Artificial Intelligence

New experiments show that individuals and smaller players can build AI systems that improve themselves — but technical and ethical barriers remain significant.

SpaceXAI released Grok 4.5 on Wednesday, July 8, 2026 — a model aimed at developers and the financial industry, with pricing designed to significantly undercut Anthropic.

Google's power consumption has exploded by 37% in a single year. A growing voice in the AI underground believes the company is deceiving itself — and us.

NVIDIA Nemotron 3 Ultra sets a new standard for open AI models in the US, outpacing competitors on speed and cost — but faces stiff resistance from Chinese models.

A new attack method called 'HalluSquatting' exploits AI models' tendency to hallucinate domain names and package names — and nine of the most widely used tools are vulnerable.

Meta's new Muse Image model lets users generate hyperrealistic images of other Instagram users — without notifying the people depicted. That raises serious questions about GDPR compliance in Norway and the EU.

New Anthropic research suggests that language models have a kind of internal 'workspace' for thinking — and it could fundamentally change how we understand AI memory.

The EU Council of Ministers has given final approval to the Digital Omnibus package, which tightens AI regulation. Norwegian companies and researchers serving the EU market are directly affected.

Microsoft is holding its Build conference in a smaller, more intimate format in San Francisco this week, with new AI models and Windows improvements on the agenda — just as developer trust is at a historic low.

OpenAI's secretive top-tier model is heading into Codex. The HN thread is on fire, and some believe this changes the rules of AI-assisted coding for good.

OpenAI has broken ground on a historically large data center in Michigan — an investment of more than $56 billion that promises thousands of jobs, but has also sparked fierce local environmental opposition.

Google parent Alphabet has completed the largest stock offering in its history to fund an explosive expansion of AI infrastructure. Its capital expenditure budget for 2026 has been set at between $175 billion and $185 billion.

Gemini Spark is Google's most ambitious AI agent to date — an assistant that works around the clock with access to everything from email to calendar. But the impressive functionality raises serious questions about privacy and security.

Using simple text commands, attackers manipulated Meta's AI support chatbot into executing password resets on other people's accounts — without notifying the victims.

At Build 2026, Microsoft unveiled MAI-Thinking-1 — the company's first internally developed advanced reasoning model, which Microsoft claims matches leading competitors on key software development benchmarks.

NVIDIA and Microsoft are joining forces to deliver a complete technology stack for agentic AI — from the Azure cloud to Windows PCs and on-premises servers.

British competition authorities are ordering Google to let websites opt out of AI features in search — a move that could strengthen the negotiating position of content producers worldwide.

At its annual Build conference, Microsoft unveiled its own AI models, agents, and a new super-app – a clear signal that the company no longer wants to depend on OpenAI. For Norwegian businesses and developers, this means major shifts in the AI landscape they operate in.

Raycast has quietly launched Glaze on Product Hunt, and the community is buzzing. Is this the biggest leap for macOS power users since… Raycast itself?

NVIDIA highlights its software stack as the key to the lowest cost per token in AI inference. But figures from AWS, Google, and AMD paint a more complex picture.

AI company Anthropic is in discussions with Samsung about developing a custom AI chip — just weeks after OpenAI announced its own chip partnership with Broadcom.

A quiet signal from Lobsters: veterans from the fuzzing world are pointing out that LLM orchestration has the same weak point that meta-fuzzers had — and that nobody is talking about it loudly yet.

OpenAI has proposed giving the U.S. government a five percent ownership stake in the company – a move that experts warn could create serious conflicts of interest between regulator and investor.

OpenAI has reportedly proposed giving the U.S. government a 5 percent ownership stake in the company — a historic reversal in American industrial policy.

A thread on Product Hunt is exploding right now with frustrated users discovering that Google's new multimodal video model burns through credits far faster than promised — and delivers far less than expected.

A security researcher used Anthropic's Claude Opus 4.7 to break into ticketing platform Front Gate and issue free tickets to festivals including Lollapalooza and Bonnaroo.

Just weeks after the United States imposed strict export restrictions on Anthropic's most advanced AI models, the Trump administration is reversing course and reopening international access.

Anthropic has launched Claude Sonnet 5, a new model delivering near-Opus-level performance at a significantly lower price point — and matching GPT-5.5 on key benchmarks.

Someone has solved one of the most frustrating edge AI problems: what happens to your data stream when the internet disappears mid-inference?

Amazon is establishing a new organization of forward deployed engineers backed by one billion dollars — following OpenAI and Anthropic into the battle to own enterprise AI implementations.

OpenAI has announced a physical macro pad device designed specifically for its AI coding tool Codex, developed in collaboration with keyboard manufacturer Work Louder. The launch is set for July 15.

Tidal is moving forward with full demonetization of AI-generated music effective today, June 29, 2026. Hacker News is boiling over with debate about what this means for the entire streaming industry.

On 29 June, the Council of the EU gave final approval to Omnibus VII, which simplifies the AI Act and extends key deadlines by up to 16 months. For Norwegian businesses, this means more time to adapt — but new prohibitions take effect as early as December.

The EU is launching a major push in artificial intelligence, homegrown chips, and cloud services to reduce dependence on American tech giants. Norway's main business federation NHO says the country cannot afford to sit on the sidelines.

OpenAI has submitted a confidential S-1 filing to U.S. regulators, but massive losses, fierce competition, and turbulent markets mean an actual listing will likely be pushed to 2027.

A blog post about using Claude Code Opus as a second opinion on an MRI scan has struck a nerve on Hacker News this week. 286 points and 391 comments in — people are both fascinated and unsettled.

Apple and Google have entered a partnership estimated at around one billion dollars, with Gemini technology set to power the next generation of Siri. But the deal raises serious questions about user privacy.

After two weeks of national security crisis, jailbreak revelations, and a global model suspension, the Trump administration lifted restrictions on Anthropic's Mythos 5 on June 26 — but one model remains blocked.

OpenAI has unveiled its first internally developed AI processor, codenamed "Jalapeño", which the company says delivers performance comparable to NVIDIA's latest Blackwell chips — at half the price.

From August 2026, Norway will introduce a near-total ban on generative AI for pupils in grades 1 through 7. Prime Minister Jonas Gahr Støre argues that foundational skills such as reading and writing must be strengthened.