Email: Muse Spark Open ✨, Microsoft’s 3T param model 🧠, d1-ultra 💡

Subject:
Muse Spark Open ✨, Microsoft’s 3T param model 🧠, d1-ultra 💡
Received:
10/16/2026, 1:31:47 PM
From:
"TLDR AI" <dan@tldrnewsletter.com>
To:
silk.theater.56@fwdnl.com
&lt;!DOCTYPE html&gt;&lt;html lang="en"&gt;&lt;head&gt;
		&lt;meta http-equiv="Content-Type" content="text/html charset=UTF-8"&gt;
		&lt;meta charset="UTF-8"&gt;
		&lt;meta name="viewport" content="width=device-width"&gt;
		&lt;meta name="x-apple-disable-message-reformatting"&gt;
		&lt;title&gt;TLDR AI&lt;/title&gt;
		&lt;meta name="color-scheme" content="light dark"&gt;
		&lt;meta name="supported-color-schemes" content="light dark"&gt;
		&lt;style type="text/css"&gt;
			:root {
				color-scheme: light dark; supported-color-schemes: light dark;
			}
			
			*,
			*:after,
			*:before {
				-webkit-box-sizing: border-box; -moz-box-sizing: border-box; box-sizing: border-box;
			}
			
			* {
				-ms-text-size-adjust: 100%; -webkit-text-size-adjust: 100%;
			}
			
			html,
			body,
			.document {
				width: 100% !important; height: 100% !important; margin: 0; padding: 0;
			}
			
			body {
				-webkit-font-smoothing: antialiased; -moz-osx-font-smoothing: grayscale; text-rendering: optimizeLegibility;
			}
			
			div[style*="margin: 16px 0"] {
				margin: 0 !important;
			}
			
			table,
			td {
				mso-table-lspace: 0pt; mso-table-rspace: 0pt;
			}
			
			table {
				border-spacing: 0; border-collapse: collapse; table-layout: fixed; margin: 0 auto;
			}
			
			img {
				-ms-interpolation-mode: bicubic; max-width: 100%; border: 0;
			}
			
			*[x-apple-data-detectors] {
				color: inherit !important; text-decoration: none !important;
			}
			
			.x-gmail-data-detectors,
			.x-gmail-data-detectors *,
			.aBn {
				border-bottom: 0 !important; cursor: default !important;
			}
			
			.btn {
				-webkit-transition: all 200ms ease; transition: all 200ms ease;
			}
			
			.btn:hover {
				background-color: #f67575; border-color: #f67575;
			}
			
			* {
				font-family: Arial, Helvetica, sans-serif; font-size: 18px;
			}
			
			@media screen and (max-width: 600px) {
				.container {
					width: 100%; margin: auto;
				}
				.stack {
					display: block!important; width: 100%!important; max-width: 100%!important;
				}
				.btn {
					display: block; width: 100%; text-align: center;
				}
			}
			
			body,
			p,
			td,
			tr,
			.body,
			table,
			h1,
			h2,
			h3,
			h4,
			h5,
			h6,
			div,
			span {
				background-color: #FEFEFE !important; color: #010101 !important;
			}
			
			@media (prefers-color-scheme: dark) {
				body,
				p,
				td,
				tr,
				.body,
				table,
				h1,
				h2,
				h3,
				h4,
				h5,
				h6,
				div,
				span {
					background-color: #27292D !important; color: #FEFEFE !important;
				}
			}
			
			a {
				color: inherit !important; text-decoration: underline !important;
			}

		&lt;/style&gt;
		&lt;!--[if mso | ie]&gt;
		&lt;style type="text/css"&gt;
			a {
				background-color: #FEFEFE !important; color: #010101 !important;
			}
			@media (prefers-color-scheme: dark) {
				a {
					background-color: #27292D !important; color: #FEFEFE !important;
				}
			}
	 &lt;/style&gt;
			&lt;![endif]--&gt;
	&lt;/head&gt;
	&lt;body class=""&gt;




