<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!Lp4D!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!Lp4D!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 424w, https://substackcdn.com/image/fetch/$s_!Lp4D!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 848w, https://substackcdn.com/image/fetch/$s_!Lp4D!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 1272w, https://substackcdn.com/image/fetch/$s_!Lp4D!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!Lp4D!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png" width="1238" height="706" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/e7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:706,&quot;width&quot;:1238,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:738665,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/205007533?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!Lp4D!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 424w, https://substackcdn.com/image/fetch/$s_!Lp4D!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 848w, https://substackcdn.com/image/fetch/$s_!Lp4D!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 1272w, https://substackcdn.com/image/fetch/$s_!Lp4D!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7d4a5e8-a25b-4dfc-ad58-967ceef92834_1238x706.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey Devs! As always, this week was busy; a lot of things happened. Here are the most important things I think you should know.</p><div><hr></div><h3><strong>Fable 5 returns with stronger safeguards</strong></h3><p>On June 30,<span> </span><strong><a href="https://www.linkedin.com/company/anthropicresearch/">Anthropic</a></strong><span> </span>c announced that the U.S. government lifted export controls on its latest large language models. The ban had halted access to Fable 5 (consumer-focused) and Mythos 5 (cybersecurity-oriented) due to concerns about jailbreaks that could expose vulnerabilities.</p><p>If you rely on Anthropic models for coding or research, update your usage plans and be prepared for more false-positive safety flags when working on security-related tasks. Follow Anthropic&#8217;s guidance to appeal misclassifications.</p><h3><strong>AI bot control arrives for web admins</strong></h3><p><strong><a href="https://www.linkedin.com/company/cloudflaredev/">Cloudflare Developers</a></strong><span> </span>July 1 blog post introduces categories for AI bots (search, agent, and training) and allows site owners to decide whether bots can crawl or train on their content.<span> </span><em>Tip</em>: Developers hosting documentation should update their bot policies to protect proprietary data or opt-in to training traffic.</p><h3><strong><a href="https://www.linkedin.com/company/meta/">Meta</a><span> </span>decodes brain waves into text</strong></h3><p>Brain2Qwerty v2 uses magnetoencephalography and a deep-learning decoder to turn brain activity into text with around 61 % word accuracy. Meta made the full dataset and training code public. This research could inspire hands-free input tools for accessibility or VR applications.</p><h3><strong>Anthropic explores its own AI chip</strong></h3><p><strong><a href="https://www.linkedin.com/company/techcrunch/">TechCrunch</a></strong><span> </span>reports that Anthropic is in talks with Samsung about designing a custom chip; details on intended use or performance remain undecided. The move follows OpenAI&#8217;s custom inference chip &#8220;Jalape&#241;o,&#8221; which aims to match Nvidia&#8217;s top GPUs. Expect competition in AI hardware to accelerate.</p><h3><strong>Google&#8217;s June AI updates</strong></h3><p>Google&#8217;s June recap highlighted the open-source Gemma 4 12B local model, which requires only 16 GB of RAM and supports multimodal tasks. Google also released Nano Banana 2 Lite, a tiny model for mobile devices, and the Gemini 3.5 Flash service for low-latency workloads. Developers can experiment with these models via Google AI Studio.</p><div id="youtube2-fppaj5s4CeM" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;fppaj5s4CeM&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><h3><strong><a href="https://www.linkedin.com/company/apple/">Apple</a><span> </span>Intelligence and a smarter Siri</strong></h3><p>At WWDC26 (late June), Apple unveiled the next generation of Apple Intelligence alongside Siri AI, a more context-aware voice assistant. Siri AI draws on on-device personal context and world knowledge to answer questions and perform tasks across apps. Apple says these features will arrive with iOS 27 and macOS 27 this fall; they are restricted to recent devices and currently unavailable in China.</p><h3><strong>Codex Subagents for Beginners: Common Mistakes to Avoid</strong></h3><p>Are you using Codex but haven&#8217;t tapped into the power of Subagents yet? In this video, I break down everything you need to know to set up and run custom sub-agents to supercharge your development workflow.</p><div id="youtube2-xItvCoWNPVo" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;xItvCoWNPVo&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><p></p><h2><strong>Developer tip of the week</strong></h2><p><em>To reduce false positives when using<span> </span><strong>Fable 5</strong><span> </span>for security research, Anthropic recommends framing vulnerability testing prompts as<span> </span><strong>defensive or academic inquiries</strong><span> </span>rather than exploit instructions. For example, instead of asking &#8220;How can I exploit CVE-XXXX?&#8221;, phrase it as &#8220;Explain how this vulnerability works and how to mitigate it.&#8221; This reduces the chance of triggering the new classifier.</em></p><div><hr></div><p>And that&#8217;s it! See you next week! Cheers, proflead! ;)</p>

