Code & Clarity

Here’s the ironic twist: companies are hiring armies of specialists just to make AI work inside their own systems. A new company called June thinks it can replace many of those human fixers with, you guessed it, more AI. June scans a business’s existing software, spots the chaos of duplicate data and clashing workflows, then builds a step by step plan to clean it up and drop in AI agents. It even tells teams exactly what to remove or connect before clicking build. A mortgage lender called CMG hit a wall trying to link Claude Code to Salesforce, then used June to map and deploy agents safely before the first official meeting. The team behind June sold their last AI company to Salesforce in 2019 and spent years watching customers wrestle with AI integration. Now they’ve raised 20 million dollars from Benioff’s Time Ventures plus backers like Michael Dell and Aaron Levie, all without a pitch deck. Their bet is that the real barrier to AI isn’t the models but the messy, outdat...

Picture calling customer service and having no idea if the person on the other end is human or AI. That is the goal of Smallest.ai, a new startup that just secured $13 million to build voice AI so natural it passes the Turing test. The company uses small, specialized models designed to mimic human conversation in real time, listening and thinking while you speak, just like we do. Unlike today’s chatbots that pause to process your entire message, this approach cuts lag to near zero. When the model hits a question it can’t answer, it hands you off to a larger AI for a quick research break, just as a human agent would. Big names like RingCentral and Truecaller are already testing it, and the company argues most customer support firms would rather buy this tech than build it themselves. Competitors like ElevenLabs focus on audio dubbing or podcasts, but Smallest.ai is all in on live, conversational voice agents. Smallest.ai was founded in late 2024 by Sudarshan Kamath, who be...

OpenAI just dropped a big update to its ChatGPT desktop app that lets you talk to your computer and get things done. The new voice mode uses OpenAI’s latest ChatGPT-Live models to handle complex, multi-step commands, like asking it to create a new thread, make a pull request, and debug code all in one go. What makes this different from the smartphone version is that it can actually control your computer. It works with ChatGPT Work and Codex, and on macOS, it can even see what’s on your screen to help with tasks. The feature is rolling out globally starting today. The background here is that OpenAI has been pushing its voice capabilities for a while, but until now, the desktop app couldn’t take action—it was more about chatting. Meanwhile, competitors like Anthropic have been updating their own voice modes to work with apps like Gmail and Slack. Why should you care This means you can now use your voice to automate tasks that used to require typing or clicking through menus...

Apple Intelligence, the company’s AI system for iPhones and Macs, just got the green light to launch in China. The approval came after Apple struck a deal with Alibaba to use its Qwen AI model, which will power features like text and image generation in Apple’s software. This is a big move for Apple, as China is a critical market where iPhone sales recently jumped 28 percent. Before settling on Alibaba, Apple reportedly considered partnerships with Baidu, ByteDance, and DeepSeek, but those didn’t work out. The delay meant Chinese users missed out on Apple Intelligence features that rolled out elsewhere in 2024. Now, with Alibaba’s tech on board, Apple can finally offer its AI tools in China, though no exact launch date has been given yet. Alibaba’s stock rose over 6 percent on the news, showing how much this deal matters for both companies. For Apple, it’s about keeping its footing in a competitive market where local brands dominate. For Alibaba, it’s a chance to showcase ...

Siri just got a major upgrade and now anyone can try it. Apple released the iOS 27 public beta, giving everyday users early access to its AI-powered assistant. This is the first time Apple has made the new Siri available beyond developers, marking its biggest test yet on a massive scale. The updated Siri can now dig into your emails, photos, and messages to answer questions or perform tasks. It can summarize group texts, add events from messages to your calendar, or even identify what’s in your camera view. It also understands context better, like answering questions about local events or news. You can trigger it the usual ways, plus new options like swiping down from the Dynamic Island or using Spotlight search. Under the hood, Siri AI runs on Apple’s own foundation models, built with Google’s help but tailored for Apple devices. These models process data locally or via Private Cloud Compute, ensuring your personal info stays private. The assistant is also more integrated...