		&lt;div style="display: none; max-height: 0px; overflow: hidden;"&gt;Muse Spark Open is an 816B-parameter multimodal open-weight model with 42B active parameters and a 1M-token context window. It was designed for&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;‌&nbsp;&lt;/div&gt;
		&lt;div style="display: none; max-height: 0px; overflow: hidden;"&gt;
			&lt;br&gt;
		&lt;/div&gt;

		&lt;table align="center" class="document"&gt;
			&lt;tbody&gt;
				&lt;tr&gt;
					&lt;td valign="top"&gt;

						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" class="container" width="600"&gt;
							&lt;tbody&gt;
								&lt;tr class="inner-body"&gt;
									&lt;td&gt;

										&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
											&lt;tbody&gt;
												&lt;tr class="header"&gt;
													&lt;td bgcolor="" class="container"&gt;

														&lt;table width="100%"&gt;
															&lt;tbody&gt;
																&lt;tr&gt;
																	&lt;td class="container"&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" style="margin-top: 0px;" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td style="padding: 0px;"&gt;

																						&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div style="text-align: center;"&gt;
																											&lt;span style="margin-right: 0px;"&gt;&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Ftldr.tech%2Fai%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/RdgikKcaC4fiSiqKCndbMS8IJ-jFXGcdt36k2-MJd8Q=452" rel="noopener noreferrer" target="_blank"&gt;&lt;span&gt;Sign Up&lt;/span&gt;&lt;/a&gt;
																											|&lt;span style="margin-right: 2px; margin-left: 2px;"&gt;&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fadvertise.tldr.tech%2F%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=advertisetopnav/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/H6xagtS4JCIWSFfTwekiMpYbatajHpQo1OVyDlK1Das=452" rel="noopener noreferrer" target="_blank"&gt;&lt;span&gt;Advertise&lt;/span&gt;&lt;/a&gt;&lt;/span&gt;|&lt;span style="margin-left: 2px;"&gt;&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fa.tldrnewsletter.com%2Fweb-version%3Fep=1%26lc=1571a9f6-84b7-11f0-b464-47c5c04ad337%26p=a7516726-c958-11f1-bd5a-2578997fd739%26pt=campaign%26t=1792157507%26s=938d61395e61a126828a644a05afd75143938e2672bf9d6332d39a3146263706/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/oKGU1TreDxrdgqwq4IqHK8nZWaWdGA5h4Or6MPPv9U8=452"&gt;&lt;span&gt;View Online&lt;/span&gt;&lt;/a&gt;&lt;/span&gt;
																											&lt;br&gt;
																											&lt;/span&gt;&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="text-align: center;"&gt;&lt;span data-darkreader-inline-color="" style="--darkreader-inline-color:#3db3ff; color: rgb(51, 175, 255) !important; font-size: 30px;"&gt;T&lt;/span&gt;&lt;span style="font-size: 30px;"&gt;&lt;span data-darkreader-inline-color="" style="color: rgb(232, 192, 96) !important; --darkreader-inline-color:#e8c163; font-size:30px;"&gt;L&lt;/span&gt;&lt;span data-darkreader-inline-color="" style="color: rgb(101, 195, 173) !important; --darkreader-inline-color:#6ec7b2; font-size:30px;"&gt;D&lt;/span&gt;&lt;/span&gt;&lt;span data-darkreader-inline-color="" style="--darkreader-inline-color:#dd6e6e; color: rgb(220, 107, 107) !important; font-size: 30px;"&gt;R&lt;/span&gt;
																										&lt;br&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;br&gt;

