<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><description>CTO building #AI</description><link>https://bsky.app/profile/m9e.bsky.social</link><title>@m9e.bsky.social - Matt Wallace</title><item><link>https://bsky.app/profile/m9e.bsky.social/post/3mf3wgcz4sc2g</link><description>This has to have been the most momentous GenAI month since 11/22. &#xA;&#xA;The best SOTA frontier models drop, and they&#39;re amazing.&#xA;4 massive open models drop and they&#39;re closer than ever to the lead.&#xA;OpenClaw blows up and creates a zeitgeist.&#xA;&#xA;Man, what a time to be alive.</description><pubDate>18 Feb 2026 01:27 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3mf3wgcz4sc2g</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3mf3vsslqis2g</link><pubDate>18 Feb 2026 01:16 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3mf3vsslqis2g</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lsfev5dpk22l</link><description>Love this detail. reuters on the ruling: https://www.reuters.com/legal/litigation/anthropic-wins-key-ruling-ai-authors-copyright-lawsuit-2025-06-24/&#xA;&#xA;[contains quote post or other embedded content]</description><pubDate>25 Jun 2025 00:41 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lsfev5dpk22l</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lquwmryxrs2n</link><description>as an AI founder who is ridiculously all in, when I’m like flabbergasted by the temerity of your product I just can’t even…</description><pubDate>05 Jun 2025 18:18 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lquwmryxrs2n</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3loygctkq5k2n</link><description>Suno is so fucking good. Wow.</description><pubDate>12 May 2025 16:47 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3loygctkq5k2n</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lolqphse522d</link><description>The year is 2030. nVidia announces new DGX with attached Fusion reactor.</description><pubDate>07 May 2025 15:48 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lolqphse522d</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lojdpimz4c24</link><description>Claude did not understand the mission and tried to write his system prompt to Notion. haha!</description><pubDate>06 May 2025 16:50 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lojdpimz4c24</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3loiythwkos2i</link><description>Home setup: rtx3090+4090. testing AWQ Qwen3; both 32B and 32B w/ 4B speculative decoding&#xA;&#xA;TL;DR: ⚠️ on spec in vanilla vllm. w/ a 75% token acceptance generation in batch-4 went from 150t/s (no speculative) -&gt; 50t/s (w/ speculative)&#xA;&#xA;I&#39;d get 400t/s+ max tput w/out speculative, larger batch</description><pubDate>06 May 2025 13:36 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3loiythwkos2i</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3llre5yvcu223</link><description>Wild and wonderful watching a keynote demo 100ft wide and being able to literally picture the code in your head. 😁</description><pubDate>01 Apr 2025 17:04 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3llre5yvcu223</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3llc5ew5mhs2v</link><description>Every schema for content should have a human vs ai flag (or perhaps an enum with human, ai, and then some hybrid roles).</description><pubDate>26 Mar 2025 15:53 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3llc5ew5mhs2v</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lkqtqao2t22u</link><description>Jensen: &#34;I&#39;d never buy a hopper!&#34;&#xA;&#xA;Azure: &#34;We don&#39;t have any Ampere GPUs to turn up even.&#34;&#xA;&#xA;Me: 😠</description><pubDate>19 Mar 2025 18:45 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lkqtqao2t22u</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lkfk6xcug22a</link><description>640KB ought to be enough for anybody.</description><pubDate>15 Mar 2025 06:55 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lkfk6xcug22a</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lk44reguw22j</link><description>It&#39;s interesting because the dials of draft model, draft model quant, --draft length, they all play into the sweet spot. and it&#39;s clearly not super consistent. like I was playing with 32B-Q8 w/ 3B vs 7B 4KL drafts; with those the default --draft 16 seems like the sweet spot. (&gt;12, &gt;32 informal test)&#xA;&#xA;[contains quote post or other embedded content]</description><pubDate>11 Mar 2025 13:01 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lk44reguw22j</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lk4463qnsc2j</link><description>32B Coder-Q8 w/ and w/out 7B-Q4_K_L draft - PSA speculative decoding is in llamacpp and works. (depending on your hardware, experiment w/ diff model sizes - ymmv vary wildly)</description><pubDate>11 Mar 2025 12:50 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lk4463qnsc2j</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3ljao3qkfbs2r</link><description>Reward function engineer</description><pubDate>28 Feb 2025 14:56 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3ljao3qkfbs2r</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lixl2sliis2k</link><description>ok. told claude 3.7 to extra some settings mgmt components from an app layer into generic fastapi router+react component. I expected that to work. Him writing this beautiful readme with emojis I did *NOT* expect.