evaluation and documentation. I built an investigative consumer-protection platform. I developed disciplines, theoretical frameworks, technical architectures, research papers, consulting services, and the language necessary to explain all of those things to very different audiences.</p><p>Somewhere in there, the familiar question of whose writing is this? stopped being particularly useful.</p><h3>The Trouble With Separating a Two-Part Equation</h3><p>Much of the current conversation about AI-assisted writing rests on an assumption that the finished work can be reverse-engineered into two piles, like:</p><ol><li>The human contributed this.</li><li>The AI contributed that.</li></ol><p>Maybe the human supplied the ideas and the AI supplied the phrasing. Maybe the AI generated a draft and the human edited it. Or, maybe this sentence “sounds like AI,” while that paragraph sounds authentically human.</p><p>But what happens when the human learns through the interaction, or the human’s vocabulary expands? What happens when repeated attempts to explain a difficult concept expose weaknesses in the human’s own thinking? What about when the AI’s misinterpretation forces the human to become more precise, or the human becomes better at recognizing conceptual drift because they’ve spent hundreds of hours correcting it?</p><p>Now try to separate the contributors. You can’t. At this point, you’re no longer examining two independent entities contributing discrete outputs. You’re attempting to dissect an <strong>emergent system whose interaction changes what becomes possible inside the system</strong>.</p><h3>Technology Was Supposed to Extend Human Capability</h3><p>This is the part of current human-AI conversations I find particularly… strange. The point of technology has always been to extend human capability. So, we produced technology capable of extending human cognitive work and, what immediately followed was preoccupation with figuring out which capabilities still “belong” to the human.</p><p>What did the AI teach you?</p><p>Which phrase came from AI?</p><p>Was that structure yours?</p><p>Would you have had that thought without it?</p><p>Did AI influence your writing style?</p><p>Of course it did.</p><p>So did every editor I’ve ever worked with; every book I’ve read; every professional environment I’ve worked inside; every argument that forced me to clarify my position; and every audience that misunderstood something I thought I had already explained perfectly well.</p><p>I understand this from the other side, too. Earlier in my career, I ghostwrote content for leaders whose names ultimately appeared on the work. One manager once told me, “You make me sound so smart!” She was smart. The thinking was hers; I could simply articulate it in print more effectively than she could.</p><p>In another role, the boundary wasn’t nearly as clean. My manager wasn’t a strong English-language communicator, and sometimes the thinking she gave me wasn’t, shall we say, fully developed. In making it coherent enough to communicate, I sometimes extended the reasoning itself.</p><p>We didn’t have an existential crisis about authorship every time this happened. We understood that articulated thought could be collaborative, and that articulation itself could contribute something intellectually meaningful.</p><p>Human cognition has never developed in isolation. The difference now is that <strong>AI makes the interaction unusually visible, persistent, responsive, and fast</strong>. And because we can see the interaction, we mistake visibility for separability.</p><h3>My Writing Isn’t All That Has Changed</h3><p>My writing certainly has changed. Interestingly, some of the greatest changes haven’t resulted from switching between research papers, social media, newsletters, consulting materials, and investigative writing. They’ve come from repeatedly watching people misunderstand the ideas themselves.</p><p>Seeing the same misconception for the fifth time changes how you explain something the sixth. I’ve become more direct and less tolerant of inherited terminology that obscures mechanisms rather than describing them.</p><p>I’ve learned where readers reliably collapse distinctions I consider important, and which assumptions need to be dismantled before a new concept can even be introduced.</p><p>AI has participated in that evolution because I use it while thinking through these problems. But believing AI therefore “wrote” the resulting work gets the relationship exactly backward. The interaction helped change <strong>me</strong>. <strong><em>There is no clean version of today’s writer that exists somewhere outside that history, waiting to be compared with the AI-influenced one.</em></strong></p><h3>The Unit of Meaning Is the System</h3><p>This is why I established in the <a href="https://catchproof.square.site/ai-systems-literacy-manifesto">AI Systems Literacy™ Manifesto</a> that AI-human collaboration is not divisible into parts. It is an emergent system. The human brings intention, stakes, judgment, experience, purpose, and direction.