																						&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr id="together-with"&gt;
																									&lt;td align="center" height="20" style="vertical-align:middle !important;" valign="middle" width="100%"&gt;&lt;strong style="vertical-align:middle !important; height: 100%;"&gt;Together With &lt;/strong&gt;
																										&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fdeveloper.godaddy.com%2Fen%2Fdocs%2Fapi-users%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=explore%26utm_content=header_agent_built_app%26utm_term=2026-10-16_tldr_ai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/0uUKWfLImSUpPYu_RgKM5EuRHV3vNVNPI1LKLjj_EXs=452"&gt;&lt;img src="https://images.tldr.tech/godaddy.png" valign="middle" style="vertical-align: middle !important; height: 100%;" alt="GoDaddy"&gt;&lt;/a&gt;&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;


																						&lt;table style="table-layout: fixed; width:100%;" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;"&gt;
																										&lt;div style="text-align: center;"&gt;

																											&lt;h1&gt;&lt;strong&gt;TLDR AI &lt;span id="date"&gt;2026-10-16&lt;/span&gt;&lt;/strong&gt;&lt;/h1&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table style="table-layout: fixed; width:100%;" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr id="sponsy-copy"&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fdeveloper.godaddy.com%2Fen%2Fdocs%2Fapi-users%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=explore%26utm_content=header_agent_built_app%26utm_term=2026-10-16_tldr_ai/2/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/zkUeHBX8A32bWiG6_sV9XCMU7atVopLzAL8dYXcDfL4=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Your agent built the app. Now help it launch. (Sponsor)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Working code is just the start. The GoDaddy Developer Platform lets your AI agent deploy your app, register an available domain with your approval, connect it to your app, and set up professional email.&lt;br&gt;&lt;br&gt;Bring hosting, domains, and email together in one connected workflow. Spend less time switching between setup screens and more time building your business.&lt;br&gt;&lt;br&gt;Explore how GoDaddy's APIs can help your agent take your project from built to online.&lt;br&gt;&lt;br&gt;→ &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fdeveloper.godaddy.com%2Fen%2Fdocs%2Fapi-users%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=explore%26utm_content=cta_developer_guides%26utm_term=2026-10-16_tldr_ai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/NiZOKl1BjqOQkc5yJ7pI8SGJ7h24kBFvU4BXUz9r9PU=452" rel="noopener noreferrer nofollow" target="_blank"&gt;&lt;span&gt;Explore the developer guides&lt;/span&gt;&lt;/a&gt;
																											&lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;
																																											&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;
																	&lt;/td&gt;
																&lt;/tr&gt;
															&lt;/tbody&gt;
														&lt;/table&gt;
													&lt;/td&gt;
												&lt;/tr&gt;
												&lt;tr bgcolor=""&gt;
													&lt;td class="container"&gt;

														&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
															&lt;tbody&gt;
																&lt;tr&gt;
																	&lt;td style="padding: 0px;"&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;&lt;span style="font-size: 36px;"&gt;&lt;span style="font-size:36px;"&gt;🚀&lt;/span&gt;&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;

																								&lt;h1&gt;&lt;strong&gt;Headlines &amp; Launches&lt;/strong&gt;&lt;/h1&gt;
																							&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table style="table-layout: fixed; width: 100%;" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top"&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fresearch.meta.ai%2Fblog%2Fmuse-spark-open%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/rKD43I4kEyfxZcjHTB45B0NKFojruhRPqalp5jq1eAg=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Muse Spark Open (4 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Muse Spark Open is an 816B-parameter multimodal open-weight model with 42B active parameters and a 1M-token context window. It was designed for coding and long-horizon agentic tasks and can run locally across two Apple M5 Ultra Mac Studios using MLX.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fmicrosoft.ai%2Fnews%2Ffrontier-pre-training-at-microsoft-ai%2F%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/HiWK092ERFxvAXrWRQSADKbvNOoSrv5n18IuG1gqyJI=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Inside Microsoft's Frontier Pre-Training Stack (18 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Microsoft detailed the training of MAI-Base-2, a 2.8T-parameter mixture-of-experts model with 80B active parameters that it said surpassed every open-weight model in its held-out pre-training evaluations. The deep dive covers its data strategy, compute efficiency, full-stack infrastructure, and Microsoft's plans to turn internally trained frontier models into enterprise offerings through Foundry.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.liquid.ai%2Fblog%2Fd1-ultra%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/VLlQXc-T9VyJ3Zv1Wl_3fr3C6XH4Aa-pPjS0psNoTV4=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;d1-ultra: Introducing the World's Best Decision Model (3 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Liquid AI's d1-ultra is a multimodal decision model designed for complex real-world tasks like detecting defects on circuit boards. The model processes text and images and provides better-calibrated answers and higher accuracy than previous models like Jev 1.14. Available on the Liquid API, d1-ultra is priced at $0.042 per million input tokens and will soon be accessible via Vercel and OpenRouter.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;


