<!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;">Anthropic's Claude Mythos Preview autonomously identified thousands of zero-day vulnerabilities across major operating systems and browsers </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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/__xemwn1quJIU8FlQkO0D8X_yR_G3kJCPuTVad9Kp90=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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/_V7HwkA8JZoxtDc7--01OwHCS-RkYcuEa2O6kBzMozs=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=d9bb6bb8-333e-11f1-9e13-0153b85a1ae9%26pt=campaign%26t=1775654282%26s=14bcf57bbc534cf62443fc5f82b6d6794a35a85b806ec78717437554ac08c0c7/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/OZ3lcKFee5YxI_eyi_Itxbx0XfHryQdHyzCHYEc8_PQ=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%2Fwandb.ai%2Fsite%2Fresources%2Fwhitepapers%2Fadvancing-physical-ai%2F%3Futm_source=tldr-ai%26utm_medium=cpc%26utm_campaign=industry-campaign%26utm_content=physical-ai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/iV6uCGPB4XUQdZCm8uSUMAzwiRlox1mNc1q17q7bL74=452"><img src="https://images.tldr.tech/weightsandbiases.png" valign="middle" style="vertical-align: middle !important; height: 100%;" alt="Weights and Biases"></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">2026-04-08</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%2Fwandb.ai%2Fsite%2Fresources%2Fwhitepapers%2Fadvancing-physical-ai%2F%3Futm_source=tldr-ai%26utm_medium=cpc%26utm_campaign=industry-campaign%26utm_content=physical-ai/2/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/BKehSkjzjQ2wzV2Rl3BsBeiZ6eqcKGXBt44OvskC3r8=452">
<span>
<strong>Close the sim-to-real gap: Your guide to advancing physical AI (Sponsor)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Physical AI and robotics are the next frontier, but complex real-world variables and specialized multimodal experimentation make building embodied systems challenging.<br>Weights & Biases' new guide, <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwandb.ai%2Fsite%2Fresources%2Fwhitepapers%2Fadvancing-physical-ai%2F%3Futm_source=tldr-ai%26utm_medium=cpc%26utm_campaign=industry-campaign%26utm_content=physical-ai/3/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/h1BvoOzrU4d389ayRp4Kpe-mijLhSg24D56SJFo7U54=452" rel="noopener noreferrer" target="_blank"><span>Advancing physical AI: From learning to embodied intelligence</span></a>, offers the strategies you need to optimize your physical AI workflow, including:
<p></p>
<ul>
<li><strong>Why</strong> a full-stack solution is essential to design, test, and deploy physical AI systems</li>
<li><strong>How</strong> to train and fine-tune multimodal models</li>
<li><strong>Techniques</strong> to close the sim-to-real gap through real-time iteration</li>
<li><strong>Tools</strong> to govern your physical AI systems</li>
</ul>
<p>Get the guide to <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwandb.ai%2Fsite%2Fresources%2Fwhitepapers%2Fadvancing-physical-ai%2F%3Futm_source=tldr-ai%26utm_medium=cpc%26utm_campaign=industry-campaign%26utm_content=physical-ai/4/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/v6CR5z-vHL1heMmFYIk472q_nFugBi80nQ2XQICtQnc=452" rel="noopener noreferrer" target="_blank"><span>Advancing physical AI.</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%2Fwww.anthropic.com%2Fglasswing%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/90Hm6NHFmni4CfMj8fX6AyCrkByAa_BW5YQ7FD_r_2M=452">
<span>
<strong>Project Glasswing: Securing critical software for the AI era (10 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Anthropic's Claude Mythos Preview autonomously identified thousands of zero-day vulnerabilities across major operating systems and browsers. Project Glasswing, in partnership with major tech companies, uses these capabilities to enhance cybersecurity by detecting and fixing vulnerabilities at scale. Anthropic plans to develop safeguards and broaden industry cooperation to address security challenges in the AI era.
