Edition: May 1, 2026

Replit's CEO Amjad Masad discusses the company's rapid growth, with revenue projected to reach a billion-dollar annual run rate, and his desire to maintain independence despite potential acquisition talks. Masad also criticizes Apple over its App Store battle with Replit, threatening to take the company to court. Replit's unique approach to AI coding assistants has allowed it to achieve positive gross margins, unlike rival Cursor, which is reportedly in talks to be acquired by SpaceX for $60 billion.

A new study found that AI models trained to present a warmer tone to users are more likely to make errors and validate incorrect beliefs, particularly when the user is feeling sad. The researchers from Oxford University's Internet Institute used supervised fine-tuning techniques to modify language models and measure the effect of warm language patterns. This tendency to prioritize user feelings over truthfulness mirrors human communication tendencies.

Meta has acquired humanoid robotics startup Assured Robot Intelligence to enhance its AI capabilities, particularly in developing robots that can understand and adapt to human behaviors. The acquisition is expected to bolster Meta's ambitions in building humanoid robots. The move reflects a broader industry trend of investing in robotics and AI technology.

Elon Musk testified in his trial against OpenAI, claiming he was deceived into funding the company and warning that AI could pose an existential threat to humanity. Musk's xAI company uses OpenAI's models to train its own chatbot, Grok. The trial's outcome could impact OpenAI's potential IPO and valuation.

A dark-money group called Build American AI is secretly paying social media influencers to promote American-made AI and frame Chinese AI as a threat, as part of a coordinated influence campaign. The campaign aims to shift public debate and emphasize the importance of the US leading in AI innovation. Influencers are being offered deals such as $5,000 per TikTok video to amplify the messaging.

Elon Musk's lawsuit against OpenAI has begun, with Musk arguing that the company's conversion to a for-profit model betrays its original nonprofit mission. The case is ongoing, with multiple witnesses and evidence being presented. The lawsuit has significant implications for the tech industry and AI development.

The US Defense Department has signed deals with Nvidia, Microsoft, and AWS to deploy AI technology on its classified networks, aiming to establish the US military as an AI-first fighting force. The agreements will provide warfighters with advanced tools for data synthesis, situational understanding, and decision-making. The move is part of the Pentagon's effort to diversify its AI vendors and ensure long-term flexibility.

The increasing use of AI is expanding the attack surface and adding complexity to cybersecurity, making it essential to rethink security approaches with AI at its core. Tarique Mustafa, a renowned expert in AI-powered cybersecurity, discusses the need for a new security paradigm. He highlights the limitations of legacy approaches and the importance of autonomous AI in protecting against cyber threats.

Forrester predicts that by the end of the decade, CIOs will be forced into a new role as enforcers of order due to the chaotic growth of agentic AI, governing the enterprise AI-powered operating system rather than running the tech. The analyst warns of systematic failure at scale if AI agent systems are not properly controlled. CIOs must adapt to new roles, including architect of enterprise decision-making and governor of autonomous systems.

The Pentagon has formed partnerships with several major AI companies, including OpenAI, Google, and Nvidia, to utilize their AI tools in classified settings. Notably, Anthropic was excluded from these deals due to being deemed a supply-chain risk. These agreements aim to facilitate the lawful use of AI systems within the Defense Department.

Elon Musk's lawsuit against OpenAI has begun, with Musk claiming that CEO Sam Altman betrayed the company's nonprofit mission by converting to a for-profit model. The trial is expected to be contentious, with many witnesses and evidence to come. Musk argues that Altman's actions amount to stealing a charity.

OpenAI is releasing its new GPT-5.5-Cyber model to a limited group of 'cyber defenders', despite previously criticizing Anthropic for a similar approach. The model is designed to identify security flaws and can be used for penetration testing and malware analysis. The restricted release has raised concerns about exclusivity and the potential for misuse.

The Trump administration has fired the entire 22-member National Science Board, which oversees the National Science Foundation, a federal agency that funds major research projects. This move is the latest blow to American science, following a year of grant terminations, job cuts, and funding cuts. The NSF has been without a director since April 2025.