																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;&lt;span style="font-size: 36px;"&gt;🧠&lt;/span&gt;&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;

																								&lt;h1&gt;&lt;strong&gt;Deep Dives &amp; Analysis&lt;/strong&gt;&lt;/h1&gt;
																							&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;


																		&lt;table style="table-layout: fixed; width: 100%;" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top"&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Falready-clean.goodfire.com%2F%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/vOXZMnuR8rBhwqgavLd44oU8G6eUeYZEjIoggSYNo4Y=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Neural Networks Are Already Clean (65 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    The cause of polysemanticity in early model organisms was thought to be superposition. This convinced much of the field that networks pack concepts arbitrarily, but what looked like arbitrary packing was mostly recursive generalization, which is less concerning. This means networks can be studied directly, instead of needing to be untangled first. The outlook for mechanistic interpretability may be far better than we thought.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FF54ab4/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/rlNfIWPVQk5zwHrxV-TBvB5y46L1qJzMfPQGeAx45B4=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;The Gradient Is a Covariance (8 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Reinforcement learning, on-policy distillation, and supervised fine-tuning on near-policy data change only a small fraction of parameters. Small and sparse are different claims. The expected gradient is a covariance, the Fisher is a variance, and the Fisher caps the gradient for any reward. At one position, the Fisher is the variance of logit sensitivities: whatever moves every candidate together counts for nothing. None of this needs on-policy sampling - it needs a score centered under the sampling distribution.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fianbarber.blog%2F2026%2F10%2F15%2Fmo-turns-mo-problems%2F%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/S-D9t1gJqnzCJGe6joNm1KKCGEl4R5j2-GOZcDfCZG8=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Mo' turns, mo' problems (6 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Doing reinforcement learning for agent use-cases is conceptually not at all that complicated. For each problem, you set up a VM with a harness and a fresh model context, send the prompt, let the model make tool calls, score the whole thing at the end, and update the weights. Agentic post-training means there's more than the model to consider. Even when the harness is consistent, you can still go off-policy with the trajectories themselves. Agents spawn subagents, make parallel tool calls, and compact large sessions, which adds complexity.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;&lt;span style="font-size: 36px;"&gt;🧑‍💻&lt;/span&gt;&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;

