<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Trei Brundrett on Medium]]></title>
        <description><![CDATA[Stories by Trei Brundrett on Medium]]></description>
        <link>https://medium.com/@clockwerks?source=rss-65700098ff3a------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*pmutTwbvRchrY3LzDXFWbw.jpeg</url>
            <title>Stories by Trei Brundrett on Medium</title>
            <link>https://medium.com/@clockwerks?source=rss-65700098ff3a------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 23:06:52 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@clockwerks/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Product & AI update: What we learned from our first AI + News workshop]]></title>
            <link>https://medium.com/american-journalism-project/product-ai-update-what-we-learned-from-our-first-ai-news-workshop-e99c0d942417?source=rss-65700098ff3a------2</link>
            <guid isPermaLink="false">https://medium.com/p/e99c0d942417</guid>
            <dc:creator><![CDATA[Trei Brundrett]]></dc:creator>
            <pubDate>Thu, 19 Sep 2024 17:35:33 GMT</pubDate>
            <atom:updated>2024-09-19T17:35:33.904Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BA8xjeTXEVXR_508ltLZJA.jpeg" /></figure><p><em>This post was written in collaboration with </em><a href="https://medium.com/@mailbackwards"><em>Liam Andrew</em></a><em>.</em></p><p><em>In July 2023, the American Journalism Project launched its Product &amp; AI Studio to explore the smart application of generative AI and other technology within local news, made possible by OpenAI with additional support from the Patrick J. McGovern Foundation.</em></p><p><em>One of the key objectives of the Product &amp; AI Studio is to foster active, collaborative learning, both within our cohort and the broader field of journalism. Here, we share a look at what we learned from a recent workshop on AI and news.</em></p><p>We regularly host virtual cohort meetings and office hours, where grantees share their progress and challenges. These sessions have surfaced great examples of journalism use cases, such as navigating the intricacies of large language model (LLM) prompt engineering, configuring custom GPTs, and formatting documents to enhance LLM responses.</p><p>Given the success of these virtual interactions, we wanted to bring people together for an in-person, hands-on workshop. Our goal was to collaborate directly on these challenges, and we’re excited to share insights from this event.</p><h4><strong>The right environment</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9gG8v0d9vPr6fPtN641-HQ.jpeg" /></figure><p>We hosted this workshop during the annual SRCCON conference, hosted by OpenNews — a network of developers, designers, journalists, and editors who collaborate on open technologies and processes within journalism.</p><p>Our goal with the workshop was to create an inclusive environment where anyone from the organizations we support could bring an AI idea or existing project and make tangible progress with support from others and the Product &amp; AI Studio team. We welcomed participants of all experience levels, from beginners to those who had already learned the tools. Over a dozen people from eight news organizations attended, including a few from outside the AJP portfolio. We were most impressed by the editors who attended with the goal of learning and feeling empowered to help their newsrooms experiment.</p><h4><strong>Key projects and ideas</strong></h4><p>There was a good mix of ideas and projects, primarily attempts to prototype solutions and assess the value of LLMs. Here are a few:</p><ul><li>One group explored image-based AI tools to understand their capabilities and limitations, including built-in bias. This group experimented with this by gathering structured data from a collection of framed photos of University of Minnesota alumni presidents on the walls of the room where we held the workshop.</li><li>Two groups experimented with using LLMs to extract insights from public meeting minutes, which has become a common theme for several AI projects inside and outside our portfolio. The large amount of unstructured data is ripe for identifying potential stories or better organizing civic information.</li><li>Other attendees focused on the basics, learning to write prompts that would return consistent results, which is crucial for understanding the limitations and possibilities of AI tools.</li></ul><p>Supporting a wide range of experience levels proved challenging, but we found that providing access to user-friendly tools, like ChatGPT, made AI experimentation accessible for most attendees. Below you’ll find how we led workshop participants through the session. You can see the full slide deck we shared here:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fdocs.google.com%2Fpresentation%2Fembed%3Fid%3D1Cnd_1L1oUM6yei02X7_oVK-LrjrV4H_fhTIdWZ7chJ8%26size%3Dl&amp;display_name=Google+Docs&amp;url=https%3A%2F%2Fdocs.google.com%2Fpresentation%2Fd%2F1Cnd_1L1oUM6yei02X7_oVK-LrjrV4H_fhTIdWZ7chJ8%2Fedit&amp;image=https%3A%2F%2Flh7-us.googleusercontent.com%2Fdocs%2FAHkbwyKsBBs7Toya8CI8ZQEr_bOf5YDlfBHsO_wr8U6TStiGETkaU26qXmUxF_Skdh7XAjr-2GJ_YWbHkNvWlKMrjRrovDggB_2wdjJCM-xIHsKOXDqhfxzY%3Dw1200-h630-p&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=google" width="700" height="559" frameborder="0" scrolling="no"><a href="https://medium.com/media/302a312a7f29635cba588f36eb025dc3/href">https://medium.com/media/302a312a7f29635cba588f36eb025dc3/href</a></iframe><p>We began the session by emphasizing that this session was about practical use cases for AI.</p><p>It’s easy to spend significant time understanding ethical issues and more existential ideas; For this workshop, we stayed focused on what could be achieved in a single afternoon, which allowed for surprising progress.</p><p>Though there are many tools worth trying, we encouraged participants to start with custom GPTs. This allowed users to test how LLMs handle specific problems, and explore techniques like prompt engineering and retrieval-augmented generation without needing code.