</description><pubDate>25 Feb 2025 00:08 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lixl2sliis2k</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lixkqrsiz22s</link><description>continue&#xA;&#xA;continue&#xA;&#xA;continue&#xA;&#xA;continue&#xA;&#xA;continue, and btw, I think you were interrupted and may have already written some stuff you can&#39;t see&#xA;&#xA;continue&#xA;&#xA;continue&#xA;&#xA;ifykyk</description><pubDate>25 Feb 2025 00:02 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lixkqrsiz22s</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lil2obeocs2u</link><description>The universe came with a game genie we just had to assemble? https://azure.microsoft.com/en-us/blog/quantum/2025/02/19/microsoft-unveils-majorana-1-the-worlds-first-quantum-processor-powered-by-topological-qubits/</description><pubDate>20 Feb 2025 00:43 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lil2obeocs2u</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3li6tqcruw227</link><description>Agent-Oriented Programming</description><pubDate>15 Feb 2025 04:07 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3li6tqcruw227</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lhtwksyrl22p</link><description>Asked ChatGPT to compare and contrast the coconut (reasoning in latent space with reasoning tokens &amp; passing hidden states) to the new recurrent depth paper.&#xA;&#xA;https://chatgpt.com/share/67aa5973-3cb8-800f-a75e-c8515a6c727b&#xA;&#xA;Coolest thing was GPT suggesting a hybrid approach which I had been thinking before I even got to the bottom</description><pubDate>10 Feb 2025 19:58 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lhtwksyrl22p</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lhkt7p3kbs26</link><description>I live in a world of language all day and startup land is not conducive to a lot of distraction, but damn if every time I pop into Suno I&#39;m not blown away.</description><pubDate>07 Feb 2025 05:04 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lhkt7p3kbs26</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lheg33hm6s2b</link><description>Based purely on the speed I feel like o3-mini is getting a solid reception. The tps is like 1/5th what it was last week</description><pubDate>04 Feb 2025 15:53 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lheg33hm6s2b</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lhbrkdo45k26</link><description>deep research seems immediately useful and may have just solved an annoying sglang issue with a particular model for me.</description><pubDate>03 Feb 2025 14:40 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lhbrkdo45k26</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lgypykucbk2y</link><description>🤣❤️</description><pubDate>31 Jan 2025 00:18 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lgypykucbk2y</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lgikpfs6oc2s</link><description>heheh.</description><pubDate>24 Jan 2025 14:01 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lgikpfs6oc2s</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lg6nx3hgh226</link><description>Feel like the &#34;peak of inflated expectations&#34; call may age poorly, although the engineering to really run AI apps is bootstrapping fast into reality, so maybe it isn&#39;t that expectations go down but that eng catches up some?</description><pubDate>20 Jan 2025 15:33 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lg6nx3hgh226</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lfflkh2pr22p</link><pubDate>10 Jan 2025 16:13 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lfflkh2pr22p</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lfdkgfuw3c22</link><description>Happy to report we @kamiwaza-ai.bsky.social have raised $11m to speed up enterprise AI with our stack- and I’m hiring.&#xA;&#xA;https://www.geekwire.com/2025/seattle-vcs-invest-in-kamiwaza-a-new-enterprise-software-startup-helping-companies-adopt-ai/</description><pubDate>09 Jan 2025 20:48 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lfdkgfuw3c22</guid></item><item><link>https://bsky.app/profile/m9e.bsky.social/post/3lfaklubvy22t</link><description>Ok, I tinkered with Suno v4 this morning and holy shit.  We are living in the best timeline.</description><pubDate>08 Jan 2025 16:13 +0000</pubDate><guid isPermaLink="false">at://did:plc:2hy2pr5qmv2li3k5xuxbnprx/app.bsky.feed.post/3lfaklubvy22t</guid></item></channel></rss>