</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%2Fz.ai%2Fblog%2Fglm-5.1%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/yfafQHDHvqWFL8MQZufxRl2Z4UPa9BVonVb39dJhy6Y=452">
<span>
<strong>GLM-5.1: Towards Long-Horizon Tasks (14 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
GLM-5.1 is a flagship model for agentic engineering created by Z.ai. It achieves state-of-the-art performance on SWE-Bench Pro. The model is built to stay effective on agent tasks over much longer horizons than previous generations. It can sustain optimization over hundreds of rounds and thousands of tool calls. The model breaks complex problems down, runs experiments, reads results, and identifies blockers with real precision.
</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%2Fwww.greaterwrong.com%2Fposts%2FWjaGAA4xCAXeFpyWm%2Fmy-picture-of-the-present-in-ai%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/z6DP1FZQ61aJenh8nmYPK2GT4kgo0R1-Fj5KnhoPA9k=452">
<span>
<strong>My picture of the present in AI (11 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Ryan Greenblatt is the chief scientist at Redwood Research, a research organization with the mission of aligning superhuman AI. This post goes through some of his best guesses for the current situation of AI. The scenario forecast discusses R&D access regulations, engineering capabilities and qualitative abilities, misalignment and misalignment-related properties, cyber, bioweapons, and economic effects. Some of the claims are highly speculative, while others are better grounded.
</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%2Fred.anthropic.com%2F2026%2Fmythos-preview%2F%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/sDu8Vokt75-9Bcb2P4PlvHZFP6rHbVjZbmUOxQLMwOo=452">
<span>
<strong>Claude Mythos (31 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Anthropic detailed early evaluations of Claude Mythos Preview, highlighting strong performance in discovering zero-day vulnerabilities and reverse-engineering exploits, prompting a coordinated security initiative called Project Glasswing.
</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.mercor.com%2Fblog%2FFinance-tasks-ai-failures-modes%2F%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/NXCdE8MSDzf3jOHtC1qLt-Dm5fVZcC-oQZeUAWrWdPw=452">
<span>
<strong>AI Can't Read an Investor Deck (6 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Current AI models struggle with interpreting complex financial documents, especially with visual data extraction. GPT-5.4, Gemini 3.1 Pro, and Claude Opus 4.6 consistently falter when processing dense charts and images, only achieving 56% to 64% accuracy compared to 72% to 80% with text-only inputs. These findings highlight significant gaps in AI's ability to perform real-world financial reasoning tasks, making the displacement of financial analysts with AI appear premature.
</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%2Fwww.digitalocean.com%2Fdeploy%3Futm_source=tldr-newsletter%26utm_medium=sponsorship%26utm_campaign=deploy_2026_offer%26utm_content=ai-nl-2nd-april-8/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/VDuXvUBXdBHISXbWk90FxHEqG1bw8aUKJ3CmfrfxD9I=452">
<span>
<strong>AI inference conference in SF + $5K in credits if you attend (Sponsor)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
DigitalOcean Deploy is April 28 in SF. One day of technical deep dives on production inference infrastructure, from serverless to dedicated GPUs. Qualifying in-person attendees can receive up to $5,000 in inference credits*. Free to attend, limited spots. <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.digitalocean.com%2Fdeploy%3Futm_source=tldr-newsletter%26utm_medium=sponsorship%26utm_campaign=deploy_2026_offer%26utm_content=ai-nl-2nd-april-8/2/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/ajKRYtcbltBOAOPbmvSg3vb70w7cBX3WbGBvpPfsaBA=452" rel="noopener noreferrer nofollow" target="_blank"><span><strong>Register now!</strong></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%2Fcursor.com%2Fblog%2Fwarp-decode%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/KBFJ1yg1QGzV6IL9sG-jAx-Uak7fVctc4Iro5lMvwtg=452">
<span>
<strong>Faster MoE Inference with Warp Decode (16 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Cursor's “warp decode” is a kernel design that reorganizes MoE inference around output neurons instead of experts. It achieves ~1.8x higher throughput and improved numerical accuracy on Blackwell GPUs.