</p><p>People built valuable tools in as little as 15 minutes, or completed their prototype in an hour when they thought it would take a week. How often does that happen?</p><h4><strong>Lessons learned: Tips on custom GPTs</strong></h4><p>Throughout the session, we all learned several helpful tips to apply to our work:</p><p><strong>A good prompt goes a long way.</strong> Often, a refined prompt is all that’s needed. We encouraged participants to keep iterating, rather than giving up and assuming something more advanced was needed. Many online resources like Joe Amiditis’ excellent <a href="https://jamditis.notion.site/Beginner-s-prompt-handbook-ChatGPT-for-local-news-publishers-15d1f07d5b194265a41fdd42955679b4">‘ChatGPT for local news publishers’</a>, can help improve prompts.</p><p><strong>Knowledge files are valuable,</strong> but they bring some challenges too:</p><ul><li>GPTs can visit websites mentioned in prompts, but often struggle to do the same within knowledge files.</li><li>PDFs are hard; GPTs convert them into plain text, losing the original formatting. A workaround is to use screenshots of the PDF for better analysis, but the underlying process remains somewhat opaque.</li><li>Knowledge files can become outdated, and rotating data sets doesn’t work well for static files that need to be deleted and re-added. This makes a document-based GPT a great prototyping tool, but not great for scaling or integrating with other systems.</li></ul><p><strong>GPTs are hard to debug. </strong>If your GPT is not answering consistently or correctly, after all tricks have been exhausted, then it might be time to level up.</p><ul><li>Fortunately a GPT offers a solid foundation, and transitioning to use the API, or another LLM, is easier with a draft prompt ready to go.</li><li>There are plenty of low-code tools that can help projects move from the prototype phase into scaling, sharing, and integration.</li><li>Some of these tools take more than an afternoon to set up, but they enable non-developers to take a project from concept to production.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pVLrfZDpeT_F3aKn" /></figure><p>We’re eager to do more workshops in the future, especially in collaboration with other groups doing similar work. Expanding our reach to include more news organizations and a broader range of skill sets will be key to our future efforts. Get in touch if you’d like to collaborate or about this more!</p><p><em>Trei Brundrett is a senior advisor working with the American Journalism Project’s Product &amp; AI Studio and Startups Studio. He is the co-founder of Vox Media and on the board of the Texas Tribune and the Guardian.</em></p><p><em>Liam Andrew is the Technology Lead for the American Journalism Project’s Product &amp; AI Studio where he builds tools and advises newsrooms on product strategy and technical resources. Previously he led the product department at The Texas Tribune, directing technology strategy and overseeing the engineering, design, product and analytics teams.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e99c0d942417" width="1" height="1" alt=""><hr><p><a href="https://medium.com/american-journalism-project/product-ai-update-what-we-learned-from-our-first-ai-news-workshop-e99c0d942417">Product &amp; AI update: What we learned from our first AI + News workshop</a> was originally published in <a href="https://medium.com/american-journalism-project">American Journalism Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Exploring emerging technologies: an update on our Product & AI Studio]]></title>
            <link>https://medium.com/american-journalism-project/exploring-emerging-technologies-an-update-on-our-product-ai-studio-423d4bb38383?source=rss-65700098ff3a------2</link>
            <guid isPermaLink="false">https://medium.com/p/423d4bb38383</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[nonprofit-news]]></category>
            <category><![CDATA[local-news]]></category>
            <category><![CDATA[journalism]]></category>
            <dc:creator><![CDATA[Trei Brundrett]]></dc:creator>
            <pubDate>Wed, 20 Dec 2023 14:43:33 GMT</pubDate>
            <atom:updated>2023-12-20T16:46:56.723Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7S-c9wEcdKvtlpzDryGJFg.jpeg" /></figure><p>This summer, the American Journalism Project launched our new <strong>Product &amp; AI Studio</strong>, a new program to explore the smart application of emerging technologies in local journalism. Our goal through this studio is to ensure that emerging technologies, including artificial intelligence, are leveraged in service of more accessible, high-quality and affordable journalism that serves the public interest. We are <a href="https://www.theajp.org/news-insights/announcements/american-journalism-project-announces-new-partnership-with-openai-to-support-local-news/">grateful to OpenAI</a> for making this studio possible, and to the Patrick J. McGovern Foundation for their additional support we recently received.</p><p>Through the Studio, we’ll help <a href="https://www.theajp.org/our-portfolio/">our portfolio of grantees</a> understand and strategically leverage emerging tech; to do this, we’re bringing in industry experts to give coaching and collaborate on the development of AI-driven tools and applications alongside our grantees and other external partners. We’re also organizing a learning community across AJP’s portfolio to document and share best practices, guidelines and lessons as these experiments unfold — and we’ll work to share these learnings with the journalism industry at large.