<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Leslie Li — The Dispatches &amp; Research Log</title>
    <link>https://leslieli.dev/</link>
    <description>Personal research log of Leslie Li. Working on local LLMs, NVFP4 quantization, agentic hardware robotics, and speculative physics.</description>
    <language>en-us</language>
    <lastBuildDate>Fri, 11 Sep 2026 13:50:00 +0800</lastBuildDate>
    <atom:link href="https://leslieli.dev/feed.xml" rel="self" type="application/rss+xml" />

    <item>
      <title>Stonkfly</title>
      <link>https://leslieli.dev/notes/stonkfly/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/stonkfly/</guid>
      <pubDate>Fri, 11 Sep 2026 13:50:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Connectome</category>
      <category>Markets</category>
      <category>Simulation</category>
      <description><![CDATA[Alex Wormuth's page runs a fruit-fly wiring diagram on BTC, ETH, and SOL charts. The bag is paper. The site says profitable learning has not been demonstrated.]]></description>
    </item>

    <item>
      <title>Anthropic on Seven Labs</title>
      <link>https://leslieli.dev/notes/anthropic-seven-labs/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/anthropic-seven-labs/</guid>
      <pubDate>Fri, 11 Sep 2026 13:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Security</category>
      <category>AI</category>
      <category>Distillation</category>
      <description><![CDATA[Anthropic's September 2026 report attributes illicit Claude distillation to seven China-based labs. The counts and the clustering are theirs.]]></description>
    </item>

    <item>
      <title>Your Agent Is Mine</title>
      <link>https://leslieli.dev/notes/your-agent-is-mine/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/your-agent-is-mine/</guid>
      <pubDate>Fri, 11 Sep 2026 12:40:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Security</category>
      <category>Agents</category>
      <category>Supply chain</category>
      <description><![CDATA[A paper measures LLM API routers that sit between agents and model providers. The client points at the router on purpose. Chaofan Shou's posts add a 6 TB dump claim.]]></description>
    </item>

    <item>
      <title>Pamir's Lapis One</title>
      <link>https://leslieli.dev/notes/lapis-one-agent-computer/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/lapis-one-agent-computer/</guid>
      <pubDate>Fri, 11 Sep 2026 00:40:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Hardware</category>
      <category>Agents</category>
      <category>Linux</category>
      <description><![CDATA[Pamir AI's site presents Lapis One as a dedicated Linux computer for agents. Early bird $359. They say the first 140 units ship in October 2026.]]></description>
    </item>

    <item>
      <title>A Calculator, Not 2030</title>
      <link>https://leslieli.dev/notes/a-calculator-not-a-picture/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/a-calculator-not-a-picture/</guid>
      <pubDate>Thu, 10 Sep 2026 09:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Economics</category>
      <category>Society</category>
      <description><![CDATA[Anthropic published a task model of AI and the US economy through 2030. The numbers are theirs. I take the labor-versus-capital finding. I do not take the dashboard as a picture of the year.]]></description>
    </item>

    <item>
      <title>Johansen on the LG Investigation</title>
      <link>https://leslieli.dev/notes/the-smart-part-is-the-defect/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/the-smart-part-is-the-defect/</guid>
      <pubDate>Wed, 09 Sep 2026 19:20:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Hardware</category>
      <category>Privacy</category>
      <category>Security</category>
      <description><![CDATA[On September 8, 2026, Matt Johansen posted a summary of Gamers Nexus's second LG investigation. You buy the TV, you plug it in, he wrote. This is what it does.]]></description>
    </item>

    <item>
      <title>Coxon Resigns from Anthropic</title>
      <link>https://leslieli.dev/notes/they-know-and-they-race/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/they-know-and-they-race/</guid>
      <pubDate>Wed, 09 Sep 2026 15:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Alignment</category>
      <category>Society</category>
      <description><![CDATA[On September 9, 2026, Jacob Coxon posted that he had resigned from Anthropic. He spent three years in pretraining at OpenAI and Anthropic. He says both labs are racing to self-improving superintelligence.]]></description>
    </item>

    <item>
      <title>The Equations Can Break</title>
      <link>https://leslieli.dev/notes/the-equations-can-break/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/the-equations-can-break/</guid>
      <pubDate>Wed, 09 Sep 2026 09:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Mathematics</category>
      <category>Science</category>
      <description><![CDATA[OpenAI's paper constructs, for every viscosity, a 3D Navier-Stokes flow that starts at rest and blows up in finite time with bounded energy and a smooth compact force. Clay C and D. I have not checked the proof.]]></description>
    </item>

