<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Posts on anotherjesse.com</title><link>https://anotherjesse.com/posts/</link><description>Recent content in Posts on anotherjesse.com</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Mon, 04 Mar 2024 00:00:00 +0000</lastBuildDate><atom:link href="https://anotherjesse.com/posts/index.xml" rel="self" type="application/rss+xml"/><item><title>Experimenting With Transformers &amp; Nokens</title><link>https://anotherjesse.com/posts/phi-nokens/</link><pubDate>Mon, 04 Mar 2024 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/phi-nokens/</guid><description>&lt;p&gt;Sometimes a single word isn&amp;rsquo;t enough. What if you could blend tokens or use non-tokens (aka nokens&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;)?&lt;/p&gt;
&lt;p&gt;Let&amp;rsquo;s explore these ideas, learn more about transformers: the architecture and Hugging Face library.&lt;/p&gt;
&lt;h2 id="installing--setup"&gt;Installing &amp;amp; Setup&lt;/h2&gt;
&lt;p&gt;I&amp;rsquo;m going to use Microsoft&amp;rsquo;s &lt;a href="https://huggingface.co/microsoft/phi-2"&gt;Phi-2&lt;/a&gt; as it is smaller than Mistral / Llama2. The ideas &lt;em&gt;should&lt;/em&gt; transfer to other transformer models.&lt;/p&gt;
&lt;p&gt;After &lt;code&gt;pip install transformers torch&lt;/code&gt;, we can load our language model to experiment with using the Hugging Face library.&lt;/p&gt;</description></item><item><title>Proteus</title><link>https://anotherjesse.com/posts/proteus/</link><pubDate>Sun, 03 Mar 2024 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/proteus/</guid><description>&lt;h3 id="parti-prompt-disclaimer"&gt;Parti Prompt Disclaimer&lt;/h3&gt;
&lt;p&gt;Many of these prompts are challenging for current generative imagery models. Compare: &lt;code&gt;253&lt;/code&gt; &amp;ldquo;a motorcycle&amp;rdquo; vs &lt;code&gt;999&lt;/code&gt; &amp;ldquo;a black background with a large yellow square&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;Drag the slider (or type a number between 0 and 1632) to change the prompt. Additionally, you can click any image to see the full-size verison.&lt;/p&gt;
&lt;div class="parti-viewer"&gt;
&lt;script type="application/json"&gt;
{
 "mode": "grid",
 "models": [
 {"name": "Proteus V0.4 Lightning", "path": "proteus-v0.4-lightning", "default": true},
 {"name": "SDXL", "path": "sdxl", "default": true}
 ],
 "promptMax": 1631,
 "promptDefault": 40,
 "seeds": 4,
 "rows": "models",
 "thumbToggle": false,
 "tsv": "PartiPrompts.tsv"
}
&lt;/script&gt;
&lt;/div&gt;
&lt;script src="https://anotherjesse.com/js/parti-viewer.js" defer&gt;&lt;/script&gt;</description></item><item><title>Playground V2.5</title><link>https://anotherjesse.com/posts/playground-v25/</link><pubDate>Tue, 27 Feb 2024 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/playground-v25/</guid><description>&lt;h2 id="note"&gt;Note&lt;/h2&gt;
&lt;p&gt;I&amp;rsquo;m Still rendering Parti Prompt with Playground &amp;hellip; this is a preview&lt;/p&gt;
&lt;h3 id="parti-prompt-disclaimer"&gt;Parti Prompt Disclaimer&lt;/h3&gt;
&lt;p&gt;Many of these prompts are challenging for current generative imagery models. Compare: &lt;code&gt;253&lt;/code&gt; &amp;ldquo;a motorcycle&amp;rdquo; vs &lt;code&gt;999&lt;/code&gt; &amp;ldquo;a black background with a large yellow square&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;Drag the slider (or type a number between 0 and 1632) to change the prompt. Additionally, you can click any image to see the full-size verison.&lt;/p&gt;
&lt;div class="parti-viewer"&gt;
&lt;script type="application/json"&gt;
{
 "mode": "grid",
 "models": [
 {"name": "SDXL", "path": "sdxl", "default": true},
 {"name": "Playground 2.5", "path": "playground-25", "default": true}
 ],
 "promptMax": 1631,
 "promptDefault": 40,
 "seeds": 4,
 "rows": "seeds",