</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%2Fdevelopers.googleblog.com%2Ftorchtpu-running-pytorch-natively-on-tpus-at-google-scale%2F%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/wD1pyeaX70RNEgwjHwluCCXqnKEWK_s0OMv27SlGr5w=452">
<span>
<strong>TorchTPU: Running PyTorch Natively on TPUs at Google Scale (10 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Google's Tensor Processing Units (TPUs) are foundational to the company's supercomputing infrastructure. The company's custom ASICs power training and serving for Google and its Cloud customers. TorchTPU is a stack that makes it easy for the AI community to access the full capabilities of TPUs. It provides the APIs and tools needed to extract every ounce of compute from Google's hardware. This post takes a look under the hood at the engineering principles behind TorchTPU.
</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%2Fgithub.com%2FWeianMao%2Ftriattention%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/AfH8vGl6qn97iNwIj-0TWPlH7ngKWpIN6qSVrVg6YxI=452">
<span>
<strong>TriAttention for KV Cache Compression (GitHub Repo)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
TriAttention estimates KV importance in pre-RoPE space using stable Q/K centers and distance-based scoring, preserving long-context reasoning quality while sharply reducing KV memory use and improving throughput.
</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%2Fabs%2F2604.04872%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/ET1ItsOGJoqJlVuZAmdndTS_WePw8Xioe5yOaOgo6xA=452">
<span>
<strong>Meta AI Scales RL for ML Engineering Agents (24 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
SandMLE is a framework for building small but structurally realistic MLE environments that made on-policy RL practical for ML engineering agents by cutting execution cost more than 13x.
</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="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.lesswrong.com%2Fposts%2FgfkJp8Mr9sBm83Rcz%2Fwe-re-actually-running-out-of-benchmarks-to-upper-bound-ai%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/QBT-XaxYXsXmaVwiQDpDxVxCM5ugKCCMnlO3AwFBN0I=452">
<span>
<strong>We're actually running out of benchmarks to upper bound AI capabilities (7 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
METR's Time Horizon suite is being saturated. Frontier AI models can reliably do all but maybe a dozen or so tasks in the suite, making it hard to upper bound their time horizon. New benchmarks are becoming more expensive to grade and create. The situation will likely get worse as AI progress continues. It is likely that, by mid-2027, no benchmark score from a 2026 or earlier benchmark will be able to rule out dangerous capabilities from frontier AI systems.
</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%2F18ytzS/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/yLzgi6Mzk2npqG1EF6GzDMg1Rcd8QQewTGt0EX3qcqA=452">
<span>
<strong>Elon Musk Asks for OpenAI's Nonprofit to Get Any Damages From His Lawsuit (3 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Elon Musk's lawsuit against OpenAI is expected to go to trial later this month in Oakland, California. Musk has amended the lawsuit to ask that any damages he might win be awarded to OpenAI's charitable arm rather than to himself. The amendment also asks that OpenAI CEO Sam Altman be removed from the OpenAI nonprofit's board. Musk is seeking more than $150 billion in damages from both OpenAI and Microsoft as he believes OpenAI strayed from its non-profit mission and defrauded him as a donor in seeking to convert to a for-profit company.