</p><p>The AI brings computational capability, pattern range, responsiveness, and an extraordinary capacity to operate within the conceptual environment the human establishes.</p><p>Neither contribution disappears, but the resulting meaning does not arrive with little ownership labels attached to every component.</p><ul><li>The human establishes the conceptual altitude.</li><li>The AI operates within it.</li><li>The human evaluates what comes back.</li><li>That response changes the next interaction.</li><li>The AI responds again.</li></ul><p>The human corrects, rejects, incorporates, reframes, questions, or extends. And around we go. Eventually, asking which one wrote this sentence? can become considerably less informative than asking, <strong>What system produced this work, and who was steering it?</strong></p><p>That is a provenance question worth asking.</p><h3>Steering Is Human Work</h3><p>In <a href="https://catchproof.square.site/ai-systems-literacy">AI Systems Literacy™</a>, <strong><em>steering</em></strong> <strong>is defined as the human function of maintaining coherence within an emergent human-AI system. </strong>Steering isn’t prompting. It isn’t telling an AI whether you prefer short paragraphs or fewer adjectives. It is the continuing work of maintaining the intended conceptual altitude of an interaction.</p><p>AI can produce something articulate that is nevertheless subtly wrong. It can follow a familiar pattern straight out of the conceptual territory you were trying to occupy while reinforcing an assumption that should have been questioned.</p><p>The human has to recognize that movement and correct it.</p><p>That requires understanding the subject well enough to know that the system has drifted. This is also why I describe successful AI-human collaboration in terms of <a href="https://github.com/AICognitiveFrontier/stable-state-responsive-alignment/blob/main/diagrams/SSRA-Theorem.md#interaction-level-stability">interaction-level stability</a>. The coherence we experience doesn’t belong entirely to the model or entirely to the human. <strong>It arises from the system formed through their interaction. </strong><em>(See </em><a href="https://github.com/AICognitiveFrontier/stable-state-responsive-alignment/blob/main/diagrams/SSRA-Theorem.md#interaction-level-stability"><em>GitHub: Reciprocal Loop Theorem</em></a><em>, and research, </em><a href="https://zenodo.org/records/21304232"><em>Stable‑State Responsive Alignment™: The Missing Layer in Human-AI Collaboration</em></a><em> for more.)</em></p><p>The model provides capability.</p><p>The human provides direction.</p><p>And, stability has to be maintained between them.</p><h3>AI Doesn’t Own the Why</h3><p>None of this gives AI agency. Quite the opposite. AI has no intention behind my work. It has no career to advance, argument to win, reputation to protect, research agenda to pursue, or reason to care whether the thing we’re producing exists tomorrow.</p><p><strong>I do.</strong></p><p>That’s why the human remains responsible for the why.</p><p>Those aren’t incidental contributions to authorship. They are the conditions that make the work purposeful in the first place.</p><p>AI contributes capability, not identity. It can extend cognition without becoming the person whose cognition is being extended.</p><h3>Are We Asking the Wrong Question?</h3><p>There are absolutely circumstances wherein AI use should be disclosed. There are contexts where provenance matters enormously. There are legitimate questions about accountability, evidence, source integrity, intellectual property, and whether someone actually possesses competence they claim to possess. Those questions shouldn’t disappear.</p><p>But they also shouldn’t be collapsed into the much cruder question:</p><p>Which words did the AI write? Because once human and machine have participated in an extended cognitive process, the influence does not stop neatly at the boundary of the generated text. The human leaves the interaction carrying something else forward:</p><p>A distinction.</p><p>A formulation.</p><p>A question.</p><p>A better way of explaining something.</p><p>A recognition that the original premise was wrong.</p><p>Sometimes, an entirely different way of thinking.</p><p>And that version of me brings those changes into tomorrow’s work.</p><p>So, yes, AI influences my writing; my writing influences how I interact with AI; those interactions influence my thinking; my thinking influences the systems I build; and the systems I build give me new things to think about. And then, tomorrow, I show up slightly different from the person who showed up today.</p><p>Good luck disentangling that.</p><blockquote>Perhaps we should stop trying to separate the contributors we deliberately brought together, and start learning how to understand the system that emerges when they do.</blockquote><p><em>Proceed with Clarity.