Welcome back to Tech Pulse β the daily, no-fluff rundown of what actually happened in tech and why it matters. Today (June 3, 2026): the big players keep gunning for each other in AI coding, Nvidia wants to put an "AI agent" inside every PC, Washington wants a look under the hood before models ship, and physicists finally caught a ghost predicted back in 1939.
π€ AI models & tools
Microsoft used its Build conference to stop leaning so hard on OpenAI. Microsoft unveiled its own in-house models: MAI-Code-1-Flash, which turns a written description into source code for apps and websites, and MAI-Thinking-1, a reasoning model β both pitched on efficiency and lower cost for developers (CNBC). Why it matters: Microsoft has ridden OpenAI for years; building cheaper first-party models is a hedge β and a price-pressure play for everyone who builds on its cloud. The fun part: the company whose empire ran on "developers, developers, developers" now wants the model itself to be the developer.
The AI coding wars are officially a four-way brawl. Microsoft's move lands as Google and Microsoft both chase Anthropic and OpenAI in coding models β the most lucrative, stickiest corner of the AI market right now (CNBC). If you're trying to make sense of why "agents" cost so differently from chat, our explainer on what AI agents actually are still holds up, and the shift to usage-based AI billing is exactly the pressure driving these cheaper models.
π» Hardware
Nvidia wants an AI agent living inside your next laptop. Nvidia revealed RTX Spark, a PC chip it's branding a 1-petaflop "superchip" built to run AI agents locally β with secure sandboxes co-developed with Microsoft, support for 1,000+ games and apps, and 100+ software partners (Adobe, Blender, Riot, Xbox). RTX Spark Windows PCs are due this fall from ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI (Acer and Gigabyte later), as Nvidia eyes a ~$200B CPU market (TechCrunch). Why it matters: it's a bet that the next PC era is local, agent-first computing β less "open the app," more "ask and it's done." The fun part: Jensen Huang's pitch in one line β "With RTX Spark and Microsoft Windows, you ask β and the PC does the work." For why running models on-device is suddenly everywhere, see our piece on on-device AI.
πΈ Big tech & policy
Washington wants an early look before models ship. A new US executive order reportedly asks leading labs β Google, OpenAI, Anthropic, and others β to voluntarily give the government early access to powerful models so it can run cybersecurity tests before public release. Google's Kent Walker called it "an important step forward," and OpenAI's Sam Altman said it "sets the balance right" (Business Today). (It's framed as voluntary; how it works in practice is the open question.)
Anthropic leaned into the security theme too. Anthropic launched a Claude Partner Network (a Services Track and Partner Hub) and published a policy report on a year of mapping AI-enabled cyber threats (Anthropic) β fresh off its confidential IPO filing and a reported ~$965B valuation. The money backdrop: Alphabet reportedly plans to raise up to $80B in equity to fund its AI-infrastructure buildout. The capital arms race isn't slowing down.
π¬ Science & inventions (the cool stuff)
Physicists caught a "ghost" predicted in 1939. Researchers reported the first direct experimental observation of the Migdal effect β a quantum process where a recoiling atomic nucleus kicks out an electron β confirming an 87-year-old prediction and giving dark-matter hunters a powerful new tool (2026 in science). Why it matters: better detection of tiny nuclear recoils sharpens the search for light dark matter β some of the most wanted "missing" stuff in the universe. (Research-stage.)
A chip that does everything with light. Scientists built a single chip that can generate, steer, and read light-based information in one device β a real step toward ultra-fast, energy-efficient photonic computing (ScienceDaily). The fun part: as AI's power bills balloon, "compute with photons, not electrons" stops sounding like sci-fi.
Cheaper green hydrogen, thanks to better chemistry. University of Birmingham researchers developed a perovskite-based catalyst that splits water into hydrogen at much lower temperatures β a possible dent in one of clean energy's stubborn cost problems (ScienceDaily). (Lab-stage, not yet at industrial scale.)
π The one-line recap
Microsoft built its own coder, Nvidia put an agent in your laptop, the US asked for a pre-launch peek, and physicists bagged a 1939 prediction. For the bigger reads, browse our AI & Automation and Future Tech topics.
Tech Pulse runs daily. Numbers and features change fast β we link primary sources so you can verify, and we flag company claims as company claims. See you tomorrow.
