<!DOCTYPE html><html lang="en"><head>
<meta http-equiv="Content-Type" content="text/html charset=UTF-8">
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width">
<meta name="x-apple-disable-message-reformatting">
<title>TLDR AI</title>
<meta name="color-scheme" content="light dark">
<meta name="supported-color-schemes" content="light dark">
<style type="text/css">
: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;
}
</style>
<!--[if mso | ie]>
<style type="text/css">
a {
background-color: #FEFEFE !important; color: #010101 !important;
}
@media (prefers-color-scheme: dark) {
a {
background-color: #27292D !important; color: #FEFEFE !important;
}
}
</style>
<![endif]-->
</head>
<body class="">
<div style="display: none; max-height: 0px; overflow: hidden;">OpenAI is releasing GPT-5.6 Sol, Terra, and Luna tomorrow. It is currently expanding preview access globally. Sol is the flagship model </div>
<div style="display: none; max-height: 0px; overflow: hidden;">
<br>
</div>
<table align="center" class="document">
<tbody>
<tr>
<td valign="top">
<table align="center" border="0" cellpadding="0" cellspacing="0" class="container" width="600">
<tbody>
<tr class="inner-body">
<td>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr class="header">
<td bgcolor="" class="container">
<table width="100%">
<tbody>
<tr>
<td class="container">
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" style="margin-top: 0px;" width="100%">
<tbody>
<tr>
<td style="padding: 0px;">
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div style="text-align: center;">
<span style="margin-right: 0px;"><a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Ftldr.tech%2Fai%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/FdXRf035PJykN_ikRSv4f3i9adr8oUvnw0AaH49_908=452" rel="noopener noreferrer" target="_blank"><span>Sign Up</span></a>
|<span style="margin-right: 2px; margin-left: 2px;"><a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fadvertise.tldr.tech%2F%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=advertisetopnav/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/YjQuU-aIIDUSHnHO_yvrlbbyhqYENDeRt2o4Ex9LEu0=452" rel="noopener noreferrer" target="_blank"><span>Advertise</span></a></span>|<span style="margin-left: 2px;"><a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fa.tldrnewsletter.com%2Fweb-version%3Fep=1%26lc=1571a9f6-84b7-11f0-b464-47c5c04ad337%26p=550aae36-7ac1-11f1-897a-ef5160069b2d%26pt=campaign%26t=1783516611%26s=dca37c5bf74e0527818076c0d3846fa3dcfcef024f6c517821e52314412d9945/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/o1crjpchAhB6hZMERsUzzu_vMA-K9sX-TYbhx4_7334=452"><span>View Online</span></a></span>
<br>
</span></div>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="text-align: center;"><span data-darkreader-inline-color="" style="--darkreader-inline-color:#3db3ff; color: rgb(51, 175, 255) !important; font-size: 30px;">T</span><span style="font-size: 30px;"><span data-darkreader-inline-color="" style="color: rgb(232, 192, 96) !important; --darkreader-inline-color:#e8c163; font-size:30px;">L</span><span data-darkreader-inline-color="" style="color: rgb(101, 195, 173) !important; --darkreader-inline-color:#6ec7b2; font-size:30px;">D</span></span><span data-darkreader-inline-color="" style="--darkreader-inline-color:#dd6e6e; color: rgb(220, 107, 107) !important; font-size: 30px;">R</span>
<br>
</td>
</tr>
</tbody>
</table>
<br>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr id="together-with">
<td align="center" height="20" style="vertical-align:middle !important;" valign="middle" width="100%"><strong style="vertical-align:middle !important; height: 100%;">Together With </strong>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fagent.minimax.io%2F%3Futm_source=morningbrew%26utm_medium=newsletter%26utm_campaign=m3_code%26utm_content=header_m3_open_weights/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/LAupoh0S_wxEixN_sN22Uc5Dpz2DQknPN7syB3Qky-M=452"><img src="https://images.tldr.tech/minimax.png" valign="middle" style="vertical-align: middle !important; height: 100%;" alt="MiniMax"></a></td>
</tr>
</tbody>
</table>
<table style="table-layout: fixed; width:100%;" width="100%">
<tbody>
<tr>
<td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;">
<div style="text-align: center;">
<h1><strong>TLDR AI <span id="date">2026-07-08</span></strong></h1>
</div>
</td>
</tr>
</tbody>
</table>
<table style="table-layout: fixed; width:100%;" width="100%">
<tbody>
<tr id="sponsy-copy">