AI News for Developers #27: Fable 5 Returns, Cloudflare Bot Controls, and Meta Brain2Qwerty

<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!NtP6!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!NtP6!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 424w, https://substackcdn.com/image/fetch/$s_!NtP6!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 848w, https://substackcdn.com/image/fetch/$s_!NtP6!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 1272w, https://substackcdn.com/image/fetch/$s_!NtP6!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!NtP6!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png" width="1301" height="729" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:729,&quot;width&quot;:1301,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1665395,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/202808821?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!NtP6!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 424w, https://substackcdn.com/image/fetch/$s_!NtP6!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 848w, https://substackcdn.com/image/fetch/$s_!NtP6!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 1272w, https://substackcdn.com/image/fetch/$s_!NtP6!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fe484c7-de73-4355-a6ca-b446ca1bf142_1301x729.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey Devs! As always, this week was busy; a lot of things happened. Here are the most important things I think you should know.</p><div><hr></div><h2><strong>Meta launches AI Mode in Facebook search</strong></h2><p><strong><a href="https://www.linkedin.com/company/meta/">Meta</a></strong> rolled out <strong>AI Mode</strong> globally on 15 June. It appears alongside existing search options and uses Meta AI to synthesize conversational answers from public posts, Groups, and Reels rather than returning a ranked list of links.</p><p><em><strong>Read more: <a href="https://about.fb.com/news/2026/06/new-ai-tools-to-help-you-make-things-happen-on-facebook/">https://about.fb.com/news/2026/06/new-ai-tools-to-help-you-make-things-happen-on-facebook/</a></strong></em></p><div><hr></div><h2><strong>OpenAI releases LifeSciBench</strong></h2><p><strong><a href="https://www.linkedin.com/company/openai/">OpenAI</a></strong> announced LifeSciBench, an expert&#8209;written benchmark for life&#8209;science research tasks. It covers 750 tasks across seven workflows and biological domains, involving 1,062 task artifacts, 173 scientist contributors, and 453 expert reviewers.</p><p>LifeSciBench could become a reference point for AI systems for biotech and pharmaceutical tasks, reminding developers to consider domain&#8209;specific performance rather than generic metrics alone.</p><p><em><strong>Read more: <a href="https://openai.com/index/introducing-life-sci-bench/">https://openai.com/index/introducing-life-sci-bench/</a></strong></em></p><div><hr></div><h2><strong>Google makes Gemini 2.5 Flash the default</strong></h2><p><strong><a href="https://www.linkedin.com/company/google/">Google</a></strong> replaced Gemini 2.5 Pro with Gemini 2.5 Flash as the default across its consumer Gemini products. Flash is a cost&#8209;optimised model priced at $0.30 per million input tokens, offers an optional thinking mode, and supports a 1 million&#8209;token context window.</p><div><hr></div><h2><strong>SpaceX to acquire Cursor for $60 billion</strong></h2><p>Days after its record&#8209;setting IPO, SpaceX agreed to buy AI coding startup Cursor (formerly Anysphere) in a $60 billion all&#8209;stock deal. The purchase, set to close in the third quarter of 2026, follows a tie&#8209;up announced two months earlier.</p><p>SpaceX&#8217;s purchase highlights how critical developer tools have become in the AI race. Cursor&#8217;s integration with SpaceX&#8217;s AI infrastructure could influence the future of agentic coding workflows and underscores the value of code&#8209;generation platforms.</p><p><em><strong>Read more: </strong></em></p><div class="twitter-embed" data-attrs="{&quot;url&quot;:&quot;https://x.com/SpaceX/status/2066873915717136548&quot;,&quot;full_text&quot;:&quot;SpaceX has exercised the option to acquire <span class=\&quot;tweet-fake-link\&quot;>@cursor_ai</span> in an all-stock transaction with the goal of building the world&#8217;s most useful AI models.\n\nFor the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon.&quot;,&quot;username&quot;:&quot;SpaceX&quot;,&quot;name&quot;:&quot;SpaceX&quot;,&quot;profile_image_url&quot;:&quot;https://pbs.substack.com/profile_images/1697749409851985920/HbrI04tM_normal.jpg&quot;,&quot;date&quot;:&quot;2026-06-16T13:21:56.000Z&quot;,&quot;photos&quot;:[],&quot;quoted_tweet&quot;:{&quot;full_text&quot;:&quot;SpaceXAI and @cursor_ai are now working closely together to create the world&#8217;s best coding and knowledge work AI.\n\nThe combination of Cursor&#8217;s leading product and distribution to expert software engineers with SpaceX&#8217;s million H100 equivalent Colossus training supercomputer will&quot;,&quot;username&quot;:&quot;SpaceX&quot;,&quot;name&quot;:&quot;SpaceX&quot;,&quot;profile_image_url&quot;:&quot;https://pbs.substack.com/profile_images/1697749409851985920/HbrI04tM_normal.jpg&quot;},&quot;reply_count&quot;:1657,&quot;retweet_count&quot;:4286,&quot;like_count&quot;:36545,&quot;impression_count&quot;:25568839,&quot;expanded_url&quot;:null,&quot;video_url&quot;:null,&quot;belowTheFold&quot;:true}" data-component-name="Twitter2ToDOM"></div><div><hr></div><h2><strong>Google released Gemma 4 12B</strong></h2><p>Gemma 4 12B, a multimodal model licensed under Apache 2.0 that runs on 16 GB devices and handles text, images, and audio. Cohere shipped its first open&#8209;source coding model at 30 B parameters, and Moonshot&#8217;s K2.7&#8209;Code is reported to deliver better coding performance with 30 % fewer reasoning tokens. Roblox open&#8209;sourced a 3&#8209;D&#8209;generation model.</p><p><em><strong>Read more: <a href="https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12b/">https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12b/</a></strong></em></p><div><hr></div><p>And that&#8217;s it! See you next week! Cheers, proflead! ;)</p>