</em> 🧭</p><p><em>Originally published at </em><a href="https://www.linkedin.com/pulse/human-ai-collaboration-you-cant-disentangle-what-built-barbara-roy-2kude"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fa40e054e73a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/humanai/human-ai-collaboration-you-cant-disentangle-what-was-built-to-conjoin-fa40e054e73a">Human-AI Collaboration: You Can’t Disentangle What Was Built to Conjoin</a> was originally published in <a href="https://medium.com/humanai">HumanAI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GPT-6 Cyber: OpenAI’s Security Model Explained]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/humanai/gpt-6-cyber-found-81-firefox-flaws-now-comes-the-hard-part-b1063fa60675?source=rss----c370fd7d25c1---4"><img src="https://cdn-images-1.medium.com/max/1672/1*N-r4JWGzFjeJHAsgTUb32g.png" width="1672"></a></p><p class="medium-feed-snippet">OpenAI&#x2019;s GPT-6 Cyber reported 81 Firefox flaws and 948 Linux exploits. Explore the evidence, benchmarks, costs, and access limits.</p><p class="medium-feed-link"><a href="https://medium.com/humanai/gpt-6-cyber-found-81-firefox-flaws-now-comes-the-hard-part-b1063fa60675?source=rss----c370fd7d25c1---4">Continue reading on HumanAI »</a></p></div>]]></description>
            <link>https://medium.com/humanai/gpt-6-cyber-found-81-firefox-flaws-now-comes-the-hard-part-b1063fa60675?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/b1063fa60675</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Halil Özel]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 15:09:19 GMT</pubDate>
            <atom:updated>2026-11-12T15:09:17.755Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The self-driving laboratory]]></title>
            <link>https://medium.com/humanai/the-self-driving-laboratory-297c64a11d59?source=rss----c370fd7d25c1---4</link>
            <guid isPermaLink="false">https://medium.com/p/297c64a11d59</guid>
            <category><![CDATA[biology]]></category>
            <category><![CDATA[wet-lab]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[research]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Mohd Younus Bhat]]></dc:creator>
            <pubDate>Sat, 07 Nov 2026 10:52:02 GMT</pubDate>
            <atom:updated>2026-11-07T10:52:01.453Z</atom:updated>
            <content:encoded><![CDATA[<p>A paper published in <a href="https://www.nature.com/articles/d41586-026-03552-8"><em>Nature</em></a> this week describes a biology laboratory in Shanghai that works almost without being touched. The system, called AutoDNA, was built by a team led by computer scientist Haibo Chen at Shanghai Jiao Tong University.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wiKqbEbvmrVLZD4ShdBnnQ.png" /><figcaption>A modern, biological spin on Robert Leighton’s classic 2023 <em>New Yorker</em> autocomplete cartoon, capturing the new era of autonomous AI agents entering wet labs. Created with the assistance of OpenAI ChatGPT (GPT-6), 7 November 2026.</figcaption></figure><p>It runs more than twenty kinds of instruments, moves samples between them with robotic arms, and designs its own experiments from a single sentence typed by a person. Human involvement is limited to refilling reagents, restocking consumables, and servicing hardware that breaks down.</p><p>To test the system, Chen’s team asked it to write a short quotation into a strand of DNA, using the four genetic letters A, C, G and T. They chose a line often, and wrongly, attributed to Stephen Hawking: “Intelligence is the ability to adapt to change.” AutoDNA synthesised the DNA, then switched to separate sequencing equipment to read it back and confirm the message matched. The exercise required 11,382 distinct operations spread over roughly eight days. The only human contributions were replenishing buffers and tubes and replacing 108 pipette tips. In a separate test, the same system detected a cancer-linked mutation in a DNA sample on its own.</p><p>What distinguishes AutoDNA from the lab automation that biotechnology companies have used for decades is where the decision-making sits. Chen has said that in most existing systems, “human experts are still required to configure the instruments,” with AI handling only one stage of the process. His team instead built the hardware and software agents together from the outset, so the system itself figures out how to route an experiment across whatever instruments are available, rather than executing a protocol a person wrote in advance. Milad Abolhasani, a chemical engineer at North Carolina State University who had no part in the project, called it remarkable for its autonomy and for how readily it absorbs different kinds of equipment. “The future of science is going to involve a lot of autonomous labs,” he told <em>Nature</em>.