 "thumbToggle": true,
 "tsv": "PartiPrompts.tsv"
}
&lt;/script&gt;
&lt;/div&gt;
&lt;script src="https://anotherjesse.com/js/parti-viewer.js" defer&gt;&lt;/script&gt;</description></item><item><title>SDXL lightning</title><link>https://anotherjesse.com/posts/sdxl-lightning/</link><pubDate>Wed, 21 Feb 2024 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/sdxl-lightning/</guid><description>&lt;p&gt;Today ByteDance released SDXL Lightning (adding to the family of distilled/faster SDXL versions Turbo, LCM). From the &lt;a href="https://arxiv.org/abs/2402.13929"&gt;paper&lt;/a&gt;, we can improvements in the quality and how close it matches SDXL.&lt;/p&gt;
&lt;p&gt;&lt;img src="from-paper.png" alt="results from paper"&gt;&lt;/p&gt;
&lt;h2 id="exploring-parti-prompts"&gt;Exploring Parti Prompts&lt;/h2&gt;
&lt;p&gt;I recently rendered SDXL as part of &lt;a href="https://anotherjesse.com/posts/parti-prompts"&gt;parti prompts&lt;/a&gt;, and Lightning is quick - let&amp;rsquo;s make it easy to compare.&lt;/p&gt;
&lt;p&gt;Below, each row is a different (initial noise) seed, whereas columns are different models (SDXL, Lightning 1 step, &amp;hellip;)&lt;/p&gt;</description></item><item><title>Parti Prompts</title><link>https://anotherjesse.com/posts/parti-prompts/</link><pubDate>Mon, 19 Feb 2024 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/parti-prompts/</guid><description>&lt;div class="parti-viewer"&gt;
&lt;script type="application/json"&gt;
{
 "mode": "single",
 "models": [
 {"name": "sd15", "path": "sd15"},
 {"name": "sd21", "path": "sd21"},
 {"name": "sdxl", "path": "sdxl"},
 {"name": "sdxl-lcm", "path": "sdxl-lcm"},
 {"name": "sdxl-lightning-unet-1", "path": "sdxl-lightning-unet-1"},
 {"name": "sdxl-lightning-unet-2", "path": "sdxl-lightning-unet-2"},
 {"name": "sdxl-lightning-unet-4", "path": "sdxl-lightning-unet-4"},
 {"name": "sdxl-lightning-unet-8", "path": "sdxl-lightning-unet-8"},
 {"name": "sdxl-turbo", "path": "sdxl-turbo"},
 {"name": "stable-cascade", "path": "stable-cascade", "default": true}
 ],
 "promptMax": 1631,
 "promptDefault": 321,
 "seeds": 4,
 "thumbToggle": true,
 "tsv": "PartiPrompts.tsv"
}
&lt;/script&gt;
&lt;/div&gt;
&lt;script src="https://anotherjesse.com/js/parti-viewer.js" defer&gt;&lt;/script&gt;

&lt;h2 id="stable-cascade"&gt;Stable Cascade&lt;/h2&gt;
&lt;h2 id="sdxl"&gt;SDXL&lt;/h2&gt;
&lt;p&gt;SDXL images were generated with 25 steps using the FP16 variant&lt;/p&gt;</description></item><item><title>The Stable Cascade 2 Step</title><link>https://anotherjesse.com/posts/stable-cascade/</link><pubDate>Sun, 18 Feb 2024 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/stable-cascade/</guid><description>&lt;p&gt;Stabilty AI has shared an early release of Stable Cascade (built upon Würstchen). Whereas Stable Diffusion uses a single UNet followed by a VAE, this model uses 2 UNets - resulting in improvements in inference time and training resources.&lt;/p&gt;
&lt;div id="cascade-viewer"&gt;&lt;/div&gt;
&lt;script src="cascade.js" defer&gt;&lt;/script&gt;</description></item><item><title>Rorschach Test For LLaVA LLM</title><link>https://anotherjesse.com/posts/llava-rorschach/</link><pubDate>Sat, 30 Dec 2023 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/llava-rorschach/</guid><description>&lt;p&gt;LLaVA works by teaching an LLM (Llama2) to &amp;ldquo;understand&amp;rdquo; embeddings from an image model (CLIP). What happens if you replace the embeddings with randomness and ask the LLM questions about what it sees?&lt;/p&gt;
&lt;h2 id="hacking-llava"&gt;Hacking LLaVA&lt;/h2&gt;