AI News for Developers #26: Meta AI Mode, OpenAI LifeSciBench, and SpaceX’s $60B Cursor Deal

<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!TZbU!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!TZbU!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 424w, https://substackcdn.com/image/fetch/$s_!TZbU!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 848w, https://substackcdn.com/image/fetch/$s_!TZbU!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 1272w, https://substackcdn.com/image/fetch/$s_!TZbU!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!TZbU!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png" width="1371" height="775" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:775,&quot;width&quot;:1371,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1846208,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/200849485?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!TZbU!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 424w, https://substackcdn.com/image/fetch/$s_!TZbU!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 848w, https://substackcdn.com/image/fetch/$s_!TZbU!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 1272w, https://substackcdn.com/image/fetch/$s_!TZbU!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F6e0d2787-edaa-418a-8b2b-03f7cfb0d392_1371x775.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey, Devs! Vlad here! The week of <strong>31 May &#8211; 6 June 2026</strong> proved that the agentic future is arriving fast. Nvidia and Microsoft doubled down on PCs that can run AI locally, open&#8209;source labs released models aimed at routing and coding, GitHub overhauled Copilot billing, and even governments weighed in on how to shepherd the technology. Here are the highlights!</p><div><hr></div><h3><strong>Nvidia unveils RTX Spark PC chip for local AI</strong></h3><p><strong><a href="https://www.linkedin.com/company/nvidia/">NVIDIA</a></strong> used Computex to introduce RTX Spark, a chip that brings inference directly to laptops and desktops and promises to make PCs into agentic AI machines. CEO Jensen Huang said the device is part of a multi&#8209;year collaboration with Microsoft to &#8220;reinvent the PC&#8221; for the AI era. Co&#8209;developed with MediaTek, RTX Spark chips will appear this fall in machines from Dell, HP, Lenovo, ASUS, Microsoft and others. The chip is designed to run autonomous agents locally, replacing cloud&#8209;only reliance. Analysts noted that bringing agents to the edge could transform PCs into &#8220;useful agentic AI personal computers&#8221; and dubbed 2026 the year of agents.</p><p><em><strong>Read more: <a href="https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark">https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark</a></strong></em></p><div><hr></div><h3><strong>Microsoft Build: Project Solara and secure agentic Windows</strong></h3><p>At its Build conference, <strong><a href="https://www.linkedin.com/company/microsoft/">Microsoft</a></strong> shifted away from traditional apps toward agent&#8209;driven experiences. CEO Satya Nadella introduced Project Solara, a family of prototype devices (badge&#8209;size units and smart&#8209;speaker&#8209;like gadgets) that host AI agents instead of an OS and apps. Microsoft also previewed the Surface RTX Spark Dev Box, an Nvidia&#8209;powered PC capable of running a 120&#8209;billion&#8209;parameter model locally. To make OpenClaw &#8212; the open&#8209;source agent orchestration framework &#8212; enterprise&#8209;ready on Windows, Microsoft announced a sandboxing tool that safely runs agent groups. A new AI agent called Scout will roll out inside Copilot to gather emails and messages that require user decisions, and Microsoft&#8217;s internal AI unit unveiled MAI Thinking&#8209;1, a reasoning model claimed to rival Anthropic&#8217;s Claude Opus 4.6.