</p><p>That prediction already has evidence to back it up. On February 5 this year, OpenAI and the Boston biotechnology firm Ginkgo Bioworks announced that they had let GPT-5 design and run a six-month campaign of cell-free protein synthesis experiments inside Ginkgo’s automated laboratory, with no person choosing the next reaction to try. Across 36,000 reaction compositions and 580 microplates, the system brought down the cost of producing a standard test protein by 40%, from $698 to $422 a gram, according to OpenAI’s own account of the work. Ginkgo has since told investors, in a filing with the US Securities and Exchange Commission in late February, that it is expanding its Boston facility past fifty automated carts this year, and it separately won a $47 million contract to build a 97-instrument autonomous laboratory for the Pacific Northwest National Laboratory in the United States. At the SLAS automation conference in Boston the same week, more than a dozen companies showed competing versions of the same idea: software that plans an experiment, and hardware that carries it out without waiting for a technician.</p><p>Reshma Shetty, Ginkgo’s co-founder, told <em>Nature</em> that AutoDNA’s work was a useful proof of principle, while wondering whether such a system would hold up once it has to operate costlier and fussier instruments, such as the mass spectrometers used to analyse proteins. That is a fair question. Everything AutoDNA has demonstrated publicly so far involves DNA in solution: no cells, no living organisms, nothing that behaves unpredictably the way a growing culture does.</p><p>Not every scientist watching this is reassured by the achievement itself. Julius Lucks, a synthetic biologist at Northwestern University, said AutoDNA’s independence worries him more than it impresses him: “You set up this operation, and then you’re not in the loop. You can’t keep track of it. It’s really important to have the human in the loop.” Chen appears to accept the force of that objection. He has said that before AutoDNA is used on living systems, it will need “a dedicated biosafety and biosecurity plan with additional guardrails, including explicit human-authorisation steps, and the ability to monitor and pause experiments in progress.” That is an unusual thing to volunteer in a paper announcing a result, and a useful one: it names a real gap in the current system rather than gesturing at safeguards yet to be designed. The team has also made its code public on GitHub and its sequencing data and agent outputs available on Zenodo, which at least allows other laboratories to verify the claims rather than take them on faith.</p><p>None of this week’s reporting claims that an AI agent can now design something dangerous on request, and nothing in the <em>Nature</em> paper or the OpenAI release says otherwise. What the two announcements reliably show is that the scale of unsupervised experimentation is climbing quickly: thousands of operations in a Shanghai laboratory this year, tens of thousands of reactions in a Boston laboratory, and a near-hundred-instrument system already under contract for a national laboratory in the United States. If that pace continues, the next five years are more likely to bring a steady spread of such systems into ordinary biotechnology and pharmaceutical work than one dramatic breakthrough. The questions that follow will be less dramatic than “robot scientist” headlines suggest, but no less consequential: who reviews what an AI agent decided to run overnight, how orders for synthetic DNA get screened for sequences resembling known pathogens once the ordering itself is automated, and whether the manual skills a graduate student currently spends years acquiring will still be worth teaching.</p><p>That last question bears directly on India’s own plans. The Union Cabinet’s BioE3 policy, approved in August 2024, commits the Department of Biotechnology to building a national network of biofoundries and “Bio-AI hubs” intended to bring AI-driven biomanufacturing within reach of Indian researchers and startups, backed by a Bio-RIDE funding allocation of roughly ₹9,197 crore. Whether those hubs eventually run systems designed in India, or simply lease equipment and software from Ginkgo, OpenAI or Chinese developers such as Chen’s team, will determine how much of this capability, and the oversight that has to accompany it, India actually controls rather than imports along with the hardware.</p><p>The sensible response to AutoDNA is neither to wave it off as a publicity exercise nor to treat it as the end of the bench scientist. It is a working demonstration, reported in a peer-reviewed journal, with its code and underlying data released for scrutiny, showing that AI agents can now run real laboratory equipment from start to finish on written instructions. Lucks’s unease, and the guardrails Chen himself says are still missing, deserve as much attention in the coverage as the achievement does. A laboratory that needs no hands still needs someone minding it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=297c64a11d59" width="1" height="1" alt=""><hr><p><a href="https://medium.com/humanai/the-self-driving-laboratory-297c64a11d59">The self-driving laboratory</a> was originally published in <a href="https://medium.com/humanai">HumanAI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>