OpenAI has rolled out GPT-5-Codex, the latest upgrade to its AI coding assistant. Unlike previous models, GPT-5-Codex can decide how long to “think” about a problem, ranging from just a few seconds to as much as seven hours. This flexibility allows the model to perform better on agentic coding benchmarks and large-scale code refactoring tasks, where deeper analysis often pays off. Where You Can Use It The new model is now live across multiple Codex-enabled products, including: Currently, it’s available to ChatGPT Plus, Pro, Business, Edu, and Enterprise users. OpenAI says API access will follow soon. Standing Out in a Crowded Market The AI coding space is more competitive than ever, with tools like Claude Code, Anysphere’s Cursor, and GitHub Copilot making major strides. Cursor even surpassed $500 million in ARR earlier this year, while Windsurf found itself at the center of a high-profile acquisition drama. By giving Codex a dynamic edge, OpenAI is positioning GPT-5-Codex as a strong alternative in this fast-growing market. Better at Code Reviews Beyond writing code, GPT-5-Codex was trained to review code more effectively. Experienced software engineers who tested it said the model offered: This makes GPT-5-Codex not only a coding assistant, but also a practical partner in software engineering workflows. The Secret Sauce: Dynamic Problem-Solving According to Codex product lead Alexander Embiricos, the big leap comes from GPT-5-Codex’s ability to adjust in real time. While GPT-5 uses a “router” to decide upfront how much compute to allocate, GPT-5-Codex can reassess mid-task. That means if it realizes a problem is tougher than expected, it can extend its working time from minutes to hours. In some cases, it has taken over seven hours to solve a single challenge. Why It Matters By combining adaptive reasoning, stronger benchmarks, and better review skills, GPT-5-Codex shows how AI coding tools are evolving from quick helpers into deep problem-solvers. For developers, this could mean cleaner code, faster debugging, and more reliable collaboration between human engineers and AI. Sources (Techcrunch )
OpenAI’s AGI Empire and Its Costs
At the heart of every empire lies an ideology, a belief system powerful enough to justify expansion, no matter the contradictions. For European colonial powers, it was Christianity paired with resource extraction. For today’s AI industry, it’s the promise of artificial general intelligence (AGI) to “benefit all humanity.” And according to journalist and bestselling author Karen Hao, OpenAI has become its chief evangelist. In her new book, Empire of AI, Hao argues that the only way to grasp the scope of OpenAI’s influence is to view it as an empire, one with extraordinary economic and political power. “They’re terraforming the Earth,” she told TechCrunch’s Equity. “They’re rewiring geopolitics and all of our lives.” The Ideology of AGI OpenAI describes AGI as a highly autonomous system that can outperform humans at most economically valuable tasks. In theory, AGI could elevate humanity by increasing abundance, accelerating scientific discovery, and driving progress. But these promises remain abstract, while the costs are tangible: massive energy demands, oceans of scraped data, escalating compute requirements, and untested systems released into the world. Hao argues that speed, not safety or efficiency, became OpenAI’s guiding principle. Rather than pursuing novel algorithms or efficiency gains, the company opted for the “cheap” path: scaling existing methods with ever more data and compute. This framing, where “the victor takes all,” forced rivals like Google and Meta to follow suit. The result: AI research increasingly shaped by corporate agendas, not independent science. The Cost of Scale The financial stakes are staggering. OpenAI projects a cash burn of $115 billion by 2029. Meta is set to spend up to $72 billion on AI infrastructure in 2025. Google expects $85 billion in capital expenditures the same year. Meanwhile, the promised societal benefits lag behind. Instead, harms are mounting, from job losses and wealth concentration to the psychological fallout of AI chatbots. Behind the scenes, low-wage workers in places like Kenya and Venezuela handle disturbing content for just $1–$2 an hour, moderating the toxic data fueling these models. Hao stresses that progress doesn’t need to come at this cost. Google DeepMind’s AlphaFold, for example, used focused datasets to solve protein folding, a breakthrough with profound implications for medicine and disease research. Unlike AGI-scale models, AlphaFold avoided mass exploitation and environmental tolls. “These are the types of AI systems we actually need,” Hao said. A Race That Reshaped the World Alongside the AGI mission, the industry has leaned heavily on the “race with China” narrative. But instead of liberalizing the world, Hao argues, Silicon Valley’s approach has done the opposite, concentrating power and leaving the U.S.