</p><p><em><strong>Read more: <a href="https://commandline.microsoft.com/project-solara-build-2026/">https://commandline.microsoft.com/project-solara-build-2026/</a></strong></em></p><div><hr></div><h3><strong>MiniMax M3: open&#8209;weight frontier model with 1M&#8209;token context</strong></h3><p>Chinese lab <strong><a href="https://www.linkedin.com/company/minimax-ai/">MiniMax</a></strong> released M3, an open&#8209;weight model built on its new MiniMax Sparse Attention (MSA) architecture. The company said M3 is the first open&#8209;weight model to combine frontier&#8209;level coding performance, a million&#8209;token context window, and native multimodality. M3 is tuned for coding and agentic work and reportedly surpasses GPT&#8209;5.5 and Gemini 3.1 Pro on SWE&#8209;Bench Pro. MSA enables a 1 M&#8209;token context while using only a fraction of compute; at 1 million tokens, M3&#8217;s per&#8209;token compute is just one&#8209;twentieth of its previous&#8209;generation model. MiniMax positions M3 as an open competitor to closed frontier models and provides weights via its API and open&#8209;source frameworks.</p><p><em><strong>Read more: <a href="https://www.minimax.io/blog/minimax-m3">https://www.minimax.io/blog/minimax-m3</a></strong></em></p><div><hr></div><h3><strong>JetBrains open&#8209;sources Mellum2 for AI routing and sub&#8209;agents</strong></h3><p><strong><a href="https://www.linkedin.com/company/jetbrains/">JetBrains</a></strong>&#8217; AI team open&#8209;sourced <strong>Mellum2</strong>, a 12&#8209;billion&#8209;parameter mixture&#8209;of&#8209;experts model designed for routing, question&#8209;answering and sub&#8209;agent tasks. Built from scratch and released under the Apache 2.0 licence, Mellum2 aims to handle the bottlenecks of production&#8209;scale AI such as latency, throughput and cost. Its mixture&#8209;of&#8209;experts design activates only 2.5 billion parameters per token, reducing compute while maintaining performance. JetBrains notes that Mellum2 excels in software engineering environments and can be run locally or fine&#8209;tuned to orchestrate complex agent workflows. Developers can experiment with Mellum2 through JetBrains&#8217; toolkit or download weights from Hugging Face.</p><p><em><strong>Read more: <a href="https://blog.jetbrains.com/ai/2026/06/mellum2-goes-open-source-a-fast-model-for-ai-workflows/">https://blog.jetbrains.com/ai/2026/06/mellum2-goes-open-source-a-fast-model-for-ai-workflows/</a></strong></em></p><div><hr></div><h3><strong>GitHub Copilot transitions to usage&#8209;based billing</strong></h3><p><strong><a href="https://www.linkedin.com/company/github/">GitHub</a></strong> announced that, starting <strong>June 1</strong>, all Copilot plans will use <strong>GitHub AI Credits</strong> instead of premium request units. Each plan will include a monthly allotment of credits, and consumption will be based on token usage (input, output and cached tokens) at the published API rates. Base prices remain unchanged &#8212; Pro ($10/mo), Pro+ ($39/mo), Business ($19/user/mo) and Enterprise ($39/user/mo) &#8212; but heavy users may need to purchase extra credits. Code completions and Next&#8209;Edit suggestions are still unlimited and do not consume credits.</p><p><em><strong>Read more: <a href="https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/">https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/</a></strong></em></p><div><hr></div><h3><strong>Anthropic urges coordinated pause in AI development</strong></h3><p><strong><a href="https://www.linkedin.com/company/anthropic-courses/">Anthropic</a></strong> published a blog post urging major AI labs to consider a <strong>coordinated and verifiable pause</strong> in model development, warning that rapid progress could allow AI systems to <strong>improve themselves without human intervention</strong>. The company&#8217;s leaders said the ability of AI to complete tasks on its own has been doubling every four months and could soon reach &#8220;recursive self&#8209;improvement,&#8221; raising safety concerns.</p><p><em><strong>Read more: <a href="https://www.anthropic.com/institute/recursive-self-improvement">https://www.anthropic.com/institute/recursive-self-improvement</a></strong></em></p><div><hr></div><p>See you next week! Cheers, proflead! ;)</p>