</p><p>Today, we’re excited to share an update on the progress the Product &amp; AI Studio has made so far. We’re starting our work in the Studio with a set of pilot investments; these direct grants will go to a cohort of 13 of our portfolio organizations, which will receive one- to two-year grants, ranging from $25,000 to $200,000. They are: Centro de Periodismo Investigativo (CPI), Cityside, Civic News Company, CT Mirror, inewsource, The Marshall Project, Montana Free Press, Outlier Media, the Nebraska Journalism Trust, Open Campus, Sahan Journal, Spotlight PA, and THE CITY. This support will enable these organizations to pilot and experiment with different AI applications, and explore more ways to utilize AI’s capabilities.</p><p>The projects we’re supporting through this cohort will allow our grantees to experiment with a wide variety of ways AI can help them solve problems, pursue their missions and serve their communities. This will allow the grantees to learn hands-on where and how to thoughtfully apply AI, to understand both the potential benefits and harms, to mitigate those risks and to become active participants in the conversation about the true opportunity of AI in local journalism. More on some of these projects:</p><ul><li><strong>Centro de Periodismo Investigativo</strong> (Puerto Rico) will launch a pilot project to test AI’s capabilities around translating content from Spanish to English, and vice versa. CPI will also collaborate with linguistic professionals for quality control. Their goal is to develop more bilingual stories more efficiently, advancing their ability to collaborate on investigations with English-language partners in the Caribbean and the U.S. and allowing them to broaden support for investigative journalism in Puerto Rico and expand to serve Puerto Ricans on the U.S. mainland.</li><li><strong>THE CITY </strong>(New York) will experiment with AI-driven tools focused on audience engagement, helping them sift through online information, answer New Yorkers’ questions and receive tips from readers. They’ll also experiment with ways to enable reporters to use AI to analyze data without needing to know how to code. THE CITY will also look at how AI can help them engage with readers who submit reporting tips, prompting their readers with initial AI-generated questions to submit more complete, detailed information, enabling the newsroom to to focus its time and resources on the more complex stages of acting on a tip.</li><li><strong>inewsource </strong>(San Diego) will experiment with AI technology and tools to produce more public records requests more quickly, targeting far more public agencies. They’ll also explore ways AI can help analyze the voluminous documents they receive in response to those requests.</li><li><strong>Outlier </strong>(Detroit) utilizes an SMS platform, which has been central to its distribution of service journalism since its founding. In addition to distributing Outlier’s work, its SMS platform enables 1:1 conversations with Detroiters that often help set Outler’s editorial strategy via answering Detroiters’ questions. Outlier will experiment with integrating AI technology into its platform and workflows in key ways to enable their team to engage more people, better understand their audiences and build deeper, more valuable direct relationships with the communities they serve.</li><li><strong>Cityside </strong>(Bay Area, California) will assess and implement revenue-related AI experiments to build learnings around using AI on the business side of news organizations. Cityside will prioritize experimenting with using AI-assisted communications to develop individual donor relationships across different giving levels.</li></ul><p>In addition to making these investments, we’re also working on building the Product &amp; AI Studio team. We’ve posted the first two roles for this team, which are now accepting applications:</p><ul><li>The <a href="https://www.theajp.org/about/work-with-us/product-partnerships-lead-ai-product-studio/"><strong>Product &amp; Partnerships Lead, Product &amp; AI Studio</strong></a> will be our point person for understanding the needs of nonprofit news organizations and their audiences and businesses, identifying new opportunities for impact with emerging technologies — specifically generative artificial intelligence — and helping launch and test products, and developing relationships with technology and platform partners.</li><li>The <a href="https://www.theajp.org/about/work-with-us/technology-lead-product-ai-studio/"><strong>Technology Lead, Product &amp; AI Studio</strong></a> will have a background in software and data engineering, and will drive technology strategy and development. They will be responsible for understanding and helping our grantee nonprofit news organizations to act upon existing and emerging technologies — specifically AI — by researching, designing, developing, and deploying solutions.</li></ul><p>As we work to advance our understanding of how to best apply these new technologies, we expect to have many lessons to share throughout our portfolio and with the broader industry. We look forward to engaging with others and providing more updates in the coming year! If you’re interested in comparing notes or partnering on this important work, <a href="/service/mailto:hello@theajp.org">please get in touch</a>.</p><p><em>Trei Brundrett is a senior advisor working with the American Journalism Project’s Product &amp; AI Studio and Startups Studio. He is the co-founder of Vox Media and on the board of the Texas Tribune and the Guardian.</em></p><p><em>This piece was originally posted on Dec. 20, 2023 at </em><a href="https://www.theajp.org/news-insights/insights/exploring-emerging-technologies-an-update-on-our-product-ai-studio/"><em>theajp.org</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=423d4bb38383" width="1" height="1" alt=""><hr><p><a href="https://medium.com/american-journalism-project/exploring-emerging-technologies-an-update-on-our-product-ai-studio-423d4bb38383">Exploring emerging technologies: an update on our Product &amp; AI Studio</a> was originally published in <a href="https://medium.com/american-journalism-project">American Journalism Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>