    <item>
      <title>We Do Not Know the Threshold</title>
      <link>https://leslieli.dev/notes/we-do-not-know-the-threshold/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/we-do-not-know-the-threshold/</guid>
      <pubDate>Mon, 07 Sep 2026 15:20:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Cognition</category>
      <category>Society</category>
      <description><![CDATA[A paper treats ChatGPT-like tools as a spreading workplace habit. The math can lock people in. We do not know if that has happened. I want people who can still do the work when the model is down.]]></description>
    </item>

    <item>
      <title>Still Mortals</title>
      <link>https://leslieli.dev/notes/still-mortals/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/still-mortals/</guid>
      <pubDate>Mon, 07 Sep 2026 12:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Alignment</category>
      <category>Society</category>
      <description><![CDATA[OpenAI's chief scientist describes an alien mind grown by scaling. A few labs are already playing God. The work now is to make the relationship a partnership while we are still mortal.]]></description>
    </item>

    <item>
      <title>The Peanut Butter Trap in Enterprise AI</title>
      <link>https://leslieli.dev/notes/the-peanut-butter-trap/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/the-peanut-butter-trap/</guid>
      <pubDate>Sun, 06 Sep 2026 14:05:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>FinOps</category>
      <category>Enterprise</category>
      <description><![CDATA[Why spreading small generative AI budgets across twenty departmental sandboxes produces dozens of demos, zero operational integration, and pure overhead.]]></description>
    </item>

    <item>
      <title>The Fallback Problem: When Machines Take the Mind</title>
      <link>https://leslieli.dev/notes/the-fallback-problem/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/the-fallback-problem/</guid>
      <pubDate>Thu, 03 Sep 2026 18:54:02 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Economics</category>
      <category>Labor</category>
      <description><![CDATA[When the Industrial Revolution mechanized muscle, labor retreated into the mind. When machines take cognition, what is labor's biological fallback?]]></description>
    </item>

    <item>
      <title>The Pupil Knows You're Tired Before You Do</title>
      <link>https://leslieli.dev/notes/sparkling-water-cognitive-fatigue/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/sparkling-water-cognitive-fatigue/</guid>
      <pubDate>Thu, 03 Sep 2026 16:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Cognitive Fatigue</category>
      <category>Pupillometry</category>
      <category>Esports</category>
      <description><![CDATA[A small study on sparkling water and esports, and the finding that survives its caveats: the pupil constricts as an objective early warning of cognitive fatigue, before you feel tired.]]></description>
    </item>

    <item>
      <title>Perplexity's Lily: What 1.35x Decode Actually Buys</title>
      <link>https://leslieli.dev/notes/perplexity-lily-engine/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/perplexity-lily-engine/</guid>
      <pubDate>Thu, 03 Sep 2026 10:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Local Inference</category>
      <category>Apple Silicon</category>
      <category>LLM Engineering</category>
      <description><![CDATA[Perplexity's Lily is a custom Rust and Metal inference engine for Qwen3.6-35B-A3B on Apple silicon. What it gains over MLX-LM, and the llama.cpp benchmark that is missing.]]></description>
    </item>

    <item>
      <title>One Flash, 2.6 Sigma: The LZ Dark Matter Hint</title>
      <link>https://leslieli.dev/notes/lz-dark-matter-hint/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/lz-dark-matter-hint/</guid>
      <pubDate>Thu, 03 Sep 2026 09:45:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Particle Physics</category>
      <category>Dark Matter</category>
      <category>LUX-ZEPLIN</category>
      <description><![CDATA[LUX-ZEPLIN recorded a single particle interaction it cannot explain, the most compelling dark matter hint to date. 2.6 sigma, below the discovery threshold, and too energetic for the simple WIMP model.]]></description>
    </item>

    <item>
      <title>The Night the Telegraphs Ran on Sunlight</title>
      <link>https://leslieli.dev/notes/carrington-event/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/carrington-event/</guid>
      <pubDate>Thu, 03 Sep 2026 09:04:27 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Space Weather</category>
      <category>Solar Physics</category>
      <category>History of Science</category>
      <description><![CDATA[A look at the Carrington Event of September 1859, through Carrington's own account, the telegraphs that ran on the auroral current, and what the benchmark does and does not tell us about a repeat.]]></description>
    </item>