The New York Times and The Daily News say OpenAI has been dishonest about its ability to track how much of their journalism appears in ChatGPT. Court documents reveal OpenAI had tools to search its training data and even built a system to detect when ChatGPT copied content word for word. This contradicts OpenAI’s earlier claims that such searches were impossible or too difficult. An OpenAI engineer admitted in a deposition that the company had a database of 78 million de-identified ChatGPT conversations and a tool called Project Giraffe that flagged regurgitated text. The news outlets argue OpenAI then provided heavily redacted chat logs to the court, making them useless for proving copyright violations. They also accuse OpenAI of deleting billions of chat records after the lawsuit began, which would violate court orders to preserve evidence. The New York Times and The Daily News are now asking the judge to sanction OpenAI. They want the company barred from using its provi...

Claude just got a feature that quietly makes a case for itself. Reflect is a dashboard that tracks how you use the AI, showing topics you discuss, usage patterns, and tasks you hand off to it. It’s not just about numbers, though. The real goal seems to be making you realize how embedded Claude has become in your daily work. The tool also nudges you to think critically about your AI habits. It might ask what tasks you’d rather keep doing yourself, even if Claude could do them faster. There are even options to set quiet hours or reminders to take breaks, a nod to how easy it is to get sucked into endless AI chats. Reflect doesn’t stop at showing your usage. It also suggests ways to use Claude more efficiently, like using its Projects feature to avoid re-explaining context. This keeps you tied closer to Claude’s ecosystem, making it harder to switch to another AI tool. Anthropic says sensitive conversations are only shown at a high level, and health-related chats are left out...

Etched, a two-year-old company taking on Nvidia, revealed it has $1 billion in contracts for its AI systems, all before finishing tests on its first chip. The startup also shared it raised $500 million late last year at a $5 billion valuation, bringing its total funding to $800 million. Its backers include major trading firms like Jane Street and Hudson River Trading, plus AI pioneers like Geoffrey Hinton and Andrej Karpathy. The company’s product is a full system, not just a chip. It combines custom-designed hardware and software to speed up AI inference, the process where models generate responses to user prompts. This is currently the most expensive and slowest part of running AI at scale, which is why investors are betting big on anyone who can improve it. Etched’s rapid rise is a sharp turn from its early days. In 2023, founders Gavin Uberti and Robert Wachen struggled to get investors interested, even with a detailed pitch arguing specialized chips would be essential...

Micron, the Idaho-based chip maker many only know for those tiny memory cards in old cameras, briefly became more valuable than Meta and Tesla last week. Its stock shot up over 236 percent in a month, closing Friday at $1,132 a share, pushing its market cap to $1.27 trillion. The reason is simple: AI needs memory, and Micron makes the kind of high-performance chips that AI servers crave. The AI boom has created a shortage of DRAM and NAND memory chips, especially the high-bandwidth variety. Companies like Nvidia, Microsoft, and Meta are snapping up every chip they can find to power their AI systems. This scarcity has driven up prices and left other industries, from PC makers to console manufacturers, scrambling for leftovers. Analysts call it RAMageddon, and they expect it to last into 2027. Micron’s latest earnings reflect this surge. Revenue hit $41.45 billion for the quarter, quadruple the same period last year, while profits jumped from $1.88 billion to $28.2 billion. ...

OpenAI’s next big AI model, GPT 5.6, won’t be released to the public right away. Instead, the company will share it only with a small group of trusted partners, and only after the Trump administration gives the green light for each one. According to reports, government officials will approve access on a case-by-case basis during a preview period. This isn’t just OpenAI being cautious. The White House reportedly pressured the company to slow down, with agencies like the Office of the National Cyber Director and the Office of Science and Technology Policy pushing for a limited rollout. If all goes well, OpenAI hopes to open access to everyone in a couple of weeks. The move mirrors what Anthropic did earlier this year with its Claude Mythos model, which it restricted to a select group of partners under its Project Glasswing program. Anthropic argued that the model was too powerful to release widely, raising questions about whether this was a genuine safety measure or a clever...