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fagent.minimax.io%2F%3Futm_source=morningbrew%26utm_medium=newsletter%26utm_campaign=m3_code%26utm_content=header_m3_open_weights/2/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/KzXzSZNI8TB-Noyyb_RPkAVUna4x_K8BiX-I7EY6gYk=452">
<span>
<strong>M3: The first open-weights model to combine frontier coding & agent capabilities (Sponsor)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
MiniMax M3 has a 1M-token context window and native multimodal understanding so it can reason through code, process visual info, and operate across massive contexts. Builders are already using it to:<p></p><p>1️⃣ Create AI agents that reason, use tools, and complete multi-step workflows.</p><p>2️⃣ Analyze screenshots, interfaces, documents, images, and visual workflows without relying on separate vision models.</p><p>3️⃣ Process entire codebases and long documents in a single conversation with a 1M-token context window.</p><p>4️⃣ Build coding assistants, computer-use agents, and video agents.</p><p>>> Get <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fplatform.minimax.io%2Fsubscribe%2Ftoken-plan%3Futm_source=morningbrew%26utm_medium=newsletter%26utm_campaign=m3_token%26utm_content=Body_5_1b_tokens_month/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/l9qHEIX1d0-p83RNnOMBlTkL_FwGsjNNrFhmq_Jho_k=452" rel="noopener noreferrer nofollow" target="_blank"><span>5.1B tokens / month </span></a>for just $500 a year.</p>
<p>>> <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fagent.minimax.io%2F%3Futm_source=morningbrew%26utm_medium=newsletter%26utm_campaign=m3_code%26utm_content=Body_minimax_code/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/mk-jF2L-Y4gCwORau799-EfNyrLsUJgewWSjypMzLKU=452" rel="noopener noreferrer nofollow" target="_blank"><span>Try MiniMax Code</span></a></p>
<p>>> <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fplatform.minimax.io%2Fdocs%2Fguides%2Fmodels-intro%3Futm_source=TLDR%26utm_medium=newsletter%26utm_campaign=2026-07-08_Primary_Minimax%26utm_content=cta_api/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/iRiDCXyJS76y2J56pEQKltfcO3RB4jCXBR_-GJy6iLs=452" rel="noopener noreferrer nofollow" target="_blank"><span>Start on the API</span></a>
</p>
</span></span></div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
<tr bgcolor="">
<td class="container">
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td style="padding: 0px;">
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;"><span style="font-size: 36px;"><span style="font-size:36px;">🚀</span></span></div></div>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;">
<h1><strong>Headlines & Launches</strong></h1>
</div>
</div>
</td>
</tr>
</tbody>
</table>
<table style="table-layout: fixed; width: 100%;" width="100%">
<tbody>
<tr>
<td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top">
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fclaude.com%2Fblog%2Fcowork-web-mobile%2F%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/f4TgeUq_wPaWkjpqXvG6p13bTIkqA7RhYPraBDJDRYo=452">
<span>
<strong>Claude Cowork on Mobile and Web (1 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Anthropic announced that Claude Cowork sessions and files would become accessible across web and mobile, allowing long-running tasks to continue without an open computer. Beta access began rolling out to Max users, with additional plans to follow.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FiWYDBH/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/O0HO_Mf5sHX-1483Rh-hcdX2U1q0G3JTo6Ns5r2h2aM=452">
<span>
<strong>GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday (1 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
OpenAI is releasing GPT-5.6 Sol, Terra, and Luna tomorrow. It is currently expanding preview access globally. Sol is the flagship model. Terra is a balanced model for everyday work. Luna is a fast and affordable model. Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FYJqOGm/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/ctNC0xwEhPwnSTHJyhhXQ0JdgGyqB2WPuTiyA7ENyhw=452">
<span>
<strong>Microsoft Replaces OpenAI, Anthropic With Own AI in Some Apps (2 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Microsoft is starting to replace OpenAI and Anthropic's models with its own in software products like Excel and Outlook. The move shows that the company is making progress in its efforts to build competitive AI models at a lower cost. Microsoft uses massive quantities of tokens in its products. It gets a lot of its tokens at a discount, but those deals are set to expire soon, and Microsoft is working to make sure it isn't stuck paying whoever the leading AI labs decide to charge.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Farstechnica.com%2Fai%2F2026%2F07%2Ffacing-us-export-controls-chinas-deepseek-plans-to-make-its-own-chips%2F%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/iT16FgfVXSpExN81K__TLcHYYpX_LrVw_ux7kAbLt6c=452">