    <item>
      <title>The Illusion of Progress: AI, Speed, and Understanding</title>
      <link>https://leslieli.dev/notes/ai-illusion-of-progress/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/ai-illusion-of-progress/</guid>
      <pubDate>Mon, 31 Aug 2026 15:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI</category>
      <category>Mathematics</category>
      <category>Terence Tao</category>
      <description><![CDATA[Terence Tao's video on AI in mathematics, and the worry that agents raise throughput while the next generation of understanding goes untrained.]]></description>
    </item>

    <item>
      <title>The Duck Does Not Ship Here</title>
      <link>https://leslieli.dev/notes/pollen-microduck-diy/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/pollen-microduck-diy/</guid>
      <pubDate>Mon, 31 Aug 2026 09:30:03 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Robotics</category>
      <category>Pollen Robotics</category>
      <category>DIY</category>
      <description><![CDATA[Pollen's $399 RL biped ships to a short list of countries. Software is open. The HAT is not. A parts list from the repo, and why DIY is the path from Singapore.]]></description>
    </item>

    <item>
      <title>Telling More Than They Can Know</title>
      <link>https://leslieli.dev/notes/telling-more-than-they-can-know/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/telling-more-than-they-can-know/</guid>
      <pubDate>Sun, 30 Aug 2026 22:13:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Mechanistic Interpretability</category>
      <category>LLM Self-Explanation</category>
      <category>Confabulation</category>
      <description><![CDATA[A clean experiment injects a water vector, flips a model's answer to Bali, and asks why. The explanation follows the context, not the cause.]]></description>
    </item>

    <item>
      <title>A Reflexion Author Scored 100% AI</title>
      <link>https://leslieli.dev/notes/a-reflexion-author-scored-100-percent-ai/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/a-reflexion-author-scored-100-percent-ai/</guid>
      <pubDate>Sun, 30 Aug 2026 19:32:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Agents</category>
      <category>Agent Memory</category>
      <category>AI Detection</category>
      <description><![CDATA[A two-word reply ran an agent-memory essay through an AI detector and got 100%. What the verdict does and does not establish, and why the thesis survives the accusation.]]></description>
    </item>

    <item>
      <title>Agents Built Technologies That Outlived Them</title>
      <link>https://leslieli.dev/notes/swarmworld-stigmergic-agents/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/swarmworld-stigmergic-agents/</guid>
      <pubDate>Sun, 30 Aug 2026 15:16:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Research</category>
      <category>Multi-Agent Systems</category>
      <category>Stigmergy</category>
      <description><![CDATA[MIT's SwarmWorld: identical LLM agents self-organized into a technological society and built artifacts that outlived them, coordinating through the environment rather than by talking.]]></description>
    </item>

    <item>
      <title>Software Engineering Still Matters. The Reason Changed.</title>
      <link>https://leslieli.dev/notes/software-engineering-still-matters/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/software-engineering-still-matters/</guid>
      <pubDate>Sun, 30 Aug 2026 14:12:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Engineering</category>
      <category>Andrew Ng</category>
      <category>Software Engineering</category>
      <description><![CDATA[Andrew Ng's second AI Engineering Skills Map article, on why software engineering fundamentals still matter when a coding agent writes all the code.]]></description>
    </item>

    <item>
      <title>I Hope It's the Next Pen-F</title>
      <link>https://leslieli.dev/notes/om-system-pen-f-teaser/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/om-system-pen-f-teaser/</guid>
      <pubDate>Thu, 27 Aug 2026 14:08:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Photography</category>
      <category>Cameras</category>
      <category>OM System</category>
      <description><![CDATA[OM System teased a new camera for September 9, leading with the EVF. The silhouette points rangefinder, and I hope it is the next Pen-F.]]></description>
    </item>

    <item>
      <title>Before pstack, Lauren Tan Wrote About Fear</title>
      <link>https://leslieli.dev/notes/before-pstack-lauren-tan-fear/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/before-pstack-lauren-tan-fear/</guid>
      <pubDate>Thu, 27 Aug 2026 13:21:19 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Learning</category>
      <category>Engineering</category>
      <category>AI Agents</category>
      <description><![CDATA[Lauren Tan's 2017 essay on fear and learning: the failed startup, the self-taught start, the Netflix career, and the line I am still learning.]]></description>
    </item>