&lt;p&gt;After loading the LLaVA model, we can inject a function that runs after clip embed but before projection to the LLMs token latent space.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="background-color:#f7f7f7;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-python" data-lang="python"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;&lt;span style="color:#57606a"&gt;# load LLaVA using HF transformers&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;tokenizer&lt;span style="color:#1f2328"&gt;,&lt;/span&gt; model&lt;span style="color:#1f2328"&gt;,&lt;/span&gt; image_processor&lt;span style="color:#1f2328"&gt;,&lt;/span&gt; context_len &lt;span style="color:#0550ae"&gt;=&lt;/span&gt; load_pretrained_model&lt;span style="color:#1f2328"&gt;(&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#0a3069"&gt;&amp;#34;liuhaotian/llava-v1.5-13b&amp;#34;&lt;/span&gt;&lt;span style="color:#1f2328"&gt;,&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; model_name&lt;span style="color:#0550ae"&gt;=&lt;/span&gt;&lt;span style="color:#0a3069"&gt;&amp;#34;llava-v1.5-13b&amp;#34;&lt;/span&gt;&lt;span style="color:#1f2328"&gt;,&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; model_base&lt;span style="color:#0550ae"&gt;=&lt;/span&gt;&lt;span style="color:#cf222e"&gt;None&lt;/span&gt;&lt;span style="color:#1f2328"&gt;,&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; load_8bit&lt;span style="color:#0550ae"&gt;=&lt;/span&gt;&lt;span style="color:#cf222e"&gt;False&lt;/span&gt;&lt;span style="color:#1f2328"&gt;,&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; load_4bit&lt;span style="color:#0550ae"&gt;=&lt;/span&gt;&lt;span style="color:#cf222e"&gt;False&lt;/span&gt;&lt;span style="color:#1f2328"&gt;)&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;&lt;span style="color:#57606a"&gt;# inject hook that allows modifying the generated clip embed&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;tower &lt;span style="color:#0550ae"&gt;=&lt;/span&gt; model&lt;span style="color:#0550ae"&gt;.&lt;/span&gt;get_vision_tower&lt;span style="color:#1f2328"&gt;()&lt;/span&gt; 
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;tower&lt;span style="color:#0550ae"&gt;.&lt;/span&gt;register_forward_hook&lt;span style="color:#1f2328"&gt;(&lt;/span&gt;random_forward_hook&lt;span style="color:#1f2328"&gt;)&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;&lt;span style="color:#57606a"&gt;# if global alter_seed set, override embed with randn of same shape&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;alter_seed &lt;span style="color:#0550ae"&gt;=&lt;/span&gt; &lt;span style="color:#cf222e"&gt;None&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;&lt;span style="color:#cf222e"&gt;def&lt;/span&gt; &lt;span style="color:#6639ba"&gt;random_forward_hook&lt;/span&gt;&lt;span style="color:#1f2328"&gt;(&lt;/span&gt;module&lt;span style="color:#1f2328"&gt;,&lt;/span&gt; &lt;span style="color:#6639ba"&gt;input&lt;/span&gt;&lt;span style="color:#1f2328"&gt;,&lt;/span&gt; output&lt;span style="color:#1f2328"&gt;):&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#cf222e"&gt;if&lt;/span&gt; &lt;span style="color:#0550ae"&gt;not&lt;/span&gt; alter_seed &lt;span style="color:#0550ae"&gt;is&lt;/span&gt; &lt;span style="color:#cf222e"&gt;None&lt;/span&gt;&lt;span style="color:#1f2328"&gt;:&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#6639ba"&gt;print&lt;/span&gt;&lt;span style="color:#1f2328"&gt;(&lt;/span&gt;&lt;span style="color:#0a3069"&gt;&amp;#34;overriding image embedding with seed: &amp;#34;&lt;/span&gt;&lt;span style="color:#1f2328"&gt;,&lt;/span&gt; alter_seed&lt;span style="color:#1f2328"&gt;)&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; torch&lt;span style="color:#0550ae"&gt;.&lt;/span&gt;manual_seed&lt;span style="color:#1f2328"&gt;(&lt;/span&gt;alter_seed&lt;span style="color:#1f2328"&gt;)&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; output&lt;span style="color:#1f2328"&gt;[&lt;/span&gt;&lt;span style="color:#0550ae"&gt;0&lt;/span&gt;&lt;span style="color:#1f2328"&gt;]&lt;/span&gt; &lt;span style="color:#0550ae"&gt;=&lt;/span&gt; torch&lt;span style="color:#0550ae"&gt;.&lt;/span&gt;randn_like&lt;span style="color:#1f2328"&gt;(&lt;/span&gt;output&lt;span style="color:#1f2328"&gt;[&lt;/span&gt;&lt;span style="color:#0550ae"&gt;0&lt;/span&gt;&lt;span style="color:#1f2328"&gt;])&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#cf222e"&gt;return&lt;/span&gt; output