																								&lt;h1&gt;&lt;strong&gt;Engineering &amp; Research&lt;/strong&gt;&lt;/h1&gt;
																							&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table style="table-layout: fixed; width: 100%;" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top"&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.atera.com%2Fwebinars%2Fautonomous-it-separating-hype-from-real-autonomy-281026%2F%3Futm_source=TLDR%26utm_medium=email%26utm_campaign=TLDR_AI/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/z67Rjx-YtAwzSMrGaeMy-7Oo0p6ibzSn4Y62f3wUiyc=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Webinar: Separating Real IT Autonomy from the Hype (Sponsor)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    “Autonomous IT” is suddenly everywhere, but almost no two vendors define it the same way – making it nearly impossible for buyers to tell the genuine autonomy from rebranded automation. &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.atera.com%2Fwebinars%2Fautonomous-it-separating-hype-from-real-autonomy-281026%2F%3Futm_source=TLDR%26utm_medium=email%26utm_campaign=TLDR_AI/2/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/eBLLu-3AkLxl0iu7hHmARtaZdzRV8hu_myOegPKvQCw=452" rel="noopener noreferrer nofollow" target="_blank"&gt;&lt;span&gt;Atera's VPs of Product Marketing and IT&lt;/span&gt;&lt;/a&gt; will break down how to define the category, assess autonomous capabilities, and use them effectively. &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.atera.com%2Fwebinars%2Fautonomous-it-separating-hype-from-real-autonomy-281026%2F%3Futm_source=TLDR%26utm_medium=email%26utm_campaign=TLDR_AI/3/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/ZJ56VM09Czosf28qficBZ2AYRystIv4oT7GQ44hYNRg=452" rel="noopener noreferrer nofollow" target="_blank"&gt;&lt;span&gt;Save your virtual seat.&lt;/span&gt;&lt;/a&gt;
																											&lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.krea.ai%2Fkrea-3%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/AQYR9n2zQ7-PLStSxb8muaXID4S-WEdee7QYoB_Kusw=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Krea 3 (Website)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Krea 3 is an image editing model built for production. It can finish an edit in under eight seconds, costing only $0.06 per image. The model was built for production and is four times faster than Nano Banana Pro. It can combine up to eight references into one picture.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fdeveloper.nvidia.com%2Fblog%2Fnvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning%2F%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/uNMNo7OGXiPw1IO9Gk4Ica8S7A8J_WAkxejLIJeSTlw=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Nvidia released Nemotron 3.5 Super VL (5 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Nvidia released Nemotron 3.5 Super VL, a multimodal reasoning model that scored 26.5 on Artificial Analysis' Intelligence Index while using substantially fewer output tokens than its predecessor. The company also introduced the smaller Nemotron 3.5 Bolt 74B-A9B and near-lossless NVFP4 quantized weights.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fthreadreaderapp.com%2Fthread%2F2110797830361436604.html%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/9um2xslZPpU7ote_Bo8braCZitrar85nmboyPNhaehc=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;GPT-Live-1.1 (1 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    GPT-Live-1.1 features new voices in French, German, Hindi, Mexican Spanish, Korean, and Japanese. The voices have better accents, and there is less 'checking'. The model is available in ChatGPT, Codex, and dots. It is rolling out now to all users and coming soon to the API.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;&lt;span style="font-size: 36px;"&gt;🎁&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;&lt;strong&gt;&lt;h1&gt;Miscellaneous&lt;/h1&gt;&lt;/strong&gt;&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table bgcolor="" style="table-layout: fixed; width: 100%;" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top"&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.anthropic.com%2Fnews%2Fan-update-on-independent-evaluators%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/q3M7YBJVmXYNtK0TAkzZUesqD2FlfG-Rf9Uf55-jhJU=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Anthropic Expanded Independent Access to Its AI Safety Work (3 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Anthropic began giving METR employee-like access to models, training information, safeguards, and other internal material so it could independently assess frontier-model risks and publish its conclusions.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fvercel.com%2Fblog%2Fai-gateway-production-index-october-2026%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/2x3ym9KnBqbyebhg3OdJtaxu9tpEPWC3ShNS34fc7_I=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Open weight takes 71.5% of tokens and 23% of spend (7 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Open-weight models now process more than seven in ten gateway tokens and account for nearly one in four dollars. Their share of spend has more than quadrupled since June. Almost one in three teams changed their most-used lab in September, an index record. Jev was the fastest-adopted model in AI Gateway history, but it only accounted for about 0.27% of gateway spend.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;


																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;&lt;span style="font-size: 36px;"&gt;⚡&lt;/span&gt;&lt;/div&gt;&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding-top: 0px; padding-bottom: 0px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;div style="text-align: center;"&gt;