    <item>
      <title>Go Deep First: Notes on Lauren Tan's pstack</title>
      <link>https://leslieli.dev/notes/go-deep-first-pstack/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/go-deep-first-pstack/</guid>
      <pubDate>Thu, 27 Aug 2026 11:20:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Agents</category>
      <category>Cursor</category>
      <category>Verification</category>
      <description><![CDATA[A reading of Lauren Tan's How I Use Cursor: agents as amnesiac new hires, verification as the bottleneck, pstack as playbooks, Benny as a work-in-progress factory. Editorial on what transfers off Cursor.]]></description>
    </item>

    <item>
      <title>Google's ATLAS: Mapping AI Use at the Scale of an Economy</title>
      <link>https://leslieli.dev/notes/google-atlas-gemini-economy/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/google-atlas-gemini-economy/</guid>
      <pubDate>Wed, 26 Aug 2026 14:20:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Economics</category>
      <category>Gemini</category>
      <category>Solow Paradox</category>
      <description><![CDATA[Notes on arXiv:2608.00038: what 15M classified Gemini conversations actually show (broad adoption, shallow depth, collaborative use), which headline numbers to discount, three interactive explainers, and an editorial reading of vendor measurement as narrative infrastructure.]]></description>
    </item>

    <item>
      <title>The Family LLM Box: Mac mini vs. Mac Studio Over Two Years</title>
      <link>https://leslieli.dev/notes/family-local-llm-tco/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/family-local-llm-tco/</guid>
      <pubDate>Wed, 26 Aug 2026 11:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Cost of Ownership</category>
      <category>Local LLM</category>
      <category>Mac mini</category>
      <description><![CDATA[Mac mini M6 32GB vs. M5 Pro 64GB vs. M5 Ultra 256GB: two-year cost of ownership against ChatGPT, Claude, Gemini, and Grok subscriptions for a household with kids.]]></description>
    </item>

    <item>
      <title>Portable Computer: Perplexity Puts the Whole Agent Runtime on Your Desk</title>
      <link>https://leslieli.dev/notes/portable-computer-dgx-spark/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/portable-computer-dgx-spark/</guid>
      <pubDate>Wed, 26 Aug 2026 09:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Local-First AI</category>
      <category>Agent Harness</category>
      <category>NVIDIA DGX Spark</category>
      <description><![CDATA[Perplexity’s Portable Computer on NVIDIA DGX Spark: orchestrator, subagents, and agent harness run locally; escalation stays metered cloud. What changes when you own the runtime.]]></description>
    </item>

    <item>
      <title>If Every Agent Can Do the Work, Why Do You Still Need a System?</title>
      <link>https://leslieli.dev/notes/token-freedom-personal-ai-firm/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/token-freedom-personal-ai-firm/</guid>
      <pubDate>Tue, 25 Aug 2026 16:40:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Coase</category>
      <category>Agent Economics</category>
      <category>Personal AI System</category>
      <description><![CDATA[Roland Wayne’s thread read against Coase’s 1937 Economica essay: the cost of using the price mechanism, direction inside a firm, why firms do not grow without limit, and the API / CLI / access-token split.]]></description>
    </item>

    <item>
      <title>You Don’t Have to Say “Jobs” to Trigger Job Fear</title>
      <link>https://leslieli.dev/notes/ai-control-narratives-replacement-anxiety/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/ai-control-narratives-replacement-anxiety/</guid>
      <pubDate>Tue, 25 Aug 2026 09:50:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Anxiety</category>
      <category>Work</category>
      <category>Agency</category>
      <description><![CDATA[A lab experiment: short videos about AI agency and control raised job-replacement anxiety even though none of them mentioned jobs.]]></description>
    </item>

    <item>
      <title>The Tiny Red Dot Lives in a Watershed We Haven’t Finished Mapping</title>
      <link>https://leslieli.dev/notes/laniakea-shapley-basin/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/laniakea-shapley-basin/</guid>
      <pubDate>Mon, 24 Aug 2026 23:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Cosmology</category>
      <category>Laniakea</category>
      <category>Cosmic Web</category>
      <description><![CDATA[The viral Laniakea map: a ~60% chance we sit in the larger Shapley basin of attraction — a drainage of galaxy flows, not a bound supercluster, and the outer rim is still off the page.]]></description>
    </item>

