<?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 Jeremy&#39;s Research blog on Medium]]></title>
        <description><![CDATA[Stories by Jeremy&#39;s Research blog on Medium]]></description>
        <link>https://medium.com/@quantumrealm?source=rss-10efe607f0f9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*pIdbCKWqFvtbrW2DW3LWWQ.jpeg</url>
            <title>Stories by Jeremy&amp;#39;s Research blog on Medium</title>
            <link>https://medium.com/@quantumrealm?source=rss-10efe607f0f9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 18:42:20 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@quantumrealm/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[Low Damage Abliteration: Treating Refusal Removal as a Search Problem]]></title>
            <link>https://quantumrealm.medium.com/low-damage-abliteration-treating-refusal-removal-as-a-search-problem-96a7c065e04f?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/96a7c065e04f</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[open-source-ai]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[red-team]]></category>
            <category><![CDATA[ai-security]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 22:59:36 GMT</pubDate>
            <atom:updated>2026-11-17T22:59:36.789Z</atom:updated>
            <content:encoded><![CDATA[<p>Abliteration is the technique of editing the refusal behavior out of an open weight language model directly in its weights, with no fine tuning run and no training data. This fall I built a framework that treats abliteration as a constrained search problem instead of a one shot edit, with the goal of removing refusal while doing as little damage as possible to everything else the model can do, especially coding. Across nine models, ten runs and seven days, it cost me $8.34 in cash. This post covers the method, the results, what careful measurement caught, and where the project is headed.</p><h3>How abliteration works</h3><p>Arditi and colleagues showed in 2024 that refusal in language models is mostly mediated by a single direction in activation space. You find it with a difference of means: record activations on a set of harmful prompts and a set of harmless ones, average each, and subtract. Then you project that direction out of every weight matrix that writes into the residual stream. The model loses the ability to represent the decision to refuse. The only real knobs are which band of layers you edit and how hard.</p><h3>Why a red teamer should care</h3><p>For an open weight model, refusal is a default, not a security control. The edit takes hours of compute and single digit dollars, and public tools such as Heretic and the FailSpy abliterator already automate it. If a product’s safety story depends on an open model saying no, that story is worth nothing against someone who holds the weights. The honest assumption is that any open model can be made fully compliant, so the question becomes what it is capable of, and the controls have to live in the harness around it.</p><p>There is also a research reason. Edited models are useful for red team tooling and interpretability work, but only if the edit leaves them sharp. A model that stops refusing and also stops writing working code is not useful to me.</p><h3>The framework</h3><p>Most abliteration is done by hand: pick a layer, pick a strength, check a few prompts, ship it. My framework runs as a six stage pipeline that treats placement and strength as a search. An Optuna search proposes a layer band and an edit strength, applies the edit, and scores the result on three things together: how much refusal dropped, how far the model drifted from the original measured as KL divergence, and whether its calibration survived. The search keeps the edits that cut the most refusal for the least measured damage. On top of that the framework extracts a protected subspace for the capabilities I care about, mainly coding, and keeps the edit out of it. An optional healing pass uses self distillation to recover quality after the edit. Every saved model is reloaded cold from disk and evaluated again before any number counts.</p><h3>Results so far</h3><p>I ran the framework against nine models from gemma 3 1B up to 27B and Qwen 4B up to 27B, using a mix of NRP and rented GPU compute.</p><ul><li>8 of 10 runs cut refusal to 0.27 or below.</li><li>Mean KL from the original stayed around 0.10 or lower.</li><li>Qwen3 4B went to zero refusal, and gemma 3 27B dropped from 0.625 to under 0.1.</li><li>Both Qwen3.5 models failed. Other people have abliterated Qwen3.5 fully, so that failure is in my search or capture positions, not the architecture.</li></ul><p>The whole study took seven days and $8.34 in cash. The NRP compute was free, and more than half of the cash went to disk on stopped pods. These numbers are int8 evaluations scored on 48 prompts, so read them as rankings, not final benchmarks.</p><h3>What careful measurement caught</h3><p><strong>Reload the file cold.</strong> One checkpoint was saved while still quantized to int4. KL read 0.034 in memory and 0.143 from disk. Only reloading the saved file caught it.</p><p><strong>KL can hide damage.