<span>
<strong>Facing US export controls, China's DeepSeek plans to make its own chips (2 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
DeepSeek is planning to enter the silicon business. The company has been meeting with potential partners in the hardware and silicon space for the past year and has been hiring engineers for the project. It is focusing on data center chips for inference, likely to reduce its reliance on both Huawei and Nvidia. Developing its own chip could give the company an advantage in a market where data center access is likely to remain constrained.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;"><span style="font-size: 36px;">🧠</span></div>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;">
<h1><strong>Deep Dives & Analysis</strong></h1>
</div>
</div>
</td>
</tr>
</tbody>
</table>
<table style="table-layout: fixed; width: 100%;" width="100%">
<tbody>
<tr>
<td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top">
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flilianweng.github.io%2Fposts%2F2026-07-04-harness%2F%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/vzKKN9f-q_adVzc1_izHT6xWqAgRqbiLsc1_mwkiPGI=452">
<span>
<strong>Harness Engineering for Self-Improvement (28 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Recursive self-improvement (RSI) is when ultraintelligent machines can design better machines to improve themselves and surpass humans in all intellectual activities. The feedback loop in modern AI involves models improving the training pipeline and deployment system, which enables better successor models with improved performance. A harness is the system surrounding a base model that orchestrates execution and decides how a model thinks and plans, calls tools and acts, perceives and manages context, stores artifacts, and evaluates results. This post looks at the research around harness engineering and how it contributes to RSI.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FiRKKeL/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/9gSvckaRXQqJ2ipV6It02r0XbRLFh1r-VWw5jVAjgp4=452">
<span>
<strong>MiniMax M3: How Sparse Attention Makes Long-Horizon Agents Practical (11 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Long-horizon tasks are demanding because the context never stops growing. The longer an agent works, the more expensive each step becomes. MiniMax Sparse Attention changes what the model reads at each step, which results in predictable cost at any length and cheap long context with minimal quality cost. It makes long context cheap and fast enough to sit inside iterative systems that need to keep state over time.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FjTUl3i/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/_CnVcxPKci1meH3l7Gu9qdN3zDlMG6qmzdX_HB_lWmA=452">
<span>
<strong>No Space Like J-Space (54 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
The 'J-space' is the area of conscious access where things are available for a model to do what humans could call conscious reasoning. It is named after an interpretability technique called the Jacobian Lens, which computes, for each layer, the average causal effect of changes in the residual stream of a model's eventual outputs, averaged across a wide variety of contexts. This allows researchers to trace what concepts are associated with each layer as the model proceeds through. The J-space and J-lens are remarkably effective tools and mark a major advance in our understanding of LLMs.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;"><span style="font-size: 36px;">🧑💻</span></div>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;">
<h1><strong>Engineering & Research</strong></h1>
</div>
</div>
</td>
</tr>
</tbody>
</table>
<table style="table-layout: fixed; width: 100%;" width="100%">
<tbody>
<tr>
<td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top">
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fstrandsagents.com%2F%3Ftrk=acf96c5e-1749-4bb0-856f-d213b19d8c74%26sc_channel=display%2Bads%26utm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/Agmz4Q5xmJ37D8aI2e1U0tS3ti2c6SXnHY6KcSt5RAk=452">
<span>