																								&lt;h1&gt;&lt;strong&gt;Quick Links&lt;/strong&gt;&lt;/h1&gt;
																							&lt;/div&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table bgcolor="" style="table-layout: fixed; width: 100%;" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top"&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fshop.md%2F%3Futm_source=Paid%26utm_medium=Newsletter%26utm_campaign=Shop-Launch%26utm_content=TLDR-AI/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/A2h0eXuqb5YUibUE29yBIEBhB29CnCSQVY4ub4iM1uY=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Turn any AI agent into your personal shopper (Sponsor)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Make any agent your personal shopper. Add the skill at &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fshop.md%2F%3Futm_source=Paid%26utm_medium=Newsletter%26utm_campaign=Shop-Launch%26utm_content=TLDR-AI/2/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/7W3nzzE1jdfHBk642rWGjub76zl0wdJzZBqLgmrTlMI=452" rel="noopener noreferrer nofollow" target="_blank"&gt;&lt;span&gt;shop.md&lt;/span&gt;&lt;/a&gt; for your size and order history from Shop. Checkout happens in the conversation and you approve every purchase. &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fshop.md%2F%3Futm_source=Paid%26utm_medium=Newsletter%26utm_campaign=Shop-Launch%26utm_content=TLDR-AI/3/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/r_YevaOY5X91GPpKYO35Qu6jFKT4Tzhqko1TvUQMCFM=452" rel="noopener noreferrer nofollow" target="_blank"&gt;&lt;span&gt;Start shopping now.&lt;/span&gt;&lt;/a&gt;
																											&lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fbfl.ai%2Fblog%2Fflux-3-image-klein%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/dBWa36vpOZUkkloiQnyqAwKwe9MxK1SKE5iBCKp7H5U=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;FLUX 3 Image Klein (4 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Black Forest Labs released FLUX 3 Image Klein, a 5.6B-parameter open-weight model that supports text-to-image generation, editing, native 4K output, and up to 10 reference images.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.perplexity.ai%2Fhub%2Fblog%2Fopen-sourcing-space%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/NyE6O86OX-jAJ_5XH3PSDdKLBsLuzmka-N14S8oq_CE=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Perplexity Open-Sourced Its Agent Sandbox (4 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Perplexity has open-sourced SPACE, the sandbox infrastructure behind its Computer agent.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FcWe75r/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/5wZaGkm6ww5Nh-PG7ZFTBv7DvHF_-xSW1Z2vVjL3Lbo=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;Introducing Handwired: JavaScript as Neural Networks (1 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    Handwired converts ordinary JavaScript code into neural networks that reproduce exact program outputs without training or GPUs.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																						&lt;table align="center" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																							&lt;tbody&gt;
																								&lt;tr&gt;
																									&lt;td class="container" style="padding: 15px 15px;"&gt;
																										&lt;div class="text-block"&gt;
																											&lt;span&gt;
                                &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fevery.to%2Fp%2Fhow-codex-happened%3Futm_source=tldrai/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/cXqfX8663HA1RLpNIKlqqQduNM0bjZTQyGJUMo7QH-o=452"&gt;
                                    &lt;span&gt;
                                        &lt;strong&gt;How Codex Happened (40 minute read)&lt;/strong&gt;
                                    &lt;/span&gt;
																											&lt;/a&gt;
																											&lt;br&gt;
																											&lt;br&gt;
																											&lt;span style="font-family: &quot;Helvetica Neue&quot;, Helvetica, Arial, Verdana, sans-serif;"&gt;
                                    OpenAI's Codex evolved from a failed cloud launch to becoming the cornerstone of OpenAI's products, gaining over 40 million weekly users.
                                &lt;/span&gt;
																											&lt;/span&gt;
																										&lt;/div&gt;
																									&lt;/td&gt;
																								&lt;/tr&gt;
																							&lt;/tbody&gt;
																						&lt;/table&gt;

																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;


																		
																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td align="left" style="word-break: break-word; vertical-align: top; padding: 5px 10px;"&gt;