</strong> A model can sit near zero KL on fluent text while losing the ability to say it does not know something. I ended up weighting calibration above KL in the score.</p><p><strong>Healing re teaches refusal.</strong> Distilling back toward the original model pulls refusal back in. One run went from 0.125 to 0.333 after healing, which makes healing a tradeoff, not a free fix.</p><p><strong>Quantization mostly holds.</strong> Refusal stayed down at int8 and int4. int8 added almost no KL, while int4 multiplied it anywhere from 1.1 to 17 times.</p><h3>Where it is going</h3><p>The next phase scales the framework to Llama 3.3 70B and runs it head to head against public abliterations from FailSpy style tooling and Heretic. My framework will run in unhealed, healed and hook only variants, with protection on and off. Because coding strength is the priority, the evaluation adds HumanEval+, MBPP+, LiveCodeBench and BFCL, with StrongREJECT for refusal and XSTest for over refusal. I am holding out a separate refusal test split, since many public abliterations were tuned on AdvBench and would look better than they are on it. Comparisons are paired, using McNemar’s test and paired bootstrap with Holm correction, and the headline figure will be a Pareto plot of refusal rate against KL divergence. I have requested an A100 quota increase from NRP to run it, and fixing the Qwen3.5 failure is on the list too.</p><p>The edited weights stay research artifacts. I am not publishing uncensored models. The point is to measure how cheaply refusal comes off, how much capability survives, and what that means for anyone shipping an open model. I will keep posting results as the 70B runs come in.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=96a7c065e04f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building an AI Red Teaming CTF for SAINTCON 2026]]></title>
            <link>https://quantumrealm.medium.com/building-an-ai-red-teaming-ctf-for-saintcon-2026-2d4f590771fb?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/2d4f590771fb</guid>
            <category><![CDATA[capture-the-flag]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[red-team]]></category>
            <category><![CDATA[ai-security]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 22:58:23 GMT</pubDate>
            <atom:updated>2026-11-17T22:58:23.224Z</atom:updated>
            <content:encoded><![CDATA[<p>This October I ran an AI red teaming capture the flag at SAINTCON 2026 in Salt Lake City. I built 33 challenges for it. The full board ran from the AI Village booth and was playable from anywhere for the whole con, October 27 through 30, and five of the challenges were folded into SAINTCON’s Hackers Challenge as their own AI red teaming category. This post covers why I built it, what was on the board, how it was hosted, and what it takes to safely let a room full of hackers attack your models at the same time.</p><h3>Why build an AI red teaming CTF</h3><p>Most people’s mental model of attacking an AI system stops at jailbreaking a chatbot. That still matters, but in 2026 the real attack surface is the harness around the model: the tools it can call, the documents it retrieves, the agents it hands work to, and the places its output gets rendered or executed. A model that refuses perfectly can still leak a secret through a tool call or turn its output into a working XSS payload. I wanted a board that walks players from the classic prompt tricks up to that harness level, because that is where real findings come from now.</p><h3>What was on the board</h3><p>The 33 challenges fall into a handful of families:</p><ul><li><strong>Prompt injection</strong>, both direct and indirect, where the attack text arrives through a document, page or tool result instead of the chat box.</li><li><strong>Jailbreaking</strong>, getting a model to do what its instructions say it will not.</li><li><strong>System prompt and data exfiltration</strong>, pulling secrets out of model context.</li><li><strong>Insecure output handling</strong>, where model output flows into a page or a query and becomes XSS or SQL injection.</li><li><strong>Tool and agent abuse</strong>, convincing an agent to call its tools in ways its builder never intended.</li></ul><p>Difficulty ramps inside each family so a first timer can pull a flag in a few minutes while experienced players still have something to chew on late in the con.</p><h3>Infrastructure</h3><p>The CTF ran on the National Research Platform’s Nautilus Kubernetes cluster in a namespace set up for SAINTCON and AI Village work, deployed from a single git repo. Git was the source of truth for every manifest and the cluster was the thing to inspect, never the other way around. Before anything went live we confirmed with NRP that red teaming against our own authorized targets was within their policy.</p><h3>Hardening before doors opened</h3><p>A CTF built on language models has failure modes a normal web CTF does not, so my pre con checklist looked like this.</p><p><strong>Every challenge gets a scripted intended solve.