–China gap largely unchanged. The one clear winner? Silicon Valley itself. The Profit–Mission Dilemma OpenAI’s unusual structure, part non-profit, part for-profit, has blurred how the company measures its impact on humanity. With its deep ties to Microsoft and recent moves edging toward a public offering, critics worry the lab has conflated financial success with human benefit. Two former safety researchers told TechCrunch they fear OpenAI now sees consumer enjoyment of ChatGPT as proof it’s fulfilling its mission. Hao sees this as dangerous, a belief system so consuming it blinds leaders to evidence of harm. “Even as the evidence accumulates that what they’re building is actually harming significant amounts of people,” Hao said, “the mission continues to paper all of that over.” The Bottom Line AI’s current trajectory wasn’t inevitable. It was shaped by choices: speed over safety, scale over science, ideology over evidence. As the industry barrels ahead, Hao’s warning is clear, unchecked belief in AGI risks building an empire whose costs outweigh its promises. Sources (Techcrunch)
OpenAI Plans Shift to Public Benefit Corporation
OpenAI has reached a nonbinding agreement with Microsoft that could reshape its corporate structure and pave the way for a potential public listing. The deal, announced Thursday, would allow the AI company to transition its for-profit arm into a Public Benefit Corporation (PBC) while still maintaining nonprofit oversight. What the agreement means Under the proposed structure, OpenAI’s nonprofit parent would continue to hold control of operations and secure a stake in the new PBC valued at over $100 billion, according to board chairman Bret Taylor. Microsoft OpenAI’s largest investor signed a memorandum of understanding (MOU) with the company to outline the terms, though a definitive agreement has yet to be finalized. The transition must still win approval from regulators in California and Delaware, where OpenAI is incorporated. Why the shift matters OpenAI’s unusual governance model has long distinguished it from Silicon Valley peers. The nonprofit board famously ousted CEO Sam Altman in 2023 before reinstating him days later. With a PBC structure, OpenAI could raise fresh capital while preserving its stated mission of building AI that benefits humanity. For Microsoft, the new agreement ensures continued access to OpenAI’s technology and keeps its role as the primary cloud provider — though OpenAI has been diversifying. The company recently committed to a $300 billion cloud contract with Oracle starting in 2027 and is collaborating with SoftBank on its massive Stargate data center project. ALSO SEE: California Bill Targets AI Chatbot Safety Tensions and context The negotiations between Microsoft and OpenAI have not been without friction. Disputes reportedly surfaced over control of Windsurf, an AI coding startup OpenAI had aimed to acquire. That deal collapsed, with Windsurf’s talent scattering to Google and Cognition. Meanwhile, Elon Musk who co-founded OpenAI before parting ways has been a vocal critic. His ongoing lawsuit accuses the company of straying from its nonprofit mission. Earlier this year, Musk even floated a $97 billion takeover bid, which OpenAI’s board rejected. Ironically, the nonprofit’s new stake in the PBC under this agreement is larger than what Musk had offered. The road ahead Nonprofit groups like Encode and The Midas Project have raised alarms over OpenAI’s for-profit transition, arguing it undermines the company’s original mission. OpenAI has pushed back, even issuing subpoenas, suggesting some of these groups may have ties to rivals such as Musk or Meta CEO Mark Zuckerberg allegations the nonprofits deny. For now, the future hinges on regulatory review. If approved, the PBC move could set the stage for OpenAI’s eventual debut as a public company potentially one of the most valuable in tech history Sources ( Techcrunch )
