Hacker Newsroom for 29 June recaps major Hacker News stories, moving through glm 5 2 benchmarks, eu chat control, kids act checks, sleep radio podcast.

1. GLM 5 2 Benchmarks

The next story is Semgrep's claim that GLM 5. 2 beats Claude in its cyber benchmarks, with the article arguing that Zhipu AI's open-weight model outperformed Claude Opus 4.

Story link

Hacker News discussion

2. EU Chat Control

The next story is Patrick Breyer's warning that Europe's chat-control fight is back on the table, with the post claiming EU leaders are trying to revive temporary message-scanning rules and rush talks on a permanent version at the same time. The article argues that the worst-case outcome would reintroduce broad scanning of private messages, allow detection orders that are not tightly limited to suspects or court approval, and make age verification a practical requirement for private communication.

Story link

Hacker News discussion

3. KIDS Act Checks

The next story is the Electronic Frontier Foundation's warning that the KIDS Act, presented as child-safety legislation, would effectively require age checks across much of the internet. The article says the bill would push sites toward ID uploads, facial age estimation, and broader moderation because platforms could be liable if they "know or should have known" a user is under 17.

Story link

Hacker News discussion

4. Sleep Radio Podcast

The next story is Marfa Public Radio Puts You to Sleep, a podcast project that turns compliance manuals, ethics codes, and other back-office paperwork into deliberate bedtime audio while doubling as a public-radio fundraiser. Hacker News loved the joke and the craft of it, with many readers saying the concept nails a real sleep niche where human voices and low-stakes material work better than music or white noise.

Story link

Hacker News discussion

5. Claude MRI Opinion

The next story is a post about using Claude Code as a second opinion on a shoulder MRI, where the model reviewed a DICOM export, contradicted the clinic's diagnosis, and left the patient caught between distrust of an aggressive treatment plan and distrust of AI itself. The article matters because it shows both the appeal of AI as a medical advocate and the risk of treating confident model output as clinical evidence, especially on complex imaging it may not reliably interpret.

Story link

Hacker News discussion

6. Robin Williams On AI

The next story is a post arguing that the best answer to AI slop and infinite online advice is the part of human work that comes from lived experience, using Robin Williams's bench-scene monologue in Good Will Hunting as its anchor. The article says language models can imitate knowledge but cannot replace firsthand feeling, memory, judgment, or the artistic choices that come from actually living through love, loss, and risk.

Story link

Hacker News discussion

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