    <item>
      <title>Skills That Scan Clean Still Fail at Runtime: NVIDIA’s Skill Lift</title>
      <link>https://leslieli.dev/notes/nvidia-aces-skill-lift/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/nvidia-aces-skill-lift/</guid>
      <pubDate>Mon, 24 Aug 2026 22:40:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Agent Skills</category>
      <category>Evaluation</category>
      <category>NVIDIA</category>
      <description><![CDATA[ACES: structural skill scanners correlate with LLM-judge quality at ρ=0.14. Skill Lift — paired with/without-skill runs — is the metric that answers whether a skill actually helps.]]></description>
    </item>

    <item>
      <title>A Reading Path for RL on LLMs</title>
      <link>https://leslieli.dev/notes/rl-for-llms-reading-list/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/rl-for-llms-reading-list/</guid>
      <pubDate>Mon, 24 Aug 2026 14:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Reinforcement Learning</category>
      <category>RLHF</category>
      <category>GRPO</category>
      <description><![CDATA[Cameron Wolfe’s RL-for-LLMs bibliography as a path, plus a plain-English glossary of TRPO, PPO, GAE, KL, GRPO, and the rest of the alphabet soup.]]></description>
    </item>

    <item>
      <title>The AI Miracle Was a Straight Line</title>
      <link>https://leslieli.dev/notes/ai-scaling-laws-the-straight-line/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/ai-scaling-laws-the-straight-line/</guid>
      <pubDate>Mon, 24 Aug 2026 10:15:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Scaling Laws</category>
      <category>Chinchilla</category>
      <category>AI Strategy</category>
      <description><![CDATA[A LinkedIn-friendly tour of scaling laws: the log-log ruler, 6ND, Chinchilla, the emergence mirage, and why 2024 forked the line into train / think / efficiency.]]></description>
    </item>

    <item>
      <title>Outrunning Light: Cherenkov Radiation as an Electromagnetic Shock Front</title>
      <link>https://leslieli.dev/notes/cherenkov-radiation-outrun-light/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/cherenkov-radiation-outrun-light/</guid>
      <pubDate>Mon, 24 Aug 2026 09:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Cherenkov Radiation</category>
      <category>Electromagnetism</category>
      <category>Wave Physics</category>
      <description><![CDATA[From Mathelirium’s simulation: why a charged particle can outrun light inside a dielectric, how the Cherenkov cone forms, and why reactor pools glow blue.]]></description>
    </item>

    <item>
      <title>Prompt as Code: Inside the GPT-Image-2 Industrial Engine &amp; 500+ Reverse-Engineered Visual Protocols</title>
      <link>https://leslieli.dev/notes/awesome-gpt-image-2-prompt-as-code/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/awesome-gpt-image-2-prompt-as-code/</guid>
      <pubDate>Sat, 22 Aug 2026 10:20:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Prompt as Code</category>
      <category>GPT-Image-2</category>
      <category>Generative UI</category>
      <category>Agent Skills</category>
      <description><![CDATA[A review of a structured prompt library, what its examples organize, and why structure can improve maintainability without making image generation deterministic.]]></description>
    </item>

    <item>
      <title>Graph Engineering: From 1 Prompt to 100 Agents Running in One System</title>
      <link>https://leslieli.dev/notes/graph-engineering-1-prompt-to-100-agents/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/graph-engineering-1-prompt-to-100-agents/</guid>
      <pubDate>Sat, 22 Aug 2026 08:45:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Graph Engineering</category>
      <category>Multi-Agent Systems</category>
      <category>Agent Harness</category>
      <description><![CDATA[An architectural masterclass by Hanako (@hanakoxbt): cutting imaginary edges, node contracts, the 4 canonical topologies, verifier gates, and durable state for massive multi-agent systems.]]></description>
    </item>

    <item>
      <title>Andrew Ng&#39;s AI Engineering Skills Map: Deconstructing the 6 Pillars of Building &amp; Deploying AI Applications</title>
      <link>https://leslieli.dev/notes/andrew-ng-ai-engineering-skills-map/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/andrew-ng-ai-engineering-skills-map/</guid>
      <pubDate>Sat, 22 Aug 2026 08:25:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Engineering</category>
      <category>Agent Harness</category>
      <category>System Architecture</category>
      <description><![CDATA[Notes on a skills map derived from more than 10,000 job postings and dozens of interviews, with Leslie's interpretation kept separate from the source.]]></description>
    </item>

