HN Flash

Hacker News. Daily summary. Top 20 stories.

2026-06-29

GLM 5.2 beats Claude in our benchmarks (Semgrep)

Open-weight GLM 5.2 beats Claude Code on an IDOR detection benchmark at one-sixth the cost.

966 pts · 452 comments

Age verification is just a precursor to automated attribution of speech (nonogra.ph)

Age verification laws are a precursor to automated identity-attribution systems for speech.

Divergent views: privacy fears vs. calls for legal engagement and free speech advocacy.

711 pts · 428 comments

HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88 (Dan Unparsed)

HackerRank's open-source ATS returns wildly inconsistent resume scores across repeated runs.

Weighting open source at 65% filters experienced engineers who do not code outside work.

697 pts · 293 comments

EU to legislate about Chat Control behind closed doors (Patrick Breyer)

EU officials push 'Chat Control' mass surveillance, sparking civil rights backlash.

Two camps: those criticizing EU tech regulation as harmful, those defending privacy gains.

686 pts · 407 comments

The KIDS Act would require age checks to get online (Electronic Frontier Foundation)

The KIDS Act would force age checks on all users, harming privacy and speech.

Strong opposition to KIDS Act as surveillance, backed by Meta lobbying.

565 pts · 459 comments

Pollen (CEO Negus-Fancey, CTO Wright) tried to remove article, and Google helped (The Pragmatic Engineer)

A blogger shows Google removed his article about Pollen's collapse after a fake DMCA claim from an uninhabited island.

The thread focuses on how copyright takedown systems are easily abused and rarely penalized.

522 pts · 76 comments

I used Claude Code to get a second opinion on my MRI (antoine.fi)

AI analysis of an MRI found no tear, contradicting a doctor's diagnosis of a Grade III partial-thickness tear.

Radiologists and commenters warn LLMs are unreliable for 3D medical imaging analysis.

496 pts · 629 comments

Professor denounces mass AI fraud on an exam at Brown (Ediciones EL PAÍS S.L.)

A Brown professor caught at least 50 students using AI on a take-home exam.

Two camps: those blaming the exam format versus those blaming students' lack of integrity.

474 pts · 622 comments

Librepods: AirPods liberated (GitHub)

Article body wasn't reachable. HN discussion still summarized.

Split between those valuing ecosystem integration and those finding AirPods mediocre compared to cheaper earbuds.

439 pts · 156 comments

Marfa Public Radio Puts You to Sleep (Marfa Public Radio, radio for a wide range.)

Marfa Public Radio's fall membership drive podcast reads boring work documents to lull you to sleep.

Most find the concept clever; some recommend similar podcasts they prefer.

405 pts · 125 comments

5k menus from the New York Public Library’s Buttolph Collection (1880-1920) (The Pudding)

The NYPL's Buttolph Collection of 5,000 menus from 1880-1920 documents the emergence of modern restaurant dining.

Two camps: those seeing few differences from modern menus, and those noting forgotten staples like tongue and oysters.

392 pts · 101 comments

Flock cameras track more than your license plate, and they're spreading fast (Engadget)

Flock Safety's ALPR cameras proliferate despite security flaws, police misuse, and false arrests.

Two camps: those seeing Flock as a manageable tool and those viewing it as an end-run around surveillance oversight.

372 pts · 291 comments

Historical memory prices 1960-2026 (DAM)

A dataset tracks DRAM, NAND, and HBM prices per GB from 1960 to 2026.

Historical, technical, and contextual debate on memory pricing data and its interpretation.

355 pts · 134 comments

Show HN: Zanagrams (Zanagrams)

A free daily word puzzle where letters and connections disappear as words are found.

337 pts · 90 comments

Michigan bill would bar employers from requiring after-hours coms with workers (CBS Detroit)

Michigan bill would prohibit employers from requiring after-hours work communication.

Two camps: supporters of worker protection versus critics of bureaucratic overreach.

264 pts · 219 comments

The curious case of the disappearing Polish S (2015) (aresluna.org)

Polish letter 'ś' was hard to type because Ctrl+S (save) blocked it.

Two camps debated whether Polish cuisine is closer to German or to Ukrainian/Russian.

248 pts · 100 comments

Ford hired AI and sacked humans. It backfired badly (The Independent)

Ford rehired hundreds of veteran engineers after its AI automation strategy backfired.

Commenters predict more such stories, noting short-term executive incentives and AI's lack of nuanced judgment.

235 pts · 4 comments

A way to exclude sensitive files issue still open for OpenAI Codex (GitHub)

Article body wasn't reachable. HN discussion still summarized.

Debate: use OS permissions vs. add an ignore feature; some note agents find workarounds.

220 pts · 136 comments

Why did this journal retract two 1940s papers by Max Planck? (Ars Technica)

A journal retracted two 1940s Max Planck papers over a likely algorithmic copyright error.

Commenters split: some faulted headline games, others debated self-plagiarism and algorithmic overreach.

179 pts · 16 comments

Tokenmaxxing is dead, long live tokenmaxxing (12 Grams of Carbon)

Tokenmaxxing policies forced AI adoption but disappearing subsidies shift incentives.

Two camps: intentional adoption ploy vs. blindly copying hype with no ROI analysis.

175 pts · 249 comments