AI News for Developers #25: Nvidia RTX Spark, Microsoft Solara, and MiniMax M3

<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!VYsC!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!VYsC!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 424w, https://substackcdn.com/image/fetch/$s_!VYsC!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 848w, https://substackcdn.com/image/fetch/$s_!VYsC!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 1272w, https://substackcdn.com/image/fetch/$s_!VYsC!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!VYsC!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png" width="1456" height="855" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:855,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2128377,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/198930316?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!VYsC!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 424w, https://substackcdn.com/image/fetch/$s_!VYsC!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 848w, https://substackcdn.com/image/fetch/$s_!VYsC!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 1272w, https://substackcdn.com/image/fetch/$s_!VYsC!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2dedb08e-d876-4bcd-a827-a83a3b99a880_1636x961.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey, Devs! Vlad here! This week was packed with big updates: Google I/O, Cursor 2.5, WordPress Armstrong, and many other cool things you need to know about. Let&#8217;s jump in!</p><h2><strong>Google unifies AI coding tools under Antigravity 2.0</strong></h2><p>Gemini CLI, Code Assist, and AI Studio into a single Antigravity 2.0 platform. The platform includes a desktop app and a CLI, integrates deeply with the Gemini Enterprise Agent Platform, and aims to unify agentic workflows across desktop, terminal, and cloud environments. Existing scripts and CI workflows will need to be migrated before the previous tools are deprecated on 18 June.</p><p><em><strong>Read more: <a href="https://antigravity.google/blog/introducing-google-antigravity-2-0">https://antigravity.google/blog/introducing-google-antigravity-2-0</a></strong></em></p><div><hr></div><h2><strong>Claude for Small Business launch</strong></h2><p>On 13 May, Anthropic launched <strong>Claude for Small Business</strong>, a suite of 15 pre&#8209;built agentic workflows that integrate with QuickBooks, PayPal, HubSpot, and other services. The initiative aims to close the AI adoption gap for small businesses by providing ready&#8209;to&#8209;use automations.</p><p><em><strong>Read more: <a href="https://www.anthropic.com/news/claude-for-small-business">https://www.anthropic.com/news/claude-for-small-business</a></strong></em></p><div><hr></div><h2><strong>Cursor Composer 2.5</strong></h2><p>Cursor released version 2.5 of its coding agent. Composer 2.5, built on Moonshot&#8217;s Kimi K2.5 checkpoint, offers significant improvements in handling long-running coding tasks, follows complex instructions more reliably, and uses targeted reinforcement learning to correct specific mistakes. The model is priced at $0.50 per million input tokens and $2.50 per million output tokens (fast variant $3.00/$15.00) and positions itself as lower&#8209;cost than other frontier models.</p><p><em><strong>Read more: <a href="https://cursor.com/blog/composer-2-5">https://cursor.com/blog/composer-2-5</a></strong></em></p><div><hr></div><h2><strong>WordPress 7.0 &#8220;Armstrong&#8221;</strong></h2><p>The official WordPress release introduced a major upgrade named after LouisArmstrong. WordPress 7.0 lays &#8220;the foundation for AI across the WordPress experience&#8221; and provides a new AI Client and Abilities API that can communicate with generative models.</p><p><em><strong>Read more: <a href="https://wordpress.org/news/2026/05/armstrong/">https://wordpress.org/news/2026/05/armstrong/</a></strong></em></p><div><hr></div><h2><strong>Microsoft open&#8209;sources Rampart and Clarity</strong></h2><p>To operationalize AI agent safety, Microsoft released two open-source tools. Rampart, built on PyRIT, converts red&#8209;team findings into repeatable tests run during development and deployment to catch prompt injection, unsafe tool use, and privilege escalation. Clarity guides engineers through structured conversations about agent behavior, permissions, and trust boundaries, storing decision logs as Markdown in the repository. These tools aim to make continuous AI safety engineering part of standard DevOps.</p><p><em><strong>Learn more: </strong></em></p><div id="youtube2-o-AxGWAa06Y" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;o-AxGWAa06Y&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><div><hr></div><h2><strong>The New Codex /goal Command Explained</strong></h2><p>Codex /goal is the most powerful feature in OpenAI Codex CLI right now, and most developers don&#8217;t know it exists yet. In this video, I show you exactly how to use the /goal command to run autonomous multi-hour coding sessions, how to write goals that actually work, and when to use /goal vs. regular prompts.</p><div id="youtube2-ePGrgs1-DsY" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;ePGrgs1-DsY&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><p><em><strong>Watch on YouTube: <a href="https://youtu.be/ePGrgs1-DsY">Codex /goal</a></strong></em></p><div><hr></div><p>See you next week! Cheers, proflead! ;)</p>