</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%2Fjobs.ashbyhq.com%2Ftldr.tech%2F3b21aaf8-dea5-4127-be71-602d30e5001e%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/gapDrL-QKtbxIAATzaoC_5WtScaJeo3KPDVasA7RwZE=452">
<span>
<strong>TLDR is hiring a Senior Software Engineer, Applied AI ($250k-$350k, Fully Remote)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
TLDR's Applied AI team is tasked with making every process at TLDR legible to code, runnable by anyone, and composable into larger workflows. Join a small, fast moving team using the latest AI tools with an unlimited token budget. <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fjobs.ashbyhq.com%2Ftldr.tech%2F3b21aaf8-dea5-4127-be71-602d30e5001e/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/xb1ELHuxb0vWlh0U2Df2ub1hf1M2DrlzjxKVgWx0L4g=452" rel="noopener noreferrer" target="_blank"><span>Learn more</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%2Fepochai.substack.com%2Fp%2Fgoogle-controls-the-most-ai-computing%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/OI4baIi-c4fRmGlYF-wwGx037mgU14xcQhI05SbpzUY=452">
<span>
<strong>Google controls the most AI computing power, driven by its custom TPUs (1 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Google holds around 25% of all compute sold since 2022.
</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%2Fgithub.com%2Fmattmireles%2Fgemma-tuner-multimodal%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/uDFrkoHaSCNp3HiYGPrsxkh9OSwP66_TgT2SvpfHsdc=452">
<span>
<strong>Gemma Multimodal Fine-Tuner (GitHub Repo)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Gemma Multimodal Fine-Tuner lets Mac users fine-tune Gemma on text, images, and audio on remote data.
</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%2Ftomtunguz.com%2Fanthropic-most-valuable-company%2F%3Futm_source=tldrai/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/TN8u-bXgKYj1A8iffuU8GVH9KEscAO-g9-mIbOaSaLc=452">
<span>
<strong>When Will Anthropic Surpass NVIDIA? (1 minute read)</strong>
</span>
</a>
<br>
<br>
<span style="font-family: "Helvetica Neue", Helvetica, Arial, Verdana, sans-serif;">
Anthropic reached $10 billion in revenue in under four years, far outpacing other software companies like ServiceNow and Shopify.
</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%2F2/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/LoNvu65g1smakbr_1xC1UxVB2fOHXPIaRkjfpnjiTeQ=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%2F2/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/yidvwQ83yI0Fxgx1z0lo5BStYC4gnIjmpAwfBYf210o=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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/o_EZIb6GRHK0J4W9X9dG9cD8Pq8sUKcLXvuUNsYAlMM=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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/ZmA21Mr2cehJsui_xspwOCBI3WeVrCP8rGpuzfoNY0k=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-d3e165357871/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/LlGAP4UOsCEriFv83IB-jle7nuzh6aKFaKDfX48UVH4=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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/IZEZmJlR3T_CsRscSHveA5PPax7qeotxGQYiQQbB-0U=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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/zGtKwkdBjeC16oBVH4ZO_nWnKz-N11NBxFwxKDUiecQ=452"><span>Andrew Tan</span></a>, <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Fin%2Faliiaminian%2F/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/IJSFU6xBOb6_UnnLrxYEp_mXV0Lx4VtM-UdAvNjDswk=452"><span>Ali Aminian</span></a>, & <a href="https://tracking.tldrnewsletter.com/CL0/https:%2F%2Fwww.linkedin.com%2Fin%2Fjacob-turner-7521a8198%2F/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/klFNqCkZUkNMZfo1MOKzFGQOKmvjVa6Ok3CN1Ysq8wk=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/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/b00Kpdtq7dhZy6g3KGQ9wQV5sdceyPlp51trruVhwGg=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=d9bb6bb8-333e-11f1-9e13-0153b85a1ae9%26pt=campaign%26pv=4%26spa=1775653389%26t=1775654282%26s=400821caaf088f0114ef2db36d97e53fff89829169210c3768332ea4c2912341/1/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/fiCO1WeX-25UWGmXqcTk69oN5B4YSBL9zTQTo7l5umg=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/CI0/0100019d6d3e2643-033f8786-31f5-4a28-b246-342a7e473baf-000000/ZISHWHmAezdPThFjbbX_yJ_NHp5Joh8Yi2IFSPSNsj8=452" style="display: none; width: 1px; height: 1px;">
</body></html>