> ## Content Index > Fetch the complete content index at: https://liverickson.com/llms.txt > Use this file to discover other available public pages before exploring further. # Authentic Attention in the Agentic Age - URL: https://liverickson.com/authentic-attention-in-the-agentic-age/ - Published: 2026-09-17T16:00:15.000Z - Updated: 2026-09-17T16:00:15.000Z - Description: Capturing attention online has, perhaps, never been easy. How do we responsibly bid for attention in an age where agents are flooding the inboxes of the people we may want to reach? - Author: Liv Erickson - Tags: Business & Leadership, Communication, Machine Learning, Philosophy, Random Thoughts How do we responsibly bid for attention in an age where agents are flooding the inboxes of the people we may want to reach? Arvind Narayanan, a professor at Princeton University who often speaks out about artificial intelligence, has been writing about AI floods as perhaps "the most under-recognized AI harm". [Floods might be the most under-recognized AI harm. By AI flood I mean overwhelming communication channels, forcing them to shut down or imposing a huge time cost on recipients. Three reasons we don’t… | Arvind NarayananFloods might be the most under-recognized AI harm. By AI flood I mean overwhelming communication channels, forcing them to shut down or imposing a huge time cost on recipients. Three reasons we don’t talk about them much: 1) There are 50+ kinds of floods, so it registers as 50 different problems instead of one huge problem. https://lnkd.in/gEbeVqqG (And no one is affected by more than a few of these, so it’s hard to grasp how pervasive it is.) 2) The harms are usually diffuse. For example, people stop opening cold emails. There’s no news story there. 3) Similarly, the harms are harder to explain. AI agent swarm hacks are legible and vivid. But as I said in the quoted post, “there was something magical about the egalitarian potential of the internet, and we lose something when we go back to clubby hierarchies.” That’s a profound shift, but it only becomes clear after we sit with it for a bit. And again, that’s just one of 50+ floods. A semi-unified response to the “flood of floods” might be critical for protecting societal resilience.![](https://liverickson.com/content/images/icon/al2o9zrvru7aqj8e1x2rzsrca-9164e5df-e3a4-4029-b0d3-5842cd8580f5)LinkedInArvind Narayanan![](https://liverickson.com/content/images/thumbnail/1789556233769-7d9d6099-824d-4936-a907-b1eb891e8ae9)](https://www.linkedin.com/feed/update/urn:li:activity:7505942874904244224/?ref=liverickson.com) His post follows on his previous commentary about how AI-generated outreach has impacted the PhD admissions cycle for 2027, with only 3 of 73 emails received human-written. Dr. Narayanan writes about how cold outreach helped him get where he was today, but that there is a profound shift happening that begins to dismantle the egalitarian potential of the internet when we consider how effective today's AI agents are at taking up this space. --- I recently decided to apply for a volunteer board role for an organization I deeply respect, and found myself curious both about authenticity and attention. Like many senior-level technology professionals, my education and work history is extensive and varied. I've worn product management, engineering leadership, and community management hats. I'm going into my eighth year with Mozilla, but I've also built a non-profit and write a lot of code in my free time. How do I best match my experience to a new potential opportunity with brand new humans in 2026 without losing my humanity, or running out of time? Step 1: Send the job listing to my closest tech-industry friends and ask them if I'm totally off-base for thinking that I'm remotely qualified for the role. I've never been on a board, but everyone has to get experience for the first time somewhere, right? I am reassured that it's not a waste of the organization's time for me to apply, so I move on to Step 2\. Step 2: Sift through my old CV, LinkedIn profile, and various assortments of my resume. I recently deleted my Canva account because the product began pushing too many crappy AI tools into the design process, so I find the .docx version of my CV - a 13 page point of reference file with virtually everything I've done, written in the longest possible way - and oscillate between feeling confident and despondent about my chances of being considered. Step 3: Consider how AI might be able to help me. This is obviously where the most tension comes into play. Is there an ethical and honest way to use AI to help me tailor my experience for this new opportunity? I turned to Ori to explore. 