																						&lt;p style="padding: 0; margin: 0; font-size: 22px; color: #000000; line-height: 1.6; font-weight: bold;"&gt;
																							Love TLDR? Tell your friends and get rewards!
																						&lt;/p&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding: 0px 10px 15px;"&gt;
																						&lt;div class="text-block"&gt;
																							Share your referral link below with friends to get free TLDR swag!
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																				&lt;tr&gt;
																					&lt;td align="left" style="padding: 10px;"&gt;
																						&lt;div class="text-block"&gt;
																							&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Frefer.tldr.tech%2F51d793d6%2F2/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/gCsHr_8Ij1vx7ZUbgDi1kBGk5jH0qZz657CbhJlP7uM=452" style="color: #464ba4; text-decoration: underline;"&gt;https://refer.tldr.tech/51d793d6/2&lt;/a&gt;
																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																				&lt;tr&gt;&lt;/tr&gt;
																				&lt;tr&gt;
																					&lt;td align="left" style="padding:5px 10px;"&gt;
																						&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fhub.sparklp.co%2Fsub_9a4c16be99eb%2F2/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/KWwyscW1R8KQyQrjue598IMxzNvVn-7Jv5_MllxbnV0=452" style="font-size: 16px; line-height: 1.6;  padding: 10px 0; display: inline-block;  text-decoration: underline;"&gt;&lt;span style="mso-text-raise:13pt; text-decoration: underline;"&gt;Track your referrals here.&lt;/span&gt;&lt;/a&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		


																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td align="left" style="word-break: break-word; vertical-align: top; padding: 5px 10px;"&gt;

																						&lt;p style="padding: 0; margin: 0; font-size: 22px; color: #000000; line-height: 1.6; font-weight: bold;"&gt;
																							Want to advertise in TLDR? 📰
																						&lt;/p&gt;
																						&lt;div class="text-block" style="margin-top: 10px;"&gt;
																							If your company is interested in reaching an audience of AI professionals and decision makers, you may want to &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fadvertise.tldr.tech%2F%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=advertisecta/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/IOn8LdFvwKnGI857c3BBH6Pv5d1gfjjJWGiu-3u58vA=452"&gt;&lt;strong&gt;&lt;span&gt;advertise with us&lt;/span&gt;&lt;/strong&gt;&lt;/a&gt;.
																						&lt;/div&gt;
																						&lt;br&gt;

																						&lt;!-- New "Want to work at TLDR?" section --&gt;

																						&lt;p style="padding: 0; margin: 0; font-size: 22px; color: #000000; line-height: 1.6; font-weight: bold;"&gt;
																							Want to work at TLDR? 💼
																						&lt;/p&gt;
																						&lt;div class="text-block" style="margin-top: 10px;"&gt;
																							&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fjobs.ashbyhq.com%2Ftldr.tech/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/CrP_2nijcSmR4x41R6aXFXLyPx3ZNPrI_lph-RK3QNs=452" rel="noopener noreferrer" style="color: #0000EE; text-decoration: underline;" target="_blank"&gt;&lt;strong&gt;Apply here&lt;/strong&gt;&lt;/a&gt;,
																							&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fjobs.ashbyhq.com%2Ftldr.tech%2Fc227b917-a6a4-40ce-8950-d3e165357871/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/lOyA1lUoMkzLc9ThQh1NG_rOforFoUwjr3IalUdOISE=452" rel="noopener noreferrer" style="color: #0000EE; text-decoration: underline;" target="_blank"&gt;&lt;strong&gt;create your own role&lt;/strong&gt;&lt;/a&gt; or send a friend's resume to &lt;a href="mailto:jobs@tldr.tech" style="color: #0000EE; text-decoration: underline;"&gt;jobs@tldr.tech&lt;/a&gt; and get $1k if we hire them! TLDR is one of &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Ffeed%2Fupdate%2Furn:li:activity:7401699691039830016%2F/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/fntyh00kCdrlkit2-xZBX0pmMeg_Wuq_kj_9wcqIMMo=452" rel="noopener noreferrer" style="color: #0000EE; text-decoration: underline;" target="_blank"&gt;&lt;strong&gt;Inc.'s Best Bootstrapped businesses&lt;/strong&gt;&lt;/a&gt; of 2025.
																						&lt;/div&gt;
																						&lt;br&gt;