Imagine walking into a tech office and seeing someone intensely playing Fortnite for hours on end. Now imagine that "someone" is actually an advanced artificial intelligence. That is exactly what visitors experienced at General Intuition, a startup that just secured a massive $320 million investment, boosting its valuation to $2.3 billion. The real surprise is what happens next. The same AI brain powering the virtual Fortnite player is also driving a large, four-legged robot roaming the office. This robot, relying only on a single camera for sight, explores its surroundings much like a curious toddler, occasionally bumping into chairs or trash cans as it learns its physical space. This quick adaptation is remarkable, especially since the AI model only needed eight minutes of outdoor real-world data to fine-tune its navigation skills for indoor spaces. General Intuition’s core idea is that AI can learn how to move and interact with the world by studying human game...

The narrative has been clear for years: AI is coming for engineering jobs. Companies have blamed AI for mass layoffs, with tech cutting more positions in May than any month in recent history. Yet fresh data shows engineering roles are holding steady while other jobs disappear. Research from SignalFire found that while overall hiring at big tech firms fell 25 percent compared to 2019, engineering jobs only dropped 11 percent. Even more striking, engineers made up 55 percent of all new hires in 2025 at companies like Google, Meta, and Amazon, up from 46 percent in 2019. Early-stage startups actually hired 7 percent more engineers than they did five years ago. SignalFire’s head of research, Asher Bantock, puts it simply: if AI were replacing engineers, hiring would be the first to drop. Instead, demand for engineering talent is growing faster than most other roles. Even Anthropic’s own economists admit they haven’t seen AI meaningfully change unemployment rates for jobs like ...

OpenAI revealed a new custom chip called Jalapeño, designed with Broadcom to handle AI tasks more efficiently. Early tests show it uses less power while delivering better performance than current options, which is a big deal for running AI models in real time. What stands out is that OpenAI used its own AI models to help design the chip, showing how deeply AI is now embedded in its own development. This chip is built specifically for inference, the process where AI models respond to user requests. OpenAI hints that while this won’t replace Nvidia’s hardware for heavy tasks like training models, it could significantly cut costs for everyday AI interactions. That’s important because inference is where most of the day-to-day AI work happens, from chatbots to coding assistants. The move follows a trend set by Google and Amazon, both of which have created their own AI-focused chips to reduce reliance on outside suppliers. OpenAI’s partnership with Broadcom was announced last Oc...

Superhuman just bought GPTZero, the startup that built a name for itself by spotting AI-generated text. The deal terms are secret, but GPTZero’s founder says the company had 19 million users and was pulling in $30 million a year. Not bad for a tool that started as a Princeton thesis project three years ago. This isn’t just a simple acquisition. Superhuman, the company formed when Grammarly took over the email app Superhuman and rebranded, already had its own AI detection feature. Their pitch is straightforward: two detectors are better than one. Grammarly’s version helps users tweak their writing so it doesn’t sound like a bot wrote it. GPTZero, on the other hand, focuses on catching AI slop in the first place. GPTZero was bootstrapped at first, then raised $13.5 million from investors like Uncork Capital and Footwork. It grew fast, turning profitable before most startups even think about it. Now, it’s joining a bigger player with deep pockets and a clear interest in makin...

Boris Cherny, the mind behind Claude Code, made a bold claim at Meta’s Scale conference last week. When asked if AI loops are just hype, he said they’re the real deal. Two years ago, humans wrote code by hand. Then AI started writing it for us. Now, we’re at the point where AI agents prompt other AI agents to do the writing. Cherny shared how he uses loops in his own work. One agent constantly looks for ways to improve code architecture, while another hunts for duplicate code that can be streamlined. These agents submit changes like human coders and keep running as the code evolves. It’s a big leap from static prompts to a system where AI works continuously in the background. The concept isn’t entirely new. Programmers have long used recursive loops where a function calls itself until a task is done. But AI loops are different. They rely on sub-agents to decide when to stop, not a fixed rule. Some loops, like the Ralph Loop, simply ask the AI if it’s finished its task, bou...