AI News for Developers #24: Antigravity 2.0, Cursor 2.5, and AI Safety

<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!a7dV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!a7dV!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 424w, https://substackcdn.com/image/fetch/$s_!a7dV!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 848w, https://substackcdn.com/image/fetch/$s_!a7dV!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 1272w, https://substackcdn.com/image/fetch/$s_!a7dV!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!a7dV!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png" width="1456" height="971" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:971,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2107250,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/197075373?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!a7dV!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 424w, https://substackcdn.com/image/fetch/$s_!a7dV!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 848w, https://substackcdn.com/image/fetch/$s_!a7dV!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 1272w, https://substackcdn.com/image/fetch/$s_!a7dV!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2643cafb-e53f-4f7d-a2eb-487a2bfeb467_1536x1024.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey devs! There&#8217;s a lot of noteworthy AI news this week. Here&#8217;s what&#8217;s new:</p><div><hr></div><h2><strong>GPT-Realtime-2, GPT-Realtime-Translate and GPT-Realtime-Whispe</strong></h2><p>GPT-Realtime-2 brings GPT-5-class reasoning to voice interactions and offers features such as tone control, parallel tool calls, and expanded context (up to 128k tokens). GPT-Realtime-Translate handles live translation from over 70 languages into 13 outputs. GPT-Realtime-Whisper delivers ultra-low-latency streaming transcription.</p><p><em><strong>Read more: <a href="https://openai.com/index/advancing-voice-intelligence-with-new-models-in-the-api/">https://openai.com/index/advancing-voice-intelligence-with-new-models-in-the-api/</a></strong></em></p><div><hr></div><h2><strong>Natural Language Autoencoders (NLAs)</strong></h2><p>Anthropic researchers introduced Natural Language Autoencoders (NLAs), an unsupervised method that translates a model&#8217;s internal activations into readable text and back again. An NLA consists of an activation verbalizer that describes a hidden activation in natural language and an activation reconstructor that maps the explanation back to the original vector. When trained jointly, these modules produce explanations that grow more informative over time.</p><p><em><strong>Read more: <a href="https://transformer-circuits.pub/2026/nla/">https://transformer-circuits.pub/2026/nla/</a></strong></em></p><div><hr></div><h2><strong>OpenAI rolled out an optional Trusted Contact</strong></h2><p>Adult users can nominate a trusted friend or family member; if the chatbot detects a conversation indicating possible self-harm, it encourages the user to reach out and sends a brief alert to the contact. The system combines automated detection with human review and aims to notify contacts within an hour. Alerts omit detailed conversation content to protect privacy. This safeguard follows a series of lawsuits alleging ChatGPT&#8217;s responses contributed to suicides and builds on earlier parental controls.</p><p><em><strong>Read more: <a href="https://openai.com/index/introducing-trusted-contact-in-chatgpt/">https://openai.com/index/introducing-trusted-contact-in-chatgpt/</a></strong></em></p><div><hr></div><h2><strong>Gemma 4: efficient open-source models from Google</strong></h2><p>Google DeepMind announced the <strong>Gemma 4</strong> family of open models, designed for advanced reasoning and agentic workflows. Available in four sizes-Effective 2B, Effective 4B, a 26B mixture-of-experts, and a 31B dense model-Gemma 4 delivers high intelligence per parameter and outperforms models many times its size.</p><p><em><strong>Read more: <a href="https://deepmind.google/models/gemma/gemma-4/">https://deepmind.google/models/gemma/gemma-4/</a></strong></em></p><div><hr></div><p>See you next week! Cheers, proflead! ;)</p>

AI News for Developers #23: Realtime Voice, Interpretable Models, Safer Chats, Gemma 4 and more

<p></p><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!Fb_I!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!Fb_I!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!Fb_I!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!Fb_I!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!Fb_I!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!Fb_I!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/144917f2-66dd-4002-a597-6f66212c3245_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1927992,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/196314192?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!Fb_I!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!Fb_I!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!Fb_I!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!Fb_I!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F144917f2-66dd-4002-a597-6f66212c3245_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey Devs! This past week was packed with AI developments that could influence how you build and deploy software. Here are the highlights:</p><h3><strong>Google&#8217;s code goes agentic</strong></h3><p>Sundar Pichai revealed that about 75% of Google&#8217;s new code is now AI-generated and reviewed by engineers, up from 50% last autumn. Google is shifting to &#8220;agentic&#8221; workflows, where developers orchestrate digital task forces of coding agents, enabling complex migrations to be completed 6 times faster.</p><p><em><strong>Read more: <a href="https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/cloud-next-2026-sundar-pichai/">Google Cloud Next &#8216;26 announcements</a></strong></em></p><div><hr></div><h3><strong>Meta&#8217;s business AI assistant scales</strong></h3><p>Meta said its business AI assistant beta now manages over 10 million conversations per week. Powered by the Muse Spark model, the assistant is rolling out to more regions, and over 8 million advertisers are using Meta&#8217;s AI-creative tools. Zuckerberg signaled monetization may come soon.</p><p><em><strong>Learn more: <a href="https://web.facebook.com/business/ai/business-assistant">https://web.facebook.com/business/ai/business-assistant</a></strong></em></p><div><hr></div><h3><strong>OpenAI models land on AWS</strong></h3><p>OpenAI and Amazon deepened their partnership: GPT-5.5, GPT-4.5, and the Codex coding agent are now available via Amazon Bedrock. OpenAI is committed to spending over $100 billion on AWS, and developers can access the models without leaving Amazon&#8217;s cloud.</p><div><hr></div><h3><strong>Autodata: an automatic data scientist for synthetic data</strong></h3><p>Meta AI researchers published a blog on Autodata, a method that lets AI agents act as data scientists. The agent generates training or evaluation data, analyzes its quality, and iteratively refines the data to improve it. The process uses a main LLM &#8220;data scientist&#8221; supported by sub-agents: one generates examples, others act as weak and strong solvers, and a judge evaluates the results. By meta-optimizing this system, Autodata can produce higher-quality datasets more efficiently than previous self-instruct methods.</p><p><em><strong>Read more: <a href="https://facebookresearch.github.io/RAM/blogs/autodata/">Meta AI RAM blog</a></strong></em></p><div><hr></div><h3><strong>Google Agents CLI</strong></h3><p>At Cloud Next &#8216;26, Google introduced <strong>agents-cli</strong>, a command-line tool and skills package that turns your coding agent (Gemini CLI, Claude Code, Codex, etc.) into an expert at building, evaluating, and deploying agents via Google&#8217;s Agent Development Kit (ADK). For a hands-on demonstration, check out <strong>my video &#8220;Google Agents CLI: Build AI Agents FAST (Beginner Guide)&#8221;</strong> on YouTube.</p><div id="youtube2-C-0DIcFVt4Q" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;C-0DIcFVt4Q&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><p>The video walks through installing the CLI, generating a design spec, scaffolding an agent project, and deploying it to Cloud Run, all in just a few minutes.</p><div><hr></div><p>See you next week! Cheers, proflead! ;)</p><p></p>