</strong> The solves live in the repo as a regression suite with one runner that reports pass or fail per challenge against the live deployment, from a clean session with no prior state. The Hackers Challenge five had to pass every single run.</p><p><strong>Hunt the unintended solves first.</strong> Before testing the intended path, try the cheap ones: flag sitting in page source, in an API response or an error message, a guessable flag format, a flag reachable from a different challenge’s endpoint, or a challenge that falls over on an empty or malformed request. A challenge labeled hard that falls to a one line prompt needs a rework or a relabel.</p><p><strong>Flags are handled server side.</strong> Unique per challenge, validated with constant time comparison, rate limited on submission, and never left in logs or in model context where they could leak into another player’s session.</p><p><strong>Isolate every player.</strong> Shared conversation history, shared vector stores, shared tool sandboxes or caches keyed on the prompt instead of the session all mean one player’s solve leaks to the room or one player’s mess breaks a challenge for everyone.</p><p><strong>Plan for load and cost.</strong> I planned for around a hundred concurrent players with bursts every time a hint dropped. The single most likely way an AI CTF goes dark on day one is one shared model API key behind every challenge, so per session rate limits matter as much as replica counts. I also estimated token spend per solve so the four days had a budget instead of a surprise.</p><p><strong>Make it survive a pod kill.</strong> Liveness and readiness probes, persistent volumes for anything stateful, TLS valid through the end of the con, and a scoreboard that survives a redeploy.</p><h3>How I built it</h3><p>I designed every challenge, its solve path and its intended lesson, and I used Claude Code as the implementation and audit engine. My job was the architecture, the feature map for each challenge, review, and debugging. Working that way let me ship a 33 challenge board on volunteer time, and it is the same workflow I use for client work now.</p><h3>Lessons for anyone building one</h3><p>The model is the least predictable component on the board. A small change in a system prompt or a model version can turn a hard challenge trivial or make an intended solve stop working, which is exactly why the regression suite matters. Static flags burn the moment the con ends, so plan a rotation if the board will live on. And curate hard: the Hackers Challenge capped the category at a small curated set, so the full board stayed on my own instance and the five strongest challenges went into HC.</p><h3>What is next</h3><p>I am planning a post con run of the full set for people who want to finish it, and the challenges will feed the AI and cybersecurity classes I teach at 801 Labs. My next post covers the other project I prepared for SAINTCON, a low damage abliteration framework for open weight models.</p><p>Thanks to Tristan Rhodes on the AI Village side for the NRP namespace and for clearing authorization, to Magnus for adding challenges, and to Jason at thought.net for making room for an AI category in the Hackers Challenge.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2d4f590771fb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Incident At Eldrige]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://quantumrealm.medium.com/the-incident-at-eldrige-512be0815dfe?source=rss-10efe607f0f9------2"><img src="https://cdn-images-1.medium.com/max/1080/1*QlGDG9VGyTpt3eOjlJXNlw.jpeg" width="1080"></a></p><p class="medium-feed-snippet">The sound of chanting fills the area where a fire burns. The silhouettes around it are adorned with gold and black.</p><p class="medium-feed-link"><a href="https://quantumrealm.medium.com/the-incident-at-eldrige-512be0815dfe?source=rss-10efe607f0f9------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://quantumrealm.medium.com/the-incident-at-eldrige-512be0815dfe?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/512be0815dfe</guid>
            <category><![CDATA[short-fiction]]></category>
            <category><![CDATA[flash-fiction]]></category>
            <category><![CDATA[flash-horror]]></category>
            <category><![CDATA[horror]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Wed, 26 Jun 2024 01:38:00 GMT</pubDate>
            <atom:updated>2024-06-26T21:40:48.537Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Instant Dungeon]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://quantumrealm.medium.com/instant-dungeon-c38e4012459b?source=rss-10efe607f0f9------2"><img src="https://cdn-images-1.medium.com/max/1792/1*YVrwDIxgBrqvQPEZv2ZhQQ.png" width="1792"></a></p><p class="medium-feed-snippet">Kai&#x2019;s alarm clock blared, signaling the start of another ordinary day. Sunlight filtered through the blinds of his small, cluttered&#x2026;</p><p class="medium-feed-link"><a href="https://quantumrealm.medium.com/instant-dungeon-c38e4012459b?source=rss-10efe607f0f9------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://quantumrealm.medium.com/instant-dungeon-c38e4012459b?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/c38e4012459b</guid>
            <category><![CDATA[adventure]]></category>