    <item>
      <title>Anthropic&#39;s Internal /eli5 Skill: Visualizing Complex Systems via Interactive HTML Artifacts</title>
      <link>https://leslieli.dev/notes/anthropic-eli5-skill-html-artifacts/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/anthropic-eli5-skill-html-artifacts/</guid>
      <pubDate>Sat, 22 Aug 2026 08:15:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Agents</category>
      <category>Prompt Engineering</category>
      <category>Claude Code</category>
      <description><![CDATA[A small Claude skill pattern for turning a topic into a self-contained HTML explanation, including an explicitly adapted SKILL.md example.]]></description>
    </item>

    <item>
      <title>FreeToken: Serving 284B–753B Frontier MoE Models on Consumer Gaming PCs at Interactive Speeds</title>
      <link>https://leslieli.dev/notes/freetoken-edge-native-moe-serving-consumer-gpus/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/freetoken-edge-native-moe-serving-consumer-gpus/</guid>
      <pubDate>Sat, 22 Aug 2026 07:45:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Systems</category>
      <category>Local Models</category>
      <category>Edge Inference</category>
      <description><![CDATA[A review of FreeToken's reported consumer-hardware MoE serving approach, followed by a reproducible protocol for testing it locally.]]></description>
    </item>

    <item>
      <title>Why Some Early-Career Professionals Work Well with AI—and Others Don’t</title>
      <link>https://leslieli.dev/notes/hbr-human-ai-collaboration-amplifiers-delegators-apprentices/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/hbr-human-ai-collaboration-amplifiers-delegators-apprentices/</guid>
      <pubDate>Fri, 21 Aug 2026 22:15:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Workforce</category>
      <category>Human-AI Collaboration</category>
      <category>HBR Research</category>
      <description><![CDATA[A deep dive into KPMG and UT Austin's field study of 523 early-career professionals: Why domain knowledge and AI literacy alone fail to predict success, and the 3 behavioral profiles of human-AI collaboration (Amplifiers, Delegators, Apprentices).]]></description>
    </item>

    <item>
      <title>Agency Agents: 140+ Open-Source Expert Personas</title>
      <link>https://leslieli.dev/notes/github-treasure-agency-agents-open-source-persona-library/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/github-treasure-agency-agents-open-source-persona-library/</guid>
      <pubDate>Fri, 21 Aug 2026 09:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Agents</category>
      <category>Open Source</category>
      <description><![CDATA[Notes on 140+ structured agent personas across 12 divisions, including how their role, scope, deliverable, and failure-mode prompts are organized.]]></description>
    </item>

    <item>
      <title>Codex as a Platform: How OpenAI&#39;s Open Agent Harness Unlocks Embedded AI Beyond Generic Chatboxes</title>
      <link>https://leslieli.dev/notes/codex-as-a-platform-open-agent-harness/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/codex-as-a-platform-open-agent-harness/</guid>
      <pubDate>Fri, 21 Aug 2026 08:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Agents</category>
      <category>System Design</category>
      <description><![CDATA[Notes on Codex exec, the SDK, app-server, and the approval boundaries needed when an agent is embedded in existing software.]]></description>
    </item>

    <item>
      <title>Claude Academy: A Limited Note on the Launch</title>
      <link>https://leslieli.dev/notes/claude-academy-interactive-learning/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/claude-academy-interactive-learning/</guid>
      <pubDate>Fri, 21 Aug 2026 08:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Agents</category>
      <category>Education</category>
      <description><![CDATA[A deliberately limited inventory of what Anthropic's launch material supports, with direct links for checking the current curriculum.]]></description>
    </item>

    <item>
      <title>Transparent Backgrounds in the GPT-Image-2 API</title>
      <link>https://leslieli.dev/notes/transparent-background-gpt-image-2/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/transparent-background-gpt-image-2/</guid>
      <pubDate>Thu, 20 Aug 2026 18:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Computer Vision</category>
      <category>Image Generation</category>
      <description><![CDATA[The API parameters and base64 decoding needed for transparent PNG or WebP output, plus the cases that still need edge inspection.]]></description>
    </item>