Imagine ending a phone call and having a summary of the conversation pop up on your screen. That’s what Reliance’s Jio is promising with its new AI assistant, which can join calls to transcribe, summarize, and even book a cab for you mid-chat. Just say “Hey Jio” and the service kicks in, built directly into the telecom network so it feels like a natural part of the call. It’s launching later this year for Jio’s 500 million users. But calls are just the start. Reliance also unveiled an AI-powered version of its MyJio app that handles tasks like activating eSIMs or picking roaming plans through simple voice requests. Then there’s TeleFrame, a home display that uses AI to proactively serve up weather alerts, schedules, and reminders. It’s a clear push into the ambient AI space, where assistants blend into daily life, much like what Amazon and Google are exploring. At its annual shareholder meeting, Reliance framed this as India’s chance to stop being just a consumer of foreig...

Baseten, a startup helping companies run AI models efficiently, is about to secure a massive $1.5 billion investment at a $13 billion valuation. This comes only five months after it raised $300 million at a $5 billion valuation. The jump in valuation is staggering, but there’s a twist: not all investors are paying the same price. Some are buying in at $13 billion, while others are getting a slightly lower $11 billion rate. This split-pricing trick lets startups inflate their headline valuation while still giving certain investors a better deal. The company launched in 2019 and specializes in AI inference, which is what happens when you ask an AI a question and it generates an answer. Baseten’s pitch is simple: it routes requests to the best model for the job, often using cheaper open-source options instead of pricier ones from big names like OpenAI. This approach has caught the eye of investors, who are pouring money into the so-called inference gold rush. Baseten is part ...

Adobe just pushed its Firefly AI assistant into three of its biggest creative apps. In Premiere, it can sort video clips into folders, rename batches of files, and even find interview questions inside footage. In Illustrator, it can rearrange layers and spot missing fonts. And in InDesign, it can pull together brand kits with logos and color palettes just from a description or uploaded sample. The assistant already lived in Express, Photoshop, and Acrobat, and Adobe says it will soon talk to Google Gemini and Slack as well. Firefly itself is also getting two new features: Elements lets you save AI-generated characters or objects to reuse in other projects, and Projects keeps all assets for a campaign or series in one place for easy sharing. Adobe has been racing to bake AI into every corner of Creative Cloud. Firefly started as an image generator and has steadily added video, storyboards, and now brand kits that can churn out product videos from a handful of photos. This ...

At a G7 lunch with AI CEOs and President Trump, French President Emmanuel Macron laid out the fear now keeping leaders up at night. If the US can block access to models like Anthropic’s Mythos 5 and Fable 5 overnight, entire economies built on American AI could freeze. That’s not hypothetical. Days earlier, the White House did exactly that, citing national security after Amazon flagged a possible safety gap. India’s Prime Minister Narendra Modi echoed the alarm. His argument is that democratic nations need steady access to the best AI to protect their own infrastructure. The worry is simple: any company or country relying on US tech could lose it without warning or explanation. The move has exposed a raw nerve. Cybersecurity experts point out that the same capabilities exist in other available models, and yet Anthropic’s remain locked. Meanwhile, firms like Canada’s Cohere are using the moment to push for digital sovereignty, arguing that dependence on a handful of US comp...

You can now order food and groceries from DoorDash using just your words or photos, thanks to the company's new AI chatbot. This feature, called Ask DoorDash, allows users to search for what they want in their own words, rather than scrolling through restaurants and stores to build a cart. For example, you can tell the chatbot that you want a filling dinner for a family of four, and it will surface restaurants that match your request. The chatbot can also help with grocery shopping by building a cart based on a photo from a cookbook or a picture of your grocery list. It will even prompt you to check if you already have staples like sugar and butter, so you don't buy something you already have. Additionally, you can ask the chatbot to reorder your last grocery cart or suggest new items based on your previous orders. This feature is rolling out on iOS in select regions and will reach more users across the US in the coming weeks. The launch of Ask DoorDash is part of ...