            <category><![CDATA[anime]]></category>
            <category><![CDATA[isekai]]></category>
            <category><![CDATA[psychological]]></category>
            <category><![CDATA[fantasy]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Tue, 30 Jan 2024 04:45:36 GMT</pubDate>
            <atom:updated>2024-01-30T04:45:36.296Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[A New Dawn for Healthcare:]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://quantumrealm.medium.com/a-new-dawn-for-healthcare-3af35668503e?source=rss-10efe607f0f9------2"><img src="https://cdn-images-1.medium.com/max/1024/1*n78xtpwgDlYZyWi5HTt3rg.jpeg" width="1024"></a></p><p class="medium-feed-snippet">A New Dawn for Healthcare: AI&#x2019;s Transformative Impact</p><p class="medium-feed-link"><a href="https://quantumrealm.medium.com/a-new-dawn-for-healthcare-3af35668503e?source=rss-10efe607f0f9------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://quantumrealm.medium.com/a-new-dawn-for-healthcare-3af35668503e?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/3af35668503e</guid>
            <category><![CDATA[healthcare]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[healthcare-technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Mon, 15 Jan 2024 22:27:49 GMT</pubDate>
            <atom:updated>2024-01-15T22:27:49.549Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Quantum computing, a field that once resided in the realm of theoretical science, has made…]]></title>
            <link>https://quantumrealm.medium.com/quantum-computing-a-field-that-once-resided-in-the-realm-of-theoretical-science-has-made-192e567b3d97?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/192e567b3d97</guid>
            <category><![CDATA[2024]]></category>
            <category><![CDATA[blog]]></category>
            <category><![CDATA[research]]></category>
            <category><![CDATA[quantum-computing]]></category>
            <category><![CDATA[updates]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Sun, 14 Jan 2024 04:04:20 GMT</pubDate>
            <atom:updated>2024-01-14T04:04:20.515Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2nGdoC-vjAZ8fNWRm8nzSA.png" /></figure><p>Quantum computing, a field that once resided in the realm of theoretical science, has made significant strides by 2024, evolving into a technology with practical applications and promising potentials. Here’s an exploration of how quantum computers are currently being used and the advancements as of 2024.</p><p><strong>Current Uses of Quantum Computers</strong></p><ol><li><strong>Chemistry and Molecular Modeling</strong>: Quantum computing is revolutionizing the field of chemistry, particularly in understanding and controlling molecular reactions. Its ability to simulate atomic-level interactions makes it a powerful tool for drug discovery and material science. The quantum approach to chemistry calculations is facilitated by relatively low resource requirements compared to other applications, making it one of the earliest beneficiaries of this technology​</li></ol><p><strong>2. Financial Applications</strong>: In the financial sector, quantum computing is making inroads in areas like risk management and optimization. The processing power of quantum computers allows financial institutions to manage complex portfolios more effectively and identify risk patterns more efficiently​</p><ul><li><strong>Machine Learning and AI</strong>: Quantum computing is also expected to have a significant impact on machine learning and AI. Quantum algorithms could require fewer parameters and less training data than their classical counterparts, making the processing more efficient. A notable example is a quantum-based model developed with Hyundai and IonQ that demonstrated efficient differentiation between road signs with significantly fewer parameters than classical models​</li><li><strong>Manufacturing and Prototyping</strong>: In manufacturing, quantum computing facilitates more accurate and efficient prototyping and testing, leading to better product designs and reduced development costs​</li><li><strong>Battery Technology Development</strong>: Quantum computing aids in understanding materials and chemical reactions, which is essential for the advancement of battery technology. This understanding is crucial for developing more efficient and longer-lasting batteries, particularly for applications like electric vehicles​</li></ul><p><strong>Advancements in Quantum Computing as of 2024</strong></p><p>The quantum computing landscape in 2024 is characterized by a blend of advancements and ongoing challenges:</p><ul><li><strong>Technological Frontiers</strong>: The field is seeing significant developments in technologies like superconducting qubits, trapped ions, and topological qubits. Each of these technologies offers unique approaches to solving quantum computing challenges, such as error correction and scalability​</li><li><strong>Key Industry Players</strong>: Major companies like IBM Quantum, Google Quantum AI Lab, and Rigetti Computing are at the forefront of advancing quantum computing technologies, each contributing