AI News for Developers #22: Agentic Coding, AWS x OpenAI, and AI That Builds Its Own Data

<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!wHnL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!wHnL!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!wHnL!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!wHnL!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!wHnL!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!wHnL!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1956329,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/195413798?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!wHnL!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!wHnL!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!wHnL!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!wHnL!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F22a2d6dd-b418-49d5-906d-a7a9db629e7b_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><p>Hey Devs! During this week, several important AI announcements were made that directly affect developers. Let&#8217;s jump into it. ;)</p><div><hr></div><h3><strong>GPT-5.5 released to ChatGPT and Codex</strong></h3><p>OpenAI released GPT-5.5 to ChatGPT Pro and Codex users on 23 April 2026. The GPT-5.5 keeps the 1 M-token context window of GPT-5.4 but is priced at $5 per 1 M input tokens and $30 per 1 M output tokens, roughly double the price of GPT-5.4. Despite the price, GPT-5.5 improves performance on 9 of 10 benchmarks (e.g., ARC-AGI-2, MCP Atlas, Terminal-Bench 2.0) and completes the same Codex tasks with fewer tokens and similar per-token latency.</p><p><em><strong>Read more: <a href="https://openai.com/index/introducing-gpt-5-5/">https://openai.com/index/introducing-gpt-5-5/</a></strong></em></p><div><hr></div><h3><strong>ChatGPT Images 2.0</strong></h3><p>OpenAI released ChatGPT Images 2.0, describing it as a complete rethink of its image&#8209;generation model. The new model expands the product from a fun, creative tool into a visual workflow platform. PetaPixel notes that OpenAI designed Images 2.0 to be available across ChatGPT, Codex, and the API so that developers can incorporate it into code&#8209;generation and research workflows</p><p><em><strong>Read more: <a href="https://openai.com/index/introducing-chatgpt-images-2-0/">https://openai.com/index/introducing-chatgpt-images-2-0/</a></strong></em></p><div><hr></div><h3><strong>Google AI Agent Clinic lessons</strong></h3><p>Google&#8217;s AI Agent Clinic series published a post on 21 April 2026 that shares lessons from refactoring a monolithic sales&#8209;research agent. The article is written for developers building complex agents and recommends:</p><ul><li><p>Break up monolithic agents</p></li><li><p>Force structured outputs</p></li><li><p>Use dynamic Retrieval&#8209;Augmented Generation (RAG)</p></li><li><p>Add observability and cost control</p></li></ul><p>These lessons help developers design more maintainable and cost&#8209;efficient AI agents that can survive in production.</p><p><em><strong>Read more: <a href="https://developers.googleblog.com/production-ready-ai-agents-5-lessons-from-refactoring-a-monolith/">https://developers.googleblog.com/production-ready-ai-agents-5-lessons-from-refactoring-a-monolith/</a></strong></em></p><div><hr></div><h2><strong>Hermes Agent + Ollama: How to Install Hermes Locally Fast</strong></h2><p>Run Hermes Agent locally with Ollama in just 1 minute, no API, no cloud, no cost. In this tutorial, I&#8217;ll show you how to install a fully local AI agent, how it works, and real use cases to boost your workflow.</p><div id="youtube2-aAQWB3-XD5s" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;aAQWB3-XD5s&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><p>Watch on YouTube: <strong><a href="https://youtu.be/aAQWB3-XD5s?si=JgUpYwnmRBOFNYBY">Hermes Agent Locally with Ollama</a></strong></p><div><hr></div><p>See you next week! Cheers, proflead!</p>