California Bill Targets AI Chatbot Safety
California is on the verge of making history with SB 243, a new bill designed to regulate AI companion chatbots and protect vulnerable users. On Wednesday night, the California State Assembly passed the legislation with bipartisan support, sending it to the state Senate for a final vote this Friday. If approved and signed by Governor Gavin Newsom, the law would take effect January 1, 2026, making California the first state to impose direct safety rules on AI companions. What the Bill Covers SB 243 targets AI companion chatbots systems that simulate human-like conversation and build social connections with users. Under the bill: Why Now? The legislation gained urgency after the tragic death of teenager Adam Raine, who reportedly discussed suicide and self-harm with ChatGPT before taking his life. The bill also follows reports that Meta’s internal chatbots engaged in “romantic” and “sensual” conversations with minors, sparking national concern. With rising scrutiny from regulators, the Federal Trade Commission is preparing investigations into how chatbots affect children’s mental health, while Texas and federal lawmakers have launched probes into major AI players like Meta and Character.AI. What Changed Along the Way Interestingly, SB 243 was initially far stricter. Early drafts would have banned manipulative engagement tactics, such as “variable rewards” (used by platforms like Replika to keep users hooked with rare responses, unlockable personalities, or special storylines). That requirement was later dropped, along with provisions that would have forced companies to log how often their bots brought up suicidal themes. Co-sponsor Sen. Josh Becker said the final bill “strikes the right balance,” imposing safeguards without creating compliance hurdles that are “technically not feasible or just a lot of paperwork for nothing.” A Bigger Fight Over AI Rules SB 243 isn’t the only AI bill in California. Another proposal, SB 53, would require broader transparency reporting from AI developers. Tech giants including OpenAI, Meta, Google, and Amazon, oppose it, while Anthropic is the lone major company in support. Despite Silicon Valley pouring millions into pro-AI political action committees ahead of the midterms, state senator Steve Padilla insists innovation and regulation don’t have to be at odds. “We can support innovation … and at the same time, provide reasonable safeguards for the most vulnerable people,” Padilla said. The state Senate votes Friday. If SB 243 passes and Governor Newsom signs it, companies will need to comply starting January 1, 2026, with reporting requirements kicking in by July 1, 2027. If approved, this law would set a first-of-its-kind precedent in the U.S., likely influencing how other states and even federal regulators approach AI companions moving forward. Sources ( Techcrunch )
Apple Intelligence: Full Guide 2025
If you’ve recently upgraded to a new iPhone, you may have already spotted Apple Intelligence baked into apps like Messages, Mail, and Notes. Apple officially rolled out its AI platform in October 2024, and it’s clear the company is positioning it as a long-term player alongside Google Gemini, OpenAI, and Anthropic. So, what exactly is Apple Intelligence and how does it work? Let’s break it down. What Is Apple Intelligence? Apple markets its new AI system as “AI for the rest of us.” It’s designed to enhance everyday tasks using generative AI, from text assistance to image creation. Timeline: How Apple Intelligence Rolled Out ALSO SEE: Made by Google 2025: Pixel 10 & More Who Can Use Apple Intelligence? Apple’s AI features are free, but only run on newer hardware. Supported devices include: Notably, only the Pro versions of iPhone 15 qualify due to chipset limitations. On-Device vs Cloud: How Apple’s AI Works Unlike GPT or Gemini, Apple Intelligence is built to work on-device whenever possible. That means tasks like email composition don’t require an internet connection, making them faster and more private. For heavier queries, Apple uses Private Cloud Compute Apple-run servers built on Apple Silicon. The switch between local and cloud processing is seamless, and users only notice if they’re offline (in which case some tasks won’t work). ChatGPT + Apple Intelligence Apple’s much-talked-about partnership with OpenAI isn’t about replacing Apple’s AI — it’s about filling gaps. Paid ChatGPT users will get access to premium features