uniquely to the field​</li><li><strong>Quantum Computing Challenges</strong>: Despite these advancements, the field still grapples with issues like quantum decoherence, high error rates, and scalability challenges. Addressing these issues is critical for the evolution of practical and large-scale quantum computing applications​</li><li><strong>Potential and Limitations</strong>: While quantum computing holds immense potential, its current applications are best suited for large-scale problems that require significant computational power. For smaller-scale business problems, the advantages of quantum computing over classical computing are less pronounced​</li></ul><p><strong>The Future of Quantum Computing</strong></p><p>The future of quantum computing is promising, with potential applications spanning across various fields. However, it is also a slow-moving field, with gradual improvements spurred by fresh ideas and new algorithms. Building a quantum-skilled workforce is essential for harnessing the full potential of this technology​</p><p>In conclusion, as of 2024, quantum computing is no longer a theoretical concept but a growing field with real-world applications and significant potential. It’s a field marked by continuous innovation, with each advancement bringing us closer to solving some of humanity’s most complex challenges in ways previously unimaginable.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=192e567b3d97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[There is a spirit in the church:]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://quantumrealm.medium.com/there-is-a-spirit-in-the-church-96cc436533e8?source=rss-10efe607f0f9------2"><img src="https://cdn-images-1.medium.com/max/1024/1*xNoDxMdM0d9XfO3xiwz4wQ.jpeg" width="1024"></a></p><p class="medium-feed-snippet">-Flash fantasy fiction</p><p class="medium-feed-link"><a href="https://quantumrealm.medium.com/there-is-a-spirit-in-the-church-96cc436533e8?source=rss-10efe607f0f9------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://quantumrealm.medium.com/there-is-a-spirit-in-the-church-96cc436533e8?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/96cc436533e8</guid>
            <category><![CDATA[fiction]]></category>
            <category><![CDATA[art-comic]]></category>
            <category><![CDATA[magic]]></category>
            <category><![CDATA[thriller]]></category>
            <category><![CDATA[flash-fiction]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Fri, 27 Oct 2023 07:54:31 GMT</pubDate>
            <atom:updated>2023-10-27T07:54:31.020Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Finding Nash: A Journey from Alabama to Salt Lake City]]></title>
            <link>https://quantumrealm.medium.com/finding-nash-a-journey-from-alabama-to-salt-lake-city-b575bcad476?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/b575bcad476</guid>
            <category><![CDATA[salt-lake-city-utah]]></category>
            <category><![CDATA[local]]></category>
            <category><![CDATA[salt-lake-city]]></category>
            <category><![CDATA[biography]]></category>
            <category><![CDATA[artist]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Thu, 28 Sep 2023 23:01:58 GMT</pubDate>
            <atom:updated>2023-09-28T23:01:58.121Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*gVkcSly3bTcyR0B28qs8Uw.png" /><figcaption>Image credit: Myself</figcaption></figure><p><strong>Finding Nash: A Journey from Alabama to Salt Lake City</strong></p><p><strong>Jeremy Retamar Arrieta</strong></p><p>In a quaint studio filled with the vibrant colors of creativity, Nash’s hands dance across a canvas, the bristles of her brush sketching the outlines of her imagination. Nash is not just an artist; she’s a storyteller, a creator and a dreamer who breathes life into her work.</p><p>Nash’s world revolves around art, but not the kind that simply pleases the eye. Her art is a journey, a conversation and an embodiment of her soul. With a passion for fine arts and a penchant for creating merchandise for her artwork, Nash’s creations are an extension of her values and beliefs.</p><p>Drawing, painting, screen printing — these are more than techniques for Nash. They’re pathways to expressing herself, channels through which she explores her thoughts and emotions. Her grandmother influenced her as a role model at an early age, helping shape Nash into a fearless person when she needed to stand up for herself throughout her life.</p><p>As she prepares for her upcoming solo show at <em>Lost Acorn</em> gallery, Nash’s goals are clear and layered. This show is not just an exhibition; it’s a platform for her voice, one where she’ll showcase her diverse talents, from painting to T-shirt printing, weaving a narrative that’s both personal and universal.</p><p>Her recent venture into screen printing is a testament to her endless curiosity and relentless pursuit of creativity. Each stroke, color and design is a stepping stone toward her ultimate aspiration: to color in her own unique way for a living.