🖥️ In case you're new here, Ori is my fully-local AI machine. When I talk about using AI, in this post and others, I am referring to what I'm able to build on my own computer. Why? "[Slow AI](https://leonardo.info/criptech/slow-ai?ref=liverickson.com)" addresses the energy, data center, and privacy issues with mainstream AI - and continued investment in local computing makes it an increasingly viable option for individuals and organizations alike. Step 4: Build a tool and do the thing. --- ![](https://liverickson.com/content/images/2026/09/image.png) SiftCV, a markdown renderer with agentic workflow instructions for personalizing resumes ## Building SiftCV A few months ago, I found myself struggling to create a slide deck for a presentation that I was giving at Columbia Business School's 'Foundation of Entrepreneurship' class. Rather than try and build a narrative deck-first, I dreamed about a tool that would let me write my slides in markdown and handle the conversion to a deck automagically. I built a personal tool - Prism - to do just that, and the result was intangible. I felt far more confident and comfortable presenting than I ever have, because the tool itself was an extension of and complement to the material. ![](https://liverickson.com/content/images/2026/09/image-1.png) Prism in 'editing' mode For me, this is part of the power of AI code generation: it allows me to blur the boundaries between tool and medium. The presentation - a personal narrative of product development and innovation 'intrapreneurship' at one of the oldest internet companies - flourished in part because the software Claude Code had generated was perfectly reflective of what *I* needed *this* presentation to have. One principle that emerges from all of the AI hype - and stands on its own as an area rich for innovation and disruption - is the way that we as consumers view data. Over the past several decades, commercial software has largely abstracted data ownership and literacy away from the end user. These moats - be it the platform itself or the infrastructure on which the data lives - have resulted in both a huge transfer of ideas, relationships, and culture from our own hands and into the pockets of major corporations. These companies benefit from keeping software extremely easy to use (to the point of literal addiction, in some cases), locked down, and hard to build. This, in turn, means that basic data literacy and architectural principles are not often deployed when we think about how to consider our lives. 💾 When I talk about 'data', I'm not just talking about quantifiable data. I'm talking about our statuses, our stories, our relationships, our photos - anything that can be digitized. Maybe with AI, anything can be 'quantifiable' these days, but data - in this sense - is just as much lived culture, preserved on disk to hopefully outlast us. We're not metrics to be maximized, but we deserve the right to our own [digital prosthetics](https://www.blinkpopshift.com/projects/prosthetic-memory?ref=liverickson.com) free from the eyes of capitalism. Building off of the [Data Introspection Project](https://misslivirose.github.io/data-introspection/?ref=liverickson.com) principle of data-as-architecture, I used [Ori](https://liverickson.com/being-a-pregnant-developer-in-the-age-of-ai-is-weird/) to break my existing CV down from a .docx file to a markdown file. The resulting output was a file where each line could be individually referenced by a future agentic call (maybe at some point I'll convert my CV into a proper sqlite database, but for now, this suffices). ``` Attached to this chat are documents that explain who I am - the is the baseline source of truth for my job history, but . Convert the attached résumé (docx or pdf) into a single, complete markdown file named -cv-complete.md. This is a SOURCE OF TRUTH FILE: it must capture EVERYTHING from the primary source of truth CV, with no page limit and no editing for length. ``` A sample prompt, representative of what I used I then had Ori - via Visual Studio Code and the Zoo Code extension - create a simple schema that could be used for rendering the markdown in a particular way: ``` Requirements: 1. Use exactly one `#` heading for the name, then a single paragraph of contact info (city · email · phone · website · github), then `---`. 2. Every major section must be a `##` heading (e.g. "Summary", "Experience", "Writing", "Governance", "Education", "Skills") so each can be used as a page-break anchor in the SiftCV app. 3. Use `**bold**` for job titles / organization names, and `-` bullet lists for accomplishments. No tables, no images, no HTML. 4. Keep it factual and line-for-line traceable to the source document. After each bullet or paragraph, add an HTML comment in this exact form: These comments are "proof": they let me verify every line came from my source résumé, and the renderer keeps them hidden (they survive as real HTML comments). ``` Then, I used a second prompt workflow to compare my source-of-truth CV file to the volunteer description, had the agent select the most relevant information and re-structure the information: ``` Here is my complete CV markdown file and the job description for at . Produce -cv-.md, a synthesized version of my CV that: 1. Reuses lines from the master file verbatim wherever possible — do not invent new claims. You may rephrase a source-of-truth-file line, but keep its REF comment and extend it with why the line was adjusted. 2. Select the most relevant experience for this role; drop or condense the rest. In the response, explain what was an intentional omission from the opportunity and why. 3. Reorder sections so the strongest relevant material appears first. 