Hacker Newsroom for 28 April recaps major Hacker News stories, moving through microsoft openai split, ai thinking, copilot pricing shift, wall staring focus.

1. Microsoft OpenAI Split

The next story is about Microsoft and OpenAI unwinding one of the defining terms of their partnership, ending the exclusivity and revenue-sharing structure that helped tie Azure to OpenAI's models. The reporting suggests Microsoft can keep hosting OpenAI products, but the arrangement is becoming less locked in as both companies try to widen distribution and keep more control over the economics.

Story link

Hacker News discussion

2. AI Thinking

The next story is an essay arguing that AI should remove drudgery and sharpen judgment, not become a way to outsource thinking altogether. The post says the real divide in software will be between people who use AI to frame problems, weigh tradeoffs, and spot risks, and people who use it to generate polished output without understanding it.

Story link

Hacker News discussion

3. Copilot Pricing Shift

The next story is GitHub's move to usage-based billing for Copilot, a sign that flat fee AI coding plans are getting harder to sustain as model costs rise. The post frames it as a way to map price to actual usage, but on Hacker News many readers immediately read it as the end of subsidized inference for agentic coding tools.

Story link

Hacker News discussion

4. Wall Staring Focus

The next story is a short post about using deliberate boredom, literally sitting and staring at a wall, as a way to recover focus instead of reaching for more stimulation. The author argues that constant screen input keeps people in a state of overload, and that a low input reset can make it easier to return to hard work.

Story link

Hacker News discussion

5. Sub Two Marathon

The next story is about Sabastian Sawe becoming the first athlete to break two hours in a competitive marathon, with multiple runners in the same race also beating the previous world record. The article presents it as a historic result, and the comments quickly widened the story into questions about how much came from athlete quality, how much came from course conditions, and how much came from modern shoe and race technology.

Story link

Hacker News discussion

6. Mercor Voice Breach

The next story is a breach report claiming that roughly four terabytes of voice samples and identity documents tied to tens of thousands of AI contractors were exposed, creating what the author describes as a deepfake ready dataset. The post focuses less on the headline number than on the practical risk of combining audio with ID scans, especially for fraud flows that still trust voiceprints or live call verification.

Story link

Hacker News discussion

That's it for today, I hope this is going to help you build some cool things.