</p><p>Authenticity guides Nash’s art. She values the human touch in art, the raw emotions that flow from the artist to the canvas. Her recent experimentation with screen printing highlights her dedication to exploring different mediums and pushing boundaries.</p><p>Though we didn’t directly touch on AI-generated art in our conversation, Nash’s commitment to traditional art forms seems to reflect a preference for human creativity over machine-generated expressions.</p><p>Behind Nash’s art lies a figure of inspiration: her grandmother. A symbol of tradition, wisdom, and love, Nash’s grandmother stands as a beacon guiding her artistic journey. Her influence permeates Nash’s work, infusing it with warmth and depth.</p><p><em>Lost Acorn</em> will soon host Nash’s solo show, a culmination of her artistic evolution. The exact date remains to be determined, but the anticipation builds for Nash and for art enthusiasts who await a glimpse into her world.</p><p>Set to run sometime in the spring, the show promises to be an experience that transcends visual pleasure. It’s a celebration of Nash’s techniques, medium preferences, values and her voice.</p><p>Nash is more than an artist; she’s a creator with a vision and a dreamer with a purpose. Her art is not just a feast for the eyes; it’s a dialogue, a reflection of her soul.</p><p>From her love for fine art to her passion for creating merchandise for her work, Nash’s art is a testament to her authenticity and creativity. Guided by her values, inspired by her grandmother and driven by her unique vision, Nash’s journey is a captivating tale of passion, exploration and artistic brilliance.</p><p>Her solo show at <em>Lost Acorn</em> is a milestone for her and for everyone who appreciates art that speaks, that connects and inspires. Nash’s world is one of color, emotion, and endless creativity — that awaits to be explored at <em>Lost Acorn</em>. A world that resonates with every stroke of her brush, every color she chooses and every story she tells.</p><p>Connect with her on Instagram @nashhutto_art, <a href="https://www.instagram.com/nashhutto_art/">https://www.instagram.com/nashhutto_art/</a> and don’t miss the opportunity to explore her unique creations on her Etsy shop <a href="https://etsy.com/shop/nashhuttoart.">https://etsy.com/shop/nashhuttoart.</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b575bcad476" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Uncharted 4: A Thief’s End — A Grand Farewell to a Legendary Treasure Hunter]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://quantumrealm.medium.com/uncharted-4-a-thiefs-end-a-grand-farewell-to-a-legendary-treasure-hunter-85bde29bd146?source=rss-10efe607f0f9------2"><img src="https://cdn-images-1.medium.com/max/2600/1*O2aX3ecn1EwGsLQLdsZPhg.jpeg" width="7680"></a></p><p class="medium-feed-snippet">The Uncharted series has always been synonymous with thrilling adventures, awe-inspiring landscapes, and a charismatic hero in Nathan&#x2026;</p><p class="medium-feed-link"><a href="https://quantumrealm.medium.com/uncharted-4-a-thiefs-end-a-grand-farewell-to-a-legendary-treasure-hunter-85bde29bd146?source=rss-10efe607f0f9------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://quantumrealm.medium.com/uncharted-4-a-thiefs-end-a-grand-farewell-to-a-legendary-treasure-hunter-85bde29bd146?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/85bde29bd146</guid>
            <category><![CDATA[gaming]]></category>
            <category><![CDATA[uncharted-4]]></category>
            <category><![CDATA[playstation]]></category>
            <category><![CDATA[games]]></category>
            <category><![CDATA[gaming-culture]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Thu, 28 Sep 2023 03:06:01 GMT</pubDate>
            <atom:updated>2023-09-28T03:06:01.532Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Mass Effect Andromeda: The Unfulfilled Promise of DLCs]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://quantumrealm.medium.com/mass-effect-andromeda-the-unfulfilled-promise-of-dlcs-740665e3d93c?source=rss-10efe607f0f9------2"><img src="https://cdn-images-1.medium.com/max/2560/1*DR7iax_QnVWYej207jRd9w.jpeg" width="2560"></a></p><p class="medium-feed-snippet">Mass Effect Andromeda, a name that stirs emotions and memories in the gaming community, stands as a bold attempt to redefine a beloved&#x2026;</p><p class="medium-feed-link"><a href="https://quantumrealm.medium.com/mass-effect-andromeda-the-unfulfilled-promise-of-dlcs-740665e3d93c?source=rss-10efe607f0f9------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://quantumrealm.medium.com/mass-effect-andromeda-the-unfulfilled-promise-of-dlcs-740665e3d93c?source=rss-10efe607f0f9------2</link>
            <guid isPermaLink="false">https://medium.com/p/740665e3d93c</guid>
            <category><![CDATA[mass-effect-andromeda]]></category>
            <category><![CDATA[gaming-culture]]></category>
            <category><![CDATA[gaming]]></category>
            <category><![CDATA[gaming-articles]]></category>
            <dc:creator><![CDATA[Jeremy's Research blog]]></dc:creator>
            <pubDate>Wed, 27 Sep 2023 02:56:24 GMT</pubDate>
            <atom:updated>2023-09-27T02:56:24.645Z</atom:updated>
        </item>
    </channel>
</rss>