4. Rewrite the Summary paragraph (2–3 sentences) for this specific role, grounded only in source-of-truth-file content. 5. Target roughly two pages at 10 pt serif / 0.6 in margins worth of content ``` The output is a human and machine-readable version of my resume, where every line is grounded in the original source. I (the human) can now make edits in a central, single place, and use that file to generate as many slices (ew, questionable word?) as my professional self as appropriate. Given that I have no plans to leave [Mozilla Data Collective](https://mozilladatacollective.com/?ref=liverickson.com) any time soon, I'm not using it for job hunting, but I think it's an interesting data introspection tool when I want to analyze my professional history in service of volunteer opportunities or future academic endeavors. 📖 I'm almost finished reading [Selfless: The Social Creation of "You"](https://www.brianloweryphd.com/writing?ref=liverickson.com) by Stanford Professor Brian Lowery, PhD, and this is strongly influencing my moment-in-time interest in 'slices of the self' via AI and reclaiming our digital footprints. The (in my opinion) least interesting part of the process was having Ori generate the actual renderer. With just a few back-and-forth iterations, I had a perfectly usable renderer with the bare set of features I need. Does it look like the vibe-coded app it is? Yes, absolutely. Did I half-heartedly try to get it to look more like a proper document editor with better information architecture and familiar layout? You bet I did. Did the baby wake up and I decided to YOLO push to GitHub? Sure thing! Am I now convinced I can vibe-code a 'clean girl' markdown-based Microsoft Office replacement? Hell yeah... some day. ![](https://liverickson.com/content/images/2026/09/image-2.png) A snapshot of the result --- ## Can Agentic Workflows be Authentic? There is a fair amount of rightful criticism about the use of AI in the hiring and outreach pipeline. More broadly, wide-scale use of agentic AI builds upon social media's influence and further questions the extent to which technology companies and their algorithms shape our connections and perceptions of our selves. What does it mean to be authentic in the age of agents? We can look to the fashion industry for a lesson in authenticity. Clothes-as-art-and-history present one option for being opinionated with your place and status in the world, and sharing your world view. 🗽 The New Yorker has an [incredible piece this week on Hermès](https://www.newyorker.com/magazine/2026/09/14/how-hermes-unharnessed-fashion?ref=liverickson.com#rid=ee29a538-5ddf-453c-9864-38b2e3019a89&q=herm%C3%A8s) and the way that the fashion house has survived generations of change. If there's one thing that I took away from The Devil Wears Prada 2, it's that Vogue can save us all from the likes of Big Tech. Note: the cover image of that New Yorker story is fairly NSFW. Artists, in general, have the role of shaping cultural expectations, beliefs, and norms - but have been largely cast out of the process of building digital tools. There is space for art within the digital realm - even more so when we embrace digital media that is integrated with the physical - and our use of AI should look more like building our paintbrushes than generating the painting. ## We Cannot be Connected without Attention Capturing attention online has, perhaps, never been easy. Now, the fragmented internet populated by bots makes it more challenging than ever to sift through the noise and find genuine connection. The rise of the attention economy creates perverse incentives to online influence, while both agentic and generative AI extend the capacity for both authentic and spurious expression. It is, perhaps, worth noting that the original, Biblical flood was sent as both punishment and warning. The flood of AI-generated content as a stand-in for authentic connection may not be as immediately catastrophic, but, as Narayanan so clearly notes - it does demand a cost. We cannot be connected without attention. Agent-to-agent connection is not the same as human-mediated evaluations of another person's self. In a world where computer programs can increasingly pass as humans, Philip Rosedale, founder of Second Life, has stated that our social relationships may become one of the only way of verifying someone's identity. Yet, if Mark Zuckerberg has his way of replacing your friends with AI agents, we might be looking at a grim, anti-social future where our identities are more shaped by Dario Amodei and Elon Musk's software programs than the people around us. ## At the end of the day, the choice isn't ours I wrestle with whether or not to disclose the use of AI in my board seat application. After all, I don't know the people on the other side yet. I know their organization and mission, and it's one that I am deeply aligned with and would love to support. I want to believe that there is space for using these tools without replacing the heart and passion and *self* that I might be able to offer. But is there space for this nuance in 2026? I want to believe there is. *You can try out SiftCV yourself at* [*https://misslivirose.github.io/siftcv/*](https://misslivirose.github.io/siftcv/?ref=liverickson.com)