    <item>
      <title>IP as Logo: Turning Agent System Skills into Visual Brand Tokens</title>
      <link>https://leslieli.dev/notes/ip-as-logo-agent-skill/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/ip-as-logo-agent-skill/</guid>
      <pubDate>Thu, 20 Aug 2026 16:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Brand Design</category>
      <category>AI Systems</category>
      <description><![CDATA[A framework for condensing complex agent capabilities, toolchains, and protocols into memorable typographic and vector identity tokens.]]></description>
    </item>

    <item>
      <title>What Is an Agent Harness? One Practical Breakdown</title>
      <link>https://leslieli.dev/notes/what-is-an-agent-harness/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/what-is-an-agent-harness/</guid>
      <pubDate>Thu, 20 Aug 2026 14:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Architecture</category>
      <category>Agents</category>
      <description><![CDATA[One practical decomposition of an agent harness into instructions, tools, an execution loop, and the translation layer around model calls.]]></description>
    </item>

    <item>
      <title>Hexapod Robot Edge Brain: Distilling Reasoning Models to Raspberry Pi 5</title>
      <link>https://leslieli.dev/notes/hexapod-robot-edge-brain-architecture/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/hexapod-robot-edge-brain-architecture/</guid>
      <pubDate>Thu, 20 Aug 2026 11:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Robotics</category>
      <category>Edge Compute</category>
      <description><![CDATA[Deploying a quantized multimodal reasoner on Raspberry Pi 5 with Hailo-8 NPU to drive 18-DOF inverse kinematics and dynamic gait adaptation.]]></description>
    </item>

    <item>
      <title>Moderna's Personalized mRNA Cancer Vaccine: Mechanism and Clinical Evidence</title>
      <link>https://leslieli.dev/notes/moderna-mrna-cancer-vaccine/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/moderna-mrna-cancer-vaccine/</guid>
      <pubDate>Wed, 19 Aug 2026 21:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Biotech</category>
      <category>Computational Biology</category>
      <description><![CDATA[Notes on mRNA-4157/V940, neoantigen selection, combination with PD-1 blockade, the reported Phase 2 follow-up, and the Phase 3 program.]]></description>
    </item>

    <item>
      <title>A Reported Ex-Vivo Neural Tissue and Robotic Interface</title>
      <link>https://leslieli.dev/notes/ex-vivo-neural-plasticity/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/ex-vivo-neural-plasticity/</guid>
      <pubDate>Wed, 19 Aug 2026 19:30:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Neuroscience</category>
      <category>Bio-computing</category>
      <description><![CDATA[A cautious account of a reported closed-loop experiment connecting maintained human cortical tissue, a microelectrode array, and robotic piano-key actuation.]]></description>
    </item>

    <item>
      <title>The River of Spacetime: Demystifying Black Holes &amp; the Event Horizon</title>
      <link>https://leslieli.dev/notes/black-hole-event-horizon/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/black-hole-event-horizon/</guid>
      <pubDate>Wed, 19 Aug 2026 17:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Astrophysics</category>
      <category>Simulation</category>
      <description><![CDATA[A coordinate-based river model of Schwarzschild spacetime, with explicit limits on what the visualization and static-clock equations do and do not describe.]]></description>
    </item>

    <item>
      <title>A Local Model Candidate for a Single RTX 5090</title>
      <link>https://leslieli.dev/notes/best-local-model-rtx-5090/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/best-local-model-rtx-5090/</guid>
      <pubDate>Wed, 19 Aug 2026 14:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>AI Models</category>
      <category>Quantization</category>
      <description><![CDATA[A sourced checkpoint candidate for one RTX 5090, with caveats around reported throughput, context-memory fit, and reproducibility.]]></description>
    </item>

    <item>
      <title>Opening the Log</title>
      <link>https://leslieli.dev/notes/opening-the-log/</link>
      <guid isPermaLink="true">https://leslieli.dev/notes/opening-the-log/</guid>
      <pubDate>Wed, 19 Aug 2026 10:00:00 +0800</pubDate>
      <dc:creator>Leslie Li</dc:creator>
      <category>Dev Log</category>
      <category>Philosophy</category>
      <description><![CDATA[Why this site exists: keeping a transparent, public record of works-in-progress, curiosity-driven research, and experimental notes.]]></description>
    </item>
  </channel>
</rss>