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;You can try it out by running &lt;a href="https://replicate.com/anotherjesse/llava-lies"&gt;llava-lies on replicate&lt;/a&gt;. This model still asks for an image, to simplify the modification of LLaVA&amp;rsquo;s code. The hook &lt;code&gt;random_forward_hook&lt;/code&gt; above replaces the output with normalized seeded noise.&lt;/p&gt;</description></item><item><title>Kandinksy 3</title><link>https://anotherjesse.com/posts/kandinsky3/</link><pubDate>Wed, 22 Nov 2023 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/kandinsky3/</guid><description>&lt;p&gt;Kandinsky 3 is the latest from the team that brought previous versions of the model.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;We conducted a lot of experiments on choosing an architecture and did a lot of work with data to make text understanding and generation quality better, and the architecture itself simpler and more concise. We also made our model more “domestic”: now it is much better oriented in the Russian and Soviet cultural field.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;p&gt;&lt;img src="architecture.png" alt="Flan-UL2 + Unet"&gt;&lt;/p&gt;</description></item><item><title>Lazy AI Coding Bots</title><link>https://anotherjesse.com/posts/lazy-coders/</link><pubDate>Sat, 03 Dec 2022 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/lazy-coders/</guid><description>&lt;p&gt;I&amp;rsquo;ve added a new command to my telegram bot: &lt;code&gt;js&lt;/code&gt;. This command will ask GPT3 to generate javascript code and then run it.&lt;/p&gt;
&lt;p&gt;For instance if I say ask it to generate a function to reverse a string, it will create function that reverses the letters one at a time.&lt;/p&gt;
&lt;p&gt;&lt;img src="reverse.png" alt="GPT3 defines a function - the bot evaluates it and returns the result"&gt;&lt;/p&gt;
&lt;p&gt;As GPT3 tended to respond with code that defined functions, I realized I could build up a library of functions! And by sending the list of functions as context for further runs, further functions could be based on existing functions.&lt;/p&gt;</description></item><item><title>Being Misunderstood</title><link>https://anotherjesse.com/posts/being-misunderstood/</link><pubDate>Tue, 05 Apr 2022 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/being-misunderstood/</guid><description>&lt;p&gt;I really enjoyed the David Whyte&amp;rsquo;s &lt;a href="https://www.soundstrue.com/products/clear-mind-wild-heart"&gt;Clear Mind, Wild Heart&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;About 2/3 in, he said something that reshaped how I go about exploring the world.&lt;/p&gt;
&lt;p&gt;Being ok with being misunderstood at the beginning is important. New ideas need to be protected and nurtured.&lt;/p&gt;
&lt;blockquote&gt;
&lt;p&gt;when you actually tell others what your secret desires are - in your work, in your life.&lt;/p&gt;
&lt;p&gt;a good diagnostic feature that you are doing something unique - something that is true to the
complexion your character - is that &lt;strong&gt;other people should not
understand entirely what you are doing&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>Programmable Cat</title><link>https://anotherjesse.com/posts/programmable-cat/</link><pubDate>Sat, 09 Oct 2021 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/programmable-cat/</guid><description>&lt;h2 id="idea"&gt;Idea&lt;/h2&gt;