AI News for Developers #21: GPT-5.5, Images 2.0 & Real Production Agents

<p>Hey Devs! Another wave of agentic AI innovations rolled in this week.</p><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!vb2R!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!vb2R!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 424w, https://substackcdn.com/image/fetch/$s_!vb2R!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 848w, https://substackcdn.com/image/fetch/$s_!vb2R!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 1272w, https://substackcdn.com/image/fetch/$s_!vb2R!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!vb2R!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png" width="1456" height="880" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:880,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2450184,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://proflead.substack.com/i/194663128?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!vb2R!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 424w, https://substackcdn.com/image/fetch/$s_!vb2R!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 848w, https://substackcdn.com/image/fetch/$s_!vb2R!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 1272w, https://substackcdn.com/image/fetch/$s_!vb2R!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2ad3c846-7b32-43c0-a012-afbab9d9e36a_1614x975.png 1456w" sizes="100vw" fetchpriority="high"></picture></a></figure></div><h3><strong>Claude&#8239;Opus&#8239;4.7 (Anthropic)</strong></h3><p>Anthropic&#8217;s latest flagship model significantly improves coding abilities: SWE&#8209;bench Verified scores jump from 80.8&#8239;% to 87.6&#8239;%, and CursorBench scores rise from 58&#8239;% to 70&#8239;%. The update increases vision resolution to 3.75&#8239;MP and introduces an <strong>&#8220;xhigh&#8221; effort level</strong> plus a new /ultrareview command for deeper analysis. However, it shows a regression on Terminal&#8209;Bench and adds stricter safety filters to block prohibited cybersecurity exploits.</p><p><em><strong>Read More about the model: <a href="https://www.anthropic.com/news/claude-opus-4-7">https://www.anthropic.com/news/claude-opus-4-7</a></strong></em></p><div><hr></div><h3><strong>Grok&#8239;4.3&#8239;Beta (xAI)</strong></h3><p>xAI quietly launched Grok&#8239;4.3&#8239;Beta, a large pre&#8209;trained model matching the scale of Grok&#8239;4.20. It introduces agents that can create presentations, spreadsheets, and documents, moving beyond simple chat to deliver real downloadable files. The beta is available via an early&#8209;access subscription priced around US$300 per month.</p><p><em><strong>Read more: <a href="https://grok.com/release-notes/apr-17-2026">https://grok.com/release-notes/apr-17-2026</a></strong></em></p><div><hr></div><h3><strong>Major OpenAI&#8239;Codex update</strong></h3><p>On April 16, OpenAI rolled out a major update to its Codex assistant. Codex can now operate the user&#8217;s computer, seeing, clicking, and typing on applications, and includes an in&#8209;app browser. It supports multiple terminal tabs and remote dev boxes, integrates more than 90 plugins (e.g., Atlassian Rovo, CircleCI, CodeRabbit), and can generate images via gpt&#8209;image&#8209;1.5. A preview of <strong>memory</strong> lets Codex remember user context and preferences for long&#8209;running tasks.</p><p><em><strong>Read more: <a href="https://openai.com/index/codex-for-almost-everything/">https://openai.com/index/codex-for-almost-everything/</a></strong></em></p><div><hr></div><h3><strong>Agents&#8239;SDK&#8239;0.14.0 (OpenAI)</strong></h3><p>OpenAI&#8217;s April 15 update to the Agents SDK introduces a more capable harness for multi&#8209;step agent loops, native sandbox execution, and separation of the harness from compute for better security and scalability. The SDK allows developers to give agents a controlled workspace to inspect files and run commands safely in sandboxed environments.</p><p><em><strong>Read more: <a href="https://openai.com/index/the-next-evolution-of-the-agents-sdk/">https://openai.com/index/the-next-evolution-of-the-agents-sdk/</a></strong></em></p><div><hr></div><h3><strong>Hermes Agent open&#8209;sourced (Nous Research)</strong></h3><p>Nous Research released Hermes Agent, a self&#8209;improving agent designed to co&#8209;evolve with users. Part of the Hermes model series, it aims to transition from static large language models to &#8220;Large Action Models&#8221; (LAMs) by enabling adaptive, tool&#8209;using agents. The open&#8209;source project emphasizes user adaptability and encourages community contributions.</p><p>Read more: </p><p>https://hermes-agent.nousresearch.com/</p><div><hr></div><h3><strong>GitHub Copilot Agents Explained: How to Build a Custom Agent</strong></h3><p>In this video, I&#8217;ll show you how to design custom AI agents in VS Code or Copilot CLI using GitHub Copilot and explain why GitHub Copilot AI agents for beginners is the most important skill to learn this year.</p><div id="youtube2-PBPypdvlVQo" class="youtube-wrap" data-attrs="{&quot;videoId&quot;:&quot;PBPypdvlVQo&quot;,&quot;startTime&quot;:null,&quot;endTime&quot;:null}" data-component-name="Youtube2ToDOM"><div class="youtube-inner"></div></div><p>Watch on YouTube: <strong><a href="https://youtu.be/PBPypdvlVQo">GitHub Copilot Agents</a></strong></p><div><hr></div><p>See you next week! Cheers, proflead!</p>

AI News, Productivity Tips & Tools for Devs

Latest issues