																						&lt;div class="text-block"&gt;
																							If you have any comments or feedback, just respond to this email!
																							&lt;br&gt;
																							&lt;br&gt; Thanks for reading,
																							&lt;br&gt;
																							&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Ftwitter.com%2Fandrewztan/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/WK0WepJT8mV-NftKgU4zV9DAOlUPlvvW0LzJ6igFU3k=452"&gt;&lt;span&gt;Andrew Tan&lt;/span&gt;&lt;/a&gt;, &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Fin%2Faliiaminian%2F/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/T86-6jr-1q7He2kSr1YfR4_fiappzNd_kp_a_oKT8GY=452"&gt;&lt;span&gt;Ali Aminian&lt;/span&gt;&lt;/a&gt;, &amp; &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Fin%2Fmeetwudi%2F/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/6Jjjgcmihh2adNonMcEkzvMKz5ZDwJZIesstkGi1fDw=452"&gt;&lt;span&gt;Di Wu&lt;/span&gt;&lt;/a&gt;
																							&lt;br&gt;
																							&lt;br&gt;
																						&lt;/div&gt;
																						&lt;br&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;

																		&lt;table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%"&gt;
																			&lt;tbody&gt;
																				&lt;tr&gt;
																					&lt;td class="container" style="padding: 15px 15px;"&gt;
																						&lt;div class="text-block" id="testing-id"&gt;
																							&lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Ftldr.tech%2Fai%2Fmanage%3Femail=silk.theater.56%2540fwdnl.com/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/-_4qb8fOiV2pZZ2Uggu3aTikw6A1huoPC9BsrA5u-5M=452"&gt;Manage your subscriptions&lt;/a&gt; to our other newsletters on tech, startups, and programming. Or if TLDR AI isn't for you, please &lt;a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fa.tldrnewsletter.com%2Funsubscribe%3Fep=1%26l=eedf6b14-3de3-11ed-9a32-0241b9615763%26lc=1571a9f6-84b7-11f0-b464-47c5c04ad337%26p=a7516726-c958-11f1-bd5a-2578997fd739%26pt=campaign%26pv=5%26spa=1792155753%26t=1792157507%26s=c844eef12b5ff4453439fb1c83aee5bf187b633be8f80b8c12774abc9c602d8b/1/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/cD6ovdy0JouFN7yHF0OVEk-GwNK6IlQKsFb-353icJ0=452"&gt;unsubscribe&lt;/a&gt;.
																							&lt;br&gt;

																						&lt;/div&gt;
																					&lt;/td&gt;
																				&lt;/tr&gt;
																			&lt;/tbody&gt;
																		&lt;/table&gt;
																	&lt;/td&gt;
																&lt;/tr&gt;
															&lt;/tbody&gt;
														&lt;/table&gt;
													&lt;/td&gt;
												&lt;/tr&gt;
											&lt;/tbody&gt;
										&lt;/table&gt;
									&lt;/td&gt;
								&lt;/tr&gt;
							&lt;/tbody&gt;
						&lt;/table&gt;
					&lt;/td&gt;
				&lt;/tr&gt;
			&lt;/tbody&gt;
		&lt;/table&gt;





	

&lt;img alt="" src="http://tracking.tldrnewsletter.com/CI0/010001a144e95f88-788613cf-531b-4901-a794-0f3fb7b3abb0-000000/D6hHfaTvLDtuVO48QJzPCBE20dNx8ronTILd-GJO6Lc=452" style="display: none; width: 1px; height: 1px;"&gt;
&lt;/body&gt;&lt;/html&gt;
Email Content