<strong>Steering can help Strands agents achieve 100% agent accuracy (Sponsor)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fstrandsagents.com%2F%3Ftrk=acf96c5e-1749-4bb0-856f-d213b19d8c74%26sc_channel=display%2Bads/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/0I2_sm2u9qra9it_ToCy0ibg8lddQfmhPdHJ_K6Ft4k=452" rel="noopener noreferrer nofollow" target="_blank"><span>Strands Agents</span></a> is the open source agent harness SDK built from prod systems inside Amazon. Build an agent harness, control it end-to-end. Steering can give your agents specific feedback so they can correct themselves. You get reliable outcomes without micromanaging. <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fstrandsagents.com%2F%3Ftrk=acf96c5e-1749-4bb0-856f-d213b19d8c74%26sc_channel=display%2Bads/2/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/5eQjCYYXfDs29R0Ns5wnJfzK4q4mfIhpsK9r5_XLdtA=452" rel="noopener noreferrer nofollow" target="_blank"><span>See how</span></a>
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Farxiv.org%2Fhtml%2F2607.02770v1%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/OJwTe_O7Qxs0-3_t8XWGjsw0mlsMWkuEnSb4Av-znuc=452">
<span>
<strong>Gemma 4 Technical Report (17 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Gemma 4 introduces a new generation of open-weight multimodal language models featuring dense and Mixture-of-Experts architectures, with parameters ranging from 2.3B to 31B. It enhances computational efficiency and reasoning via innovations like a thinking mode, long-context efficiency, and a unified encoder-free architecture. These models achieve performance comparable to larger models across various benchmarks and are available under an Apache 2.0 license for widespread adoption.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fabout.fb.com%2Fnews%2F2026%2F07%2Fintroducing-muse-image-meta-ai%2F%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/F7HVJjhwO9lQlTN1G7slo7uByz090simWAT-_9EkfT0=452">
<span>
<strong>Introducing Muse Image: Image Generation Built for Your World (4 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Meta launched Muse Image, an AI-powered model for generating high-quality visuals, now available in Meta AI and integrated with Instagram and WhatsApp. Users can create images using simple prompts, utilize AI effects on Instagram, or redesign rooms with real products via Facebook Marketplace. Soon available globally, Muse Image is free for general use, with more advanced options included in Meta's subscription plans.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.liquid.ai%2Fblog%2Fantidoom%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/1Z390Ri3tM4ToXoqE-fKk0tSJ36f1ZdKvMI-59KJcVg=452">
<span>
<strong>Reducing Doom Loops with Final Token Preference Optimization (12 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Repetitive degeneration is a common failure mode during inference. Antidoom training is a technique that repairs the degenerative repetitive behavior commonly seen after training. It selectively targets the problematic tokens that begin loops to achieve the near-complete elimination of repetitive loops.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fblog.google%2Finnovation-and-ai%2Ftechnology%2Fdevelopers-tools%2Fexpanding-managed-agents-gemini-api%2F%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/EM7GUldJXkqxmSIKkB82UxJD7rFzNdBr7CFD9_RydG4=452">
<span>
<strong>Expanding Managed Agents in Gemini API: background tasks, remote MCP, and more (2 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Google has announced new capabilities for Managed Agents in Gemini API, including background execution, remote MCP server integration, custom function calling, and refreshing credentials across interactions. Users still call a single endpoint, and Gemini handles reasoning, code execution, package installation, file management, and web information, all inside an isolated cloud sandbox. The updates turn managed agents into asynchronous workers that operate inside real development environments without blocking applications.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;"><span style="font-size: 36px;">🎁</span></div></div>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;"><strong><h1>Miscellaneous</h1></strong></div>
</div>
</td>
</tr>
</tbody>
</table>
<table bgcolor="" style="table-layout: fixed; width: 100%;" width="100%">
<tbody>
<tr>
<td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top">
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="mailto:ai@tldr.tech?utm_source=tldrai">
<span>
<strong>TLDR is hiring a curator for TLDR AI! (TLDR Curator, ~5 hrs/week)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Over 1M subscribers read TLDR AI to stay on top of the latest in AI models, research, engineering, and more. If you work in AI and want to help curate it, send your LinkedIn or resume to ai@tldr.tech!