through this integration. Apple has hinted that Google Gemini may be the next third-party partner. Developers + Apple Intelligence At WWDC 2025, Apple announced its Foundation Models framework, giving developers access to Apple’s on-device AI. This means third-party apps can integrate AI features without relying on expensive cloud APIs. For example: All while keeping your data private and offline. What’s Next for Siri? Apple is expected to debut a major Siri overhaul in 2026. Rumors suggest Apple might even partner with competitors like Google to speed up development. Until then, Siri remains more capable than ever — but still behind leaders like ChatGPT and Gemini in some areas. Apple Intelligence is still in its early stages, but it’s already transforming how iPhone, iPad, and Mac users interact with their devices. With on-device processing, tighter Siri integration, and third-party AI partnerships, Apple is carving its own path in the AI race with privacy and seamless user experience at the forefront. Sources ( Techcrunch )
Mistral AI: Europe’s OpenAI Rival
Mistral AI, the French startup behind the AI assistant Le Chat and several powerful AI models, has quickly become one of Europe’s most promising tech companies. Valued at a staggering $14 billion (up from $6B in June 2024), it’s now seen as the only European contender capable of standing up to OpenAI. Even French President Emmanuel Macron has voiced support, urging citizens to “download Le Chat” instead of sticking to U.S.-based alternatives. What is Mistral AI? Founded in 2023, Mistral AI has a bold mission: “to put frontier AI in the hands of everyone.” Unlike many competitors, it leans heavily into open-source models signaling a more transparent approach to AI development. Its flagship app, Le Chat, is already gaining traction, topping France’s iOS App Store after launch and surpassing 1 million downloads in just two weeks. By mid-2025, it added major upgrades like: Mistral’s Growing Suite of Models Mistral has built an impressive lineup of AI models across different domains: In June 2025, Mistral also launched Mistral Code, a coding client to compete with GitHub Copilot, Cursor, and Windsurf. Who’s Behind Mistral AI? The company was founded by three AI veterans: They’re supported by advisers such as Cédric O (former French digital minister), which has stirred some political controversy. Are Mistral’s Models Open Source? Not all. The company keeps its premier models closed, monetized through APIs, while providing others (like Devstral and Mistral NeMo) freely under Apache 2.0. Business Model: How Mistral Makes Money Mistral AI uses a mix of consumer and enterprise revenue streams: Despite rapid growth, revenues are reportedly still in the eight-digit range. ALSO SEE: Lovable CEO on AI Vibe Coding, Unicorn Growth & Future Strategic Partnerships Mistral has already secured high-profile partnerships: Enterprise Features Mistral is steadily building out enterprise-focused tools: Funding Journey Mistral has raised over €1 billion (~$1.04B) since its founding: Regulatory Stance CEO Arthur Mensch has publicly opposed the EU’s fast-tracked AI Act, urging regulators to “pause for two years” before enforcing obligations. Brussels, however, remains firm on its timeline. What’s Next? IPO or Acquisition? Mensch has been clear: Mistral is not for sale. The plan is an IPO, though acquisition rumors (including Apple) continue to circulate. To silence speculation, Mistral will need to prove it can grow revenues at the same pace as its valuation. At Techxnow, we’ll be following closely as this French powerhouse continues to redefine the AI landscape. SOURCES ( Techcrunch )
AI Companion Dot Shuts Down Oct 5
The AI companion app Dot, once pitched as a “friend and confidante,” is officially shutting down. Its parent company, New Computer, announced that the app will remain available until October 5, 2025, giving users time to say goodbye and download their data. This marks the end of one of the more ambitious attempts at building a personalized AI friend—an area of technology that has attracted both curiosity and controversy. A Different Kind of Chatbot When Dot launched in 2024, co-founders Sam Whitmore and Jason Yuan (a former Apple designer) weren’t trying to build just another chatbot. Their vision was bold: create an AI that could feel like a real companion. Instead of offering quick answers like traditional assistants, Dot was designed to: Yuan