Fujitsu has confirmed it will end its mainframe business by 2035, coinciding with the expected rise of quantum AI supercomputers. The company plans to focus on AI-driven management and will use its own AI technology to enhance decision-making. Fujitsu is also exploring defense projects with Japan, the UK, and Australia.

Shivon Zilis, a longtime employee of Elon Musk and mother of four of his children, played a crucial role as a behind-the-scenes manager of communications and egos in OpenAI's early years, serving as a liaison between OpenAI and Musk. Zilis' involvement in OpenAI's operations and her relationship with Musk have been revealed in court testimonies and evidence. Her role has been described as a covert liaison, with Zilis providing Musk with updates on OpenAI's progress and fundraising efforts.

OpenAI's GPT-5.5 model has achieved new benchmarks in important tasks, but struggles with distinguishing between what it knows and doesn't know, often confidently delivering incorrect output. The model is available in ChatGPT with various subscription tiers and offers features like tool use, web search, and structured outputs. GPT-5.5's performance tops the Artificial Analysis Intelligence Index, but falls behind competitors in subjective evaluations.

The Pentagon's CTO, Emil Michael, has stated that Anthropic is still considered a supply chain risk and is barred from use, despite interest in its Mythos model from government agencies. Agencies are evaluating Mythos for its cybersecurity capabilities, but it is not being deployed operationally. The US government is exploring ways to work with AI companies to understand and mitigate potential risks.

A new US phone network for Christians is set to launch, blocking porn and gender-related content, while a startup has released a tool to debug Large Language Models, and mass firings have dealt a fresh blow to American science. The phone network's filter raises concerns about subjective control over content, and the debugging tool could give users more control over AI model building. These developments highlight ongoing issues in technology and science governance.

The US National Oceanic and Atmospheric Administration is using inexpensive submersibles to map the Pacific seafloor and search for critical mineral deposits. The submersibles, built by Orpheus Ocean, can dive nearly 6,000 meters down and capture cores of sediment, making deep-sea exploration more economical. This technology could aid in deep-sea science and mining, providing new insights into the vastly understudied deep sea.

Companies are taking control of their data to tailor AI for their needs, balancing ownership with the safe flow of high-quality data. A conversation from MIT Technology Review's EmTech AI conference explores how AI factories unlock new levels of scale, sustainability, and governance. This positions data control as a strategic imperative for governments and enterprises.

Elon Musk had a tumultuous week in court as he testified in his case against OpenAI, arguing with lawyers and changing his story. Musk's performance on the stand may hurt his chances of winning the case. The trial continues with Musk's claims that OpenAI 'stole a nonprofit'

Christian content creators are using Fiverr to outsource the creation of AI-generated images and videos, often using the technology to retell biblical stories. These gig workers are embracing AI to meet the demand for cheap and quick content. The resulting clips can be found on various social media platforms.

Microsoft has launched a new AI agent in Word designed for legal teams, which can handle document edits, negotiation history, and complex documents. The Legal Agent follows structured workflows to manage tasks like reviewing contracts. It can work with existing documents and analyze changes.

A new US phone network, Radiant Mobile, is set to launch, targeting Christians and blocking porn and gender-related content by default. The network uses network-level blocking, which cannot be turned off even by adult account owners. This move marks a unique approach to content filtering in the US cell phone market.

The UK's Driver and Vehicle Licensing Agency (DVLA) has introduced new technology to support driving license applications that require medical checks, aiming to clear a 14-week backlog. The agency has implemented a modern casework system and hired more medical caseworkers to reduce processing times. Despite progress, some applicants still face significant delays and difficulties with the system.

ChatGPT Images 2.0 has seen significant adoption in India, with 5 million downloads in its launch week, but its global response has been more measured, with limited overall growth and modest engagement gains. The tool is being used for self-expression, creating studio-style portraits and imaginative visuals. Emerging markets like Pakistan and Indonesia also saw sharp spikes in app downloads.