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fsupport.claude.com%2Fen%2Farticles%2F15424964-claude-fable-5-promotional-access%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/qPyTGEA0tacdMzNtbL-Z0ZZEnPRXApQ-4i5WhPV0xG4=452">
<span>
<strong>Claude Fable 5 promotional access (7 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Anthropic is offering access to Claude Fable 5 at no extra cost until July 12. The promotion is available on Pro, Max, Team, and premium seats on seat-based Enterprise plans. Up to 50% of users' weekly subscription limits can be used on Claude Fable 5 during the promotional period. After reaching the limit, users can continue using Fable 5 with usage credits or switch to another model and keep working within their remaining limits.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2FioxpHo/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/zg7kBz4au0vopAVd0gWYd5l50c07CVwvRrhL-jBTkck=452">
<span>
<strong>Microsoft's Real AI Strategy Is Not the Chatbot (8 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Microsoft aims to vertically integrate the enterprise AI chain, controlling the path from customer interaction to cloud services. This strategy extends beyond its chatbot initiatives.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Flinks.tldrnewsletter.com%2F3v92r5/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/TEPJmis_LjbKRJZW2PNXbYvjrnmSmSg2safjzB8yQNE=452">
<span>
<strong>Building a Moat: Self Learning Agents (12 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Self-learning agents should combine agent traces with in-browser user activity to capture both AI failures and human corrections. CopilotKit uses the AG-UI protocol to turn these interactions into procedural and episodic memory, allowing agents to improve over time while keeping learning scoped by user, team, or application.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;"><span style="font-size: 36px;">⚡</span></div></div>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding-top: 0px; padding-bottom: 0px;">
<div class="text-block">
<div style="text-align: center;">
<h1><strong>Quick Links</strong></h1>
</div>
</div>
</td>
</tr>
</tbody>
</table>
<table bgcolor="" style="table-layout: fixed; width: 100%;" width="100%">
<tbody>
<tr>
<td style="padding:0;border-collapse:collapse;border-spacing:0;margin:0;" valign="top">
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fthreadreaderapp.com%2Fthread%2F2074621759891132710.html%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/tLHf-Y7BOaOt3F6e3YCHzfg17GqodVETHzfdg633h_M=452">
<span>
<strong>SpaceX and Cursor are set to release their first jointly developed AI model as soon as Wednesday (1 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
The model is expected to be competitive with Anthropic's Opus 4.8 and OpenAI's GPT-5.5 in some respects.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fthreadreaderapp.com%2Fthread%2F2074255408013955466.html%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/2gAIQGNmaUUCBc-1MU_f1mjtOnd_Fb_ZxRamvJu1PkM=452">
<span>
<strong>GPT-Realtime-2.1-mini is now available in the API (1 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
GPT-Realtime-2.1-mini brings reasoning and tool use to OpenAI's Realtime mini lineup at the same cost as GPT-Realtime-mini.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.lesswrong.com%2Fposts%2FmWpo4Tu87ZSFzwFWB%2Fcalibrating-alignment-evals%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/IwEmkPUGhMiMyU1FRnqLxXOm7XeYMnE-QsUFVpKW4zU=452">
<span>
<strong>Why Alignment Evals Need Calibration (8 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Current alignment benchmarks may overstate safety because models can recognize evaluation settings, game scoring rules, or conceal sleeper behaviors.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
<table align="center" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block">
<span>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fdeveloper.microsoft.com%2Fblog%2Fdont-rewrite-your-cli-for-agents%3Futm_source=tldrai/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/PDH_PUg2zWUB5cgYTTVslqU6s2s4Wz_sjHWA-UkDRZw=452">
<span>
<strong>Agents Work Better with Conventional CLIs (12 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Microsoft tested whether agents performed better when command-line arguments were replaced by a single JSON payload.
</span>
</span>
</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td align="left" style="word-break: break-word; vertical-align: top; padding: 5px 10px;">
<p style="padding: 0; margin: 0; font-size: 22px; color: #000000; line-height: 1.6; font-weight: bold;">
Love TLDR? Tell your friends and get rewards!
</p>
</td>
</tr>
<tr>
<td class="container" style="padding: 0px 10px 15px;">
<div class="text-block">
Share your referral link below with friends to get free TLDR swag!
</div>
</td>
</tr>
<tr>
<td align="left" style="padding: 10px;">
<div class="text-block">
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Frefer.tldr.tech%2F51d793d6%2F2/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/_s_yAno1tUjxs4vLNvEPFgBYV_LG85o-s_76o696LWQ=452" style="color: #464ba4; text-decoration: underline;">https://refer.tldr.tech/51d793d6/2</a>
</div>
</td>
</tr>
<tr></tr>
<tr>
<td align="left" style="padding:5px 10px;">
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fhub.sparklp.co%2Fsub_9a4c16be99eb%2F2/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/2bf6AH29Hp4NzY8hwMm60hDoqUgl7nSqLwKbWHE4agY=452" style="font-size: 16px; line-height: 1.6; padding: 10px 0; display: inline-block; text-decoration: underline;"><span style="mso-text-raise:13pt; text-decoration: underline;">Track your referrals here.</span></a>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td align="left" style="word-break: break-word; vertical-align: top; padding: 5px 10px;">
<p style="padding: 0; margin: 0; font-size: 22px; color: #000000; line-height: 1.6; font-weight: bold;">
Want to advertise in TLDR? 📰
</p>
<div class="text-block" style="margin-top: 10px;">
If your company is interested in reaching an audience of AI professionals and decision makers, you may want to <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fadvertise.tldr.tech%2F%3Futm_source=tldrai%26utm_medium=newsletter%26utm_campaign=advertisecta/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/nwwpuyMuiqpgzyxomRJcPqL-TSZ58at1qsvNsxI8SN8=452"><strong><span>advertise with us</span></strong></a>.