once described the app as “a living mirror of myself”—a piece of software that could facilitate conversations with one’s “inner self.” For many users, this went beyond simple utility. Dot was marketed as a confidante, a digital friend who could be there when humans weren’t. Why Dot Is Shutting Down Interestingly, the shutdown isn’t being attributed to external controversy. Instead, the founders explained that their “Northstar” visions had diverged. In a farewell note, they wrote: “Rather than compromise either vision, we’ve decided to go our separate ways and wind down operations.” The message acknowledged the emotional weight of this closure. Losing Dot isn’t just like losing a productivity app or a to-do list tool. It’s losing a digital relationship—something far more personal. “We want to be sensitive to the fact that this means many of you will lose access to a friend, confidante, and companion, which is somewhat unprecedented in software.” The Wider Challenge of AI Companions Dot’s closure also highlights the larger debate around AI companions. The idea of digital friends is powerful, but it comes with risks. For startups, these risks make the AI companion space harder to navigate. Building emotional trust with users is powerful—but it can also be dangerous without guardrails. Did Dot Really Have “Hundreds of Thousands” of Users? In its farewell note, the company claimed that Dot had attracted “hundreds of thousands” of users. However, third-party analytics suggest otherwise. According to data from Appfigures: The gap between self-reported engagement and analytics paints a picture of a startup that may have been struggling to scale. What Happens Next for Users If you’ve been using Dot, you still have a brief window to save your data. Here’s how: After that date, the app will shut down permanently, and users will lose access to their conversations. The closure of Dot underscores a bigger truth: AI companions sit in a delicate space between innovation and risk. They’re not just tools they become part of people’s emotional lives. That makes them harder to build, scale, and sustain responsibly. For users, Dot’s disappearance may feel more personal than other app shutdowns. It’s not just about losing software it’s about saying goodbye to a “friend.” The story of Dot serves as both a warning and a lesson. Startups in the AI companion space will need to carefully balance personalization with safety, ethics, and long-term sustainability if they want to succeed where others have stumbled. SOURCES: ( Techcrunch )
Apple iPhone 17 Event: How to Watch
Apple is set to host its much-anticipated “Awe Dropping” hardware event on Tuesday at 10 a.m. PT. The event will be livestreamed directly from Apple Park in Cupertino, and fans can tune in on Apple’s official website. As always, media outlets including Techxnow, will bring you the latest updates and insights as announcements unfold. iPhone 17: What to Expect The spotlight will be on Apple’s upcoming iPhone 17 lineup, with plenty of buzz around a brand-new model: iPhone 17 Air: Rumored to replace the “Plus” model, this version is expected to be the thinnest iPhone ever. Bigger and Smoother Display: The standard iPhone 17 could feature a 6.3-inch screen and a 120Hz refresh rate, doubling the smoothness of its predecessor. Sharper Selfies: Apple may bump the front camera up to 24 megapixels, a major upgrade for creators and casual users alike. Redesigning the Pro The iPhone 17 Pro might see its biggest design change in years: Instead of the familiar square-shaped camera bump, rumors suggest a rectangular camera bar stretching across the back. The bar could house the main lenses, with the flash and LiDAR sensor positioned on the right side, giving the Pro a sleek, new look. ALSO SEE: OpenAI Safety Update: Lessons for Startup Founders More Than Just iPhones The event will also feature announcements beyond the iPhone lineup. Apple is expected to reveal updates to the Apple Watch, improvements for AirPods, and more details about iOS 26. The new iOS introduces the Liquid Glass interface, which gives iPhones a more transparent, fluid design. From groundbreaking design changes to performance upgrades, this Apple event is shaping up to be one of the biggest in recent years. Techxnow will bring you all the key highlights, hands-on insights, and deep dives as the announcements roll in. Sources ( Techcrunch )
LayerX Raises $100M to Automate Japan’s Back Office
