Top Stories
GPT-5.6
1321 points · openai.com
OpenAI has pushed out GPT-5.6, and the front page lit up almost instantly. The incremental version bump signals the now-familiar cadence of tightening the same core model rather than a headline leap, but the community is parsing the release notes closely for gains in reasoning, tool use, and cost. As always, the interesting debate isn’t the benchmark table — it’s whether the improvements are real in day-to-day coding and agent workflows or just eval-chasing.
EU Parliament Greenlights Chat Control 1.0
1423 points · patrick-breyer.de
The long-running fight over the EU’s client-side message-scanning proposal took a decisive turn, and HN is treating it as a five-alarm privacy story. MEP Patrick Breyer, a persistent critic, frames the vote as a loss for the very children it claims to protect while opening the door to mass surveillance of private communications. Expect a wave of discussion about end-to-end encryption, the technical impossibility of “safe” backdoors, and what compliance would mean for messaging apps operating in Europe.
John Deere Owners Get the Right to Repair Under FTC Settlement
1328 points · apnews.com
A years-long grievance from farmers reached a milestone: an FTC settlement will force John Deere to let owners and independent shops repair their own equipment. It’s a landmark for the broader right-to-repair movement, which HN has championed for a decade, and a signal that regulators are willing to take on locked-down firmware and dealer-only diagnostics. The thread digs into what the settlement actually mandates versus what the fine print leaves open.
Postgres Rewritten in Rust, Passing 100% of the Regression Tests
693 points · github.com
An ambitious project claims a from-scratch Postgres reimplementation in Rust that now clears the full Postgres regression suite. Passing 100% of those tests is a genuinely impressive bar, and the comments weigh what “compatible” really means in practice — wire protocol, extensions, edge-case behavior — against the memory-safety and maintainability upside of a Rust codebase. A recurring HN favorite: rewrite-it-in-Rust, but this time with receipts.
Show HN: Getting GLM 5.2 Running on My Slow Computer
707 points · github.com
A developer shows off Colibri, a project for running the GLM 5.2 model on modest, non-GPU hardware. The appeal is obvious to the HN crowd: local inference without a datacenter, and the ongoing quest to squeeze large models onto whatever machine you already own. The discussion covers quantization tricks, real-world token throughput, and where the line sits between “usable” and “technically runs.”
Show HN: 18 Words
1010 points · 18words.com
A deceptively simple word game rocketed up the front page, proving again that a tight, well-executed browser toy can out-trend most product launches. Beyond the fun, HN is dissecting the design constraints, the daily-puzzle mechanic, and why minimal, no-signup web games keep finding an audience. Sometimes the best Show HN is the one you can’t stop playing.
Hy3
487 points · hy.tencent.com
Tencent’s research arm dropped Hy3, drawing a crowd curious about how China’s big labs are pushing generative models. The interest is partly technical and partly geopolitical — HN watches these releases as a barometer for how fast the global frontier is spreading beyond the usual US players. Commenters are probing the capabilities, licensing, and how it stacks up against Western counterparts.
Interview with Mitchell Hashimoto About Ghostty and Zig
266 points · alexalejandre.com
The HashiCorp co-founder talks about Ghostty, his fast terminal emulator, and his bet on the Zig language. It’s catnip for the systems-programming crowd: a respected builder explaining why he chose Zig over Rust or C, how he thinks about performance and polish in a terminal, and what it’s like to work on infrastructure software purely for the craft of it. The thread is a good window into current sentiment on Zig’s trajectory.
Also Trending
- Muse Spark 1.1 (380 points) — Meta ships an update to its Muse Spark model with a new API, adding to the week’s steady drumbeat of model releases. ai.meta.com
- The Glass Backbone: Why the Army’s Logistics Will Break in the Next War (384 points) — A sharp military-analysis piece arguing modern logistics networks are dangerously fragile. mwi.westpoint.edu
- A Possible Future for Damn Interesting (288 points) — The beloved long-form site muses publicly about its next chapter and sustainability. damninteresting.com
- Why American Ambulance Rides Are So Expensive (229 points) — A deep dive into the tangled billing and market structure behind sky-high US ambulance bills. davidoks.blog
- Launch HN: Context.dev (YC S26) — API to Get Structured Data from Any Website (98 points) — A fresh YC startup pitches structured web scraping as an API. context.dev