</div>
<br>
<!-- New "Want to work at TLDR?" section -->
<p style="padding: 0; margin: 0; font-size: 22px; color: #000000; line-height: 1.6; font-weight: bold;">
Want to work at TLDR? 💼
</p>
<div class="text-block" style="margin-top: 10px;">
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fjobs.ashbyhq.com%2Ftldr.tech/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/vfMVRXzMlNIioytEf65L2cSvEMIV2tHxaniyg3uKSUQ=452" rel="noopener noreferrer" style="color: #0000EE; text-decoration: underline;" target="_blank"><strong>Apply here</strong></a>,
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fjobs.ashbyhq.com%2Ftldr.tech%2Fc227b917-a6a4-40ce-8950-d3e165357871/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/cBueHhUz_ToSRVmgVrZmvsOZSs3mRKgtxRDhZeS7KZk=452" rel="noopener noreferrer" style="color: #0000EE; text-decoration: underline;" target="_blank"><strong>create your own role</strong></a> or send a friend's resume to <a href="mailto:jobs@tldr.tech" style="color: #0000EE; text-decoration: underline;">jobs@tldr.tech</a> and get $1k if we hire them! TLDR is one of <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Ffeed%2Fupdate%2Furn:li:activity:7401699691039830016%2F/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/gWQdnY3cUsP4tVpnAVv9h89Q4tI8Zm4Z-ABs3fXtqMQ=452" rel="noopener noreferrer" style="color: #0000EE; text-decoration: underline;" target="_blank"><strong>Inc.'s Best Bootstrapped businesses</strong></a> of 2025.
</div>
<br>
<div class="text-block">
If you have any comments or feedback, just respond to this email!
<br>
<br> Thanks for reading,
<br>
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Ftwitter.com%2Fandrewztan/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/ln5M-6hx572Nz0uAhIaA8a0zuWaoeVf9b4wda0VYnuI=452"><span>Andrew Tan</span></a>, <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Fin%2Faliiaminian%2F/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/KogKqX8Ed7tjpI19khXimMrobN9P5ySGSFAGRRjblkc=452"><span>Ali Aminian</span></a>, & <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Fin%2Fjacob-turner-7521a8198%2F/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/7dVWPG8zYhxENRodBS_g8GtXxAtoTBAxwM6r8afRx9Q=452"><span>Jacob Turner</span></a>
<br>
<br>
</div>
<br>
</td>
</tr>
</tbody>
</table>
<table align="center" bgcolor="" border="0" cellpadding="0" cellspacing="0" width="100%">
<tbody>
<tr>
<td class="container" style="padding: 15px 15px;">
<div class="text-block" id="testing-id">
<a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Ftldr.tech%2Fai%2Fmanage%3Femail=silk.theater.56%2540fwdnl.com/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/qhPRQ-SlgrNpTuX2-z3BjciAx7Xuv51WGOs3_Lw3Ais=452">Manage your subscriptions</a> to our other newsletters on tech, startups, and programming. Or if TLDR AI isn't for you, please <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=550aae36-7ac1-11f1-897a-ef5160069b2d%26pt=campaign%26pv=4%26spa=1783515648%26t=1783516611%26s=156fc47867bc276790267846faf464d3a1e0285bf329e64a19015a83fd7c29f0/1/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/aF60AbDavmd1TdDjBjfBNYLKTQ26u2TKK_t2csab6t4=452">unsubscribe</a>.
<br>
</div>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
</tbody>
</table>
<img alt="" src="http://tracking.tldrnewsletter.com/CI0/0100019f41dfc533-365619f2-0a39-4bc1-871c-8aaa25a77664-000000/2JWGxDBLQYhUpIO-goHz-JACQWBk7lK5IRapGGMzYQA=452" style="display: none; width: 1px; height: 1px;">
</body></html>