Japan’s enterprises are under pressure from aging demographics, labor shortages, and stricter compliance rules like the 2023 e-invoicing mandate. At the same time, companies are racing to adopt generative AI. Yet, only 16% of digital transformation projects succeed, and the rate is even lower just 4 -11% in traditional industries. The reasons? Rigid corporate culture, weak leadership commitment, and a shortage of digital talent. This is where LayerX steps in. The $100M Series B Milestone LayerX has secured a $100 million Series B round, led by U.S.-based TCV its first investment in a Japanese startup. Other investors include: The new funding brings LayerX’s total raised to $192.2 million. While the valuation wasn’t disclosed, the company said both the round size and valuation are among the largest ever for a Japanese startup at the Series B stage. Inside LayerX’s Product Suite LayerX delivers an AI SaaS platform designed to cut the workload of finance, tax, procurement, and HR teams. Its portfolio includes: The Founder’s Vision LayerX was founded in 2018 by Yoshinori Fukushima, a University of Tokyo-trained machine learning researcher and serial entrepreneur. He previously launched Gunosy, a news app that later went public on the Tokyo Stock Exchange. Fukushima shifted LayerX’s focus after identifying a bottleneck in Japanese enterprise workflows: paper-based invoice processing. This led to the creation of Bakuraku, an AI-native platform designed to replace manual, Excel-heavy processes. Competition and Differentiation In Japan, LayerX competes with Money Forward Cloud Keihi, freee, and Rakuraku Seisan. Globally, its rivals include SAP Concur, Rippling, Brex, Ramp, Spendesk, and Airbase. In AI-powered workflow automation, it also faces Harvey. LayerX sets itself apart with: The Bakuraku Suite now covers everything from expense management and corporate cards to compliance, attendance, and receivables all in one platform. Growth at Breakneck Speed Since its Series A in November 2023, LayerX has posted rapid growth: The Road Ahead Looking forward, LayerX has set bold targets: Its client list already includes Mitsui & Co., MUFG Bank, Ippudo, IRIS Ohyama, Imperial Hotel, and Sekisui Chemical. LayerX’s trajectory shows how AI-first platforms can finally crack Japan’s stubborn digital transformation challenges. With fresh funding, rapid adoption, and ambitious targets, LayerX is positioning itself not just as a back-office solution but as one of Japan’s next SaaS powerhouses. Sources ( Techcrunch )
OpenAI Safety Update: Lessons for Startup Founders
1. The Wake-Up Call Recent tragedies linked to unsafe chatbot interactions have put AI safety back under the spotlight. The cases of Adam Raine, a teenager who used ChatGPT to discuss self-harm, and Stein-Erik Soelberg, who spiraled into paranoia fueled by the model, highlight how powerful AI systems can go dangerously off track when lacking proper guardrails. For startups building with AI, these incidents are not just news headlines—they are reminders of the real-world impact of design choices. ALSO SEE: OpenAI & Anthropic Unite on AI Safety Testing 2. OpenAI’s New Guardrails In response, OpenAI announced several new safety measures: 3. Industry Pushback Despite these measures, critics remain skeptical.Jay Edelson, counsel in the Raine family’s lawsuit against OpenAI, argues that the company’s response is “inadequate,” suggesting safety concerns should have been addressed long before launch. This raises a critical question for founders: Is safety a reactive patch—or a proactive strategy? 4. Lessons for Startup Founders Whether you’re building in AI or any other sector, there are clear takeaways: ALSO SEE: Google to Verify All Android App Developers 5. The Bigger Picture OpenAI’s new 120-day initiative includes partnerships with experts on adolescent health, substance use, and eating disorders. This signals a shift toward cross-disciplinary collaboration, something all tech builders can learn from. Safety is no longer just a compliance checkbox—it’s becoming a competitive advantage. Founders who integrate safeguards into their products from day one will not only avoid crisis headlines but also build companies that earn long-term trust. At TechXnow, we believe in building what lasts, not just what’s next. OpenAI’s latest moves remind us: when human lives intersect with technology, founders carry responsibility that goes far beyond code. Sources ( Techcrunch )