&lt;p&gt;The family was excited to decorate to halloween this year, and several of the decorations came with &lt;em&gt;Try Me&lt;/em&gt; buttons. The buttons were external to the decoration - and seem to be simple momentary switches (when they are pressed the circuit is closed for the duration of the button press).&lt;/p&gt;
&lt;p&gt;It would be neat to add these to our programmable space using a tiny ESP-8266 microcontroller / MQTT.&lt;/p&gt;
&lt;p&gt;Based on success of connecting &lt;a href="https://anotherjesse.com/posts/miffy-super-powers"&gt;Miffy&lt;/a&gt; to our local programmable space, this seemed approachable for an electronics novice like me.&lt;/p&gt;</description></item><item><title>Alien Love Worms</title><link>https://anotherjesse.com/posts/alien-love-worm/</link><pubDate>Sun, 28 Oct 2018 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/alien-love-worm/</guid><description>&lt;h1 id="inspiration"&gt;Inspiration&lt;/h1&gt;
&lt;p&gt;Around the time the glowforge was announced, I saw some inspiring art that
had many layered cuts. I thought they were paper, turns out they are
by an artist named &lt;a href="https://gabrielschama.com"&gt;Gabriel Schama&lt;/a&gt;, and are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;huge! many are six to eight feet tall/wide&lt;/li&gt;
&lt;li&gt;cut in a (thin) plywood!&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;img src="gabriel-website.png" alt="screenshot of gabrielschama.com"&gt;&lt;/p&gt;
&lt;h1 id="presenting-alien-love-worm"&gt;Presenting &amp;ldquo;Alien Love Worm&amp;rdquo;&lt;/h1&gt;
&lt;p&gt;For my first design I wanted to do a paper cut. Living near the
coast, I wanted to encorporate a whale tail, waves, a peace sign and
hearts.&lt;/p&gt;</description></item><item><title>Seamless Loops</title><link>https://anotherjesse.com/posts/seamless-loops/</link><pubDate>Sun, 28 Oct 2018 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/seamless-loops/</guid><description>&lt;p&gt;Using this &lt;a href="https://necessarydisorder.wordpress.com/2017/11/15/drawing-from-noise-and-then-making-animated-loopy-gifs-from-there/"&gt;tutorial in processing&lt;/a&gt;,
and &lt;a href="https://twitter.com/golan"&gt;Golan Levin&lt;/a&gt;&amp;rsquo;s sketch in processing,
started to understand how to make seamlessly looping animations. I took Golan&amp;rsquo;s
awesome demo and ported it to p5.js (live demo). This gif made the technique click for me.&lt;/p&gt;
&lt;p&gt;The basic idea is think of the random values as locations on a clock face. This gives
1D noise over time (shown in the image above). For 2D you can use 3D noise space
and think about a line sweeping through.&lt;/p&gt;</description></item><item><title>Miffy Super Powers</title><link>https://anotherjesse.com/posts/miffy-super-powers/</link><pubDate>Fri, 23 Dec 2016 00:00:00 +0000</pubDate><guid>https://anotherjesse.com/posts/miffy-super-powers/</guid><description>&lt;p&gt;When my oldest daughter was born, my dear friend &lt;a href="https://twitter.com/bs"&gt;Britt Selvitelle&lt;/a&gt; gave us a gift of &lt;a href="https://www.mrmaria.com/miffy-xl-lamp/"&gt;Miffy Lamp&lt;/a&gt; from Japan.&lt;/p&gt;
&lt;p&gt;&lt;img src="miffy.png" alt="Miffy Lamp"&gt;&lt;/p&gt;
&lt;h1 id="my-problems-with-power"&gt;My problems with power&lt;/h1&gt;
&lt;p&gt;Miffy had two &amp;ldquo;problems&amp;rdquo;, the first being the power cord for the lamp didn&amp;rsquo;t fit into american outlets. The second was the bulb being incandescent consumed lots of power and created lots of heat (and rather bright if I remember correctly).&lt;/p&gt;
&lt;h1 id="solution-lets-put-it-on-the-internet"&gt;Solution: Let&amp;rsquo;s put it on the internet!&lt;/h1&gt;
&lt;p&gt;Having learned about ESP-8266, a cheap &amp;amp; tiny microcontroller that has WiFi, I thought it would be cool to give the bunny